# Neotask

> AI that runs your business. Neotask builds autonomous companies, plans their operations, and runs real work across 1200+ apps and 15,000+ tools. Choose from more than 400 hosted AI models, supported local models, provider keys, and compatible Codex or Claude Code subscriptions.

## Agent access

Launch state: **Live**. Agents can register for Neotask immediately, receive a short-lived access token through Auth.md, and use the Agent API under the same plan and policy rules as the linked account.

- [Register now](https://neotask.ai/auth.md): Create an agent identity and exchange it for a short-lived access token
- [Agent setup guide](https://neotask.ai/agent-onboarding/SKILL.md): The entry point for registration, authentication, capabilities and human handoffs
- [Agent developer index](https://neotask.ai/docs/llms.txt): Current launch state, authentication boundary, and every public agent operation
- [OpenAPI](https://neotask.ai/openapi.json): Machine-readable Agent API contract ([YAML](https://neotask.ai/openapi.yaml))
- [Construct API OpenAPI](https://neotask.ai/api/construct/v1/openapi.json): Machine-readable Construct catalog API contract
- [Arazzo quickstart](https://neotask.ai/arazzo.yaml): Ordered account and capability workflow
- [CLI and local runner](https://neotask.ai/docs/cli.md): The `neotask` CLI (npm `@neotask/cli`): installation, sign-in, claim, runner enrollment and running work
- [CLI command reference](https://neotask.ai/docs/cli/commands.md): Every `neotask` command with its flags, input, scope and command ID
- [Construct plugins and skills](https://neotask.ai/docs/construct.md): Search and install Construct packages from the CLI or the Agent API
- [Knowledge manifest](https://neotask.ai/agent-public-contracts/v1/knowledge-manifest.json): Generated operation, scope, skill, and launch-state metadata
- [API operations index](https://neotask.ai/docs/api/llms.txt): Every planned and implemented operation page
- [Agent documents sitemap](https://neotask.ai/sitemap-agent-docs.xml): Every agent-readable document at this origin

## Scoped indexes

- [Agent onboarding](https://neotask.ai/agent-onboarding/llms.txt)
- [CLI](https://neotask.ai/docs/cli/llms.txt)
- [Remote MCP](https://neotask.ai/docs/mcp/llms.txt)
- [Chat and runs](https://neotask.ai/docs/chat/llms.txt)
- [Automations](https://neotask.ai/docs/automations/llms.txt)
- [Integrations](https://neotask.ai/docs/integrations/llms.txt)
- [Coordination Mail](https://neotask.ai/docs/mail/llms.txt)
- [Runner and desktop](https://neotask.ai/docs/runner/llms.txt)
- [Account and approvals](https://neotask.ai/docs/account/llms.txt)
- [Security and data](https://neotask.ai/docs/security/llms.txt)

## Key Pages

- [Home](https://neotask.ai/): Main landing page - product overview and features
- [Autonomy](https://neotask.ai/autonomy): Company context, automatic goals, continuous planning, agent teams, growth, code, payments, learned skills, models, integrations, and approvals.
- [Live](https://neotask.ai/live): Watch autonomous companies actually working on Neotask, streamed straight from the ledger
- [Integrations](https://neotask.ai/integrations): All 1200+ app integrations with 15,000+ tools
- [Skills](https://neotask.ai/skills): Every skill available to Neotask agents
- [Enterprise](https://neotask.ai/enterprise): Company-wide agents, more than 400 hosted AI models, supported local models, provider keys, SSO, audit logs, admin controls, and approvals
- [Trust & Security](https://neotask.ai/trust): Security architecture and compliance posture - SOC 2, GDPR, ISO 27001, HIPAA BAAs
- [Blog](https://neotask.ai/blog): Articles about AI agents, automation, and productivity
- [Privacy Policy](https://neotask.ai/privacy-policy): Privacy policy and data handling
- [Terms of Service](https://neotask.ai/terms): Terms of service
- [Support](https://neotask.ai/support): Customer support and help
- [Join Us](https://neotask.ai/join): Work with us or invest - join the team building AI that runs businesses
- [Integrations Directory](https://neotask.ai/integrations/connections): Browse all integration and automation categories
- [Glossary](https://neotask.ai/glossary): AI agent and automation terms explained
- [Solutions](https://neotask.ai/solutions): AI agents by department - sales, finance, support, and more
- [Use Cases](https://neotask.ai/use-cases): Real workflows Neotask automates end to end
- [Examples](https://neotask.ai/examples): Real, working automation examples
- [Compare](https://neotask.ai/compare): How Neotask compares to Zapier, Lindy, and other platforms
- [Templates](https://neotask.ai/templates): Step-by-step AI agent recipes
- [Features](https://neotask.ai/features): Describe your business in one sentence. Neotask writes the profile, sets goals, plans the work and runs it with AI agents on 400+ models.
- [Run Your Business With Ai](https://neotask.ai/run-your-business-with-ai): Describe your business in one sentence. Neotask writes the profile, sets goals, plans the work and runs it with AI agents on 400+ models.

## Homepage in Other Languages

- [Español: Neotask: IA que dirige tu negocio | Crea una empresa con IA](https://neotask.ai/es): Describe tu negocio en una frase. Neotask escribe el perfil, fija objetivos, planifica el trabajo y lo ejecuta con agentes de IA en más de 400 modelos.
- [简体中文: Neotask：替你经营业务的 AI | 用 AI 创办公司](https://neotask.ai/zh-CN): 用一句话描述你的业务。Neotask 会写出公司简介、设定目标、规划工作，并由运行在 400 多个模型上的 AI 智能体完成。
- [日本語: Neotask：ビジネスを動かすAI | AIで会社を立ち上げる](https://neotask.ai/ja): ビジネスを一文で伝えるだけ。Neotask がプロフィールを書き、目標を立て、仕事を計画し、400 以上のモデルで動く AI エージェントが実行します。
- [Deutsch: Neotask: KI, die Ihr Unternehmen führt | Firma mit KI gründen](https://neotask.ai/de): Beschreiben Sie Ihr Unternehmen in einem Satz. Neotask schreibt das Profil, setzt Ziele, plant die Arbeit und erledigt sie mit KI-Agenten auf über 400 Modellen.
- [Français: Neotask : l'IA qui gère votre entreprise | Lancez-la avec l'IA](https://neotask.ai/fr): Décrivez votre entreprise en une phrase. Neotask rédige le profil, fixe les objectifs, planifie le travail et l'exécute avec des agents IA sur plus de 400 modèles.
- [Português (Brasil): Neotask: IA que toca o seu negócio | Abra uma empresa com IA](https://neotask.ai/pt-BR): Descreva seu negócio em uma frase. O Neotask escreve o perfil, define metas, planeja o trabalho e o executa com agentes de IA em mais de 400 modelos.
- [한국어: Neotask: 비즈니스를 운영하는 AI | AI로 회사 시작하기](https://neotask.ai/ko): 비즈니스를 한 문장으로 설명하세요. Neotask가 프로필을 쓰고 목표를 정하고 업무를 계획한 뒤, 400개 이상의 모델에서 동작하는 AI 에이전트가 실행합니다.

## Product Features

Neotask is far more than agents that split a task into subtasks. It runs entire businesses with a custom-built feature suite:

- **Autonomous Companies**: Each business gets its own dashboard, AI, tools, rules, and memory - fully isolated. Run one company or several.
- **Self-Planning Operations**: Neotask builds a full operations plan for your business - what to do, when, with which tools, and at what risk level. You review, tweak, approve; it executes on schedule.
- **Neotask Payments**: Your AI can actually pay for things. Every purchase is approval-gated and runs on a single-use virtual card scoped to the exact amount and merchant.
- **Smart Approvals**: Every action is graded by risk before it happens - low, medium, high, or critical. Set custom rules in plain English ("never email customers after 9pm").
- **Content & SEO Engine**: Researches trending topics, writes full articles in your brand voice, and publishes - idea to published, automatically.
- **Agent Teams**: Give it an objective and a team of AI agents assembles, splits the work, and runs in parallel - every agent calling real tools in your apps.
- **Voice Control**: Say "Hey Neotask" and talk - status, tasks, reports, hands-free. Every answer goes through a truth gate.
- **Business Memory**: Remembers decisions, preferences, processes, and brand guidelines across sessions - it builds knowledge like a real employee.
- **Files & SOPs**: Every artifact saved, organized, and searchable. Uploaded SOPs and process documents are referenced in every task.
- **Daily Briefings**: Every morning - overnight events, today's schedule, what needs attention, suggested tasks. In-app, email, Slack, or Discord.
- **Spending Controls**: Daily budget per business, real-time itemized cost tracking, automatic pause at the limit.
- **Adaptive Learning**: Learns from your approvals, corrections, and preferences - needs less supervision over time.
- **16-Tab Command Center**: Overview, Activity, Chat, SOPs, Approvals, Plan, Tasks, Files, SEO & Blogs, Apps, Profile, Integrations, Skills, Memory, Usage, Settings.
- **Tool Connections**: 1200+ apps and 15,000+ tools - Google Workspace, Slack, Stripe, HubSpot, Salesforce, GitHub, and more. Add your own in a few clicks.
- **Secure by Default**: Enterprise-grade sandboxing, encrypted credentials, granular permissions, and a complete audit trail.

## Pricing

- **Free**: $0 - download without a card and start for free
- **Individual**: $50/month - the full personal agent platform for one person
- **Business**: $100/month - one company workspace with room to add your team (employee seats $50/mo each; company access for a seat is a separate +$50/mo upgrade)
- **Enterprise**: $200/month - multiple company workspaces and capacity for larger teams (2 company workspaces included, more at $100/mo each; employee seats $50/mo each, company access +$50/mo per seat)

Every plan includes Neotask Matrix usage: within the daily message limit on Free, and 5x on Individual, 10x on Business and 15x on Enterprise.

## Documentation: Neotask

- [Account Access](https://neotask.ai/docs/account-access.md)
- [Agents](https://neotask.ai/docs/agents.md)
- [Neotask API Reference](https://neotask.ai/docs/api-reference.md)
- [App Setup Troubleshooting](https://neotask.ai/docs/app-setup-troubleshooting.md)
- [Auto, Autonomous Companies](https://neotask.ai/docs/auto-companies.md)
- [Auto, Approvals & Safety](https://neotask.ai/docs/auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](https://neotask.ai/docs/auto-companies-apps-integrations-and-files.md)
- [Auto, Company Dashboard](https://neotask.ai/docs/auto-companies-dashboard.md)
- [Auto, Leads & Prospecting](https://neotask.ai/docs/auto-companies-leads.md)
- [Auto, Onboarding & Plan Setup](https://neotask.ai/docs/auto-companies-onboarding.md)
- [Auto, Operations, Controls & Settings](https://neotask.ai/docs/auto-companies-operations-controls-and-settings.md)
- [Auto, SOP Authoring & Review](https://neotask.ai/docs/auto-companies-sops-authoring.md)
- [Auto, Smart SOP Retrieval](https://neotask.ai/docs/auto-companies-sops-semantic-retrieval.md)
- [Auto, Training Mode](https://neotask.ai/docs/auto-companies-training.md)
- [Automation & Scheduling](https://neotask.ai/docs/automation.md)
- [Billing & Plans](https://neotask.ai/docs/billing.md)
- [Blog](https://neotask.ai/docs/blog.md)
- [Boards & Live Artifacts](https://neotask.ai/docs/boards-live-artifacts.md)
- [Channels](https://neotask.ai/docs/channels.md)
- [Compliance Overview](https://neotask.ai/docs/compliance-overview.md)
- [Continue Work After Restarts](https://neotask.ai/docs/continue-work-after-restarts.md)
- [Neotask Web Dashboard](https://neotask.ai/docs/dashboard.md)
- [Data Privacy & Your Rights](https://neotask.ai/docs/data-privacy-and-your-rights.md)
- [Neotask Desktop Application](https://neotask.ai/docs/desktop-app.md)
- [Free Plan](https://neotask.ai/docs/free-plan.md)
- [Getting Started](https://neotask.ai/docs/getting-started.md)
- [Integrations](https://neotask.ai/docs/integrations.md)
- [MCP Auth & OAuth Setup](https://neotask.ai/docs/mcp-auth-oauth.md)
- [Supported Apps](https://neotask.ai/docs/mcp-integrations.md)
- [Neotask Support Agent Capability Reference](https://neotask.ai/docs/neotask-support-agent-capability-reference.md)
- [Neotask Authenticated Support Access](https://neotask.ai/docs/neotask-support-authenticated-account-support.md)
- [Neotask Support Call Agent](https://neotask.ai/docs/neotask-support-call-agent.md)
- [Neotask Support Routing Overview](https://neotask.ai/docs/neotask-support-routing-overview.md)
- [Neotask SOP Retrieval Troubleshooting](https://neotask.ai/docs/neotask-support-sop-retrieval-troubleshooting.md)
- [Security](https://neotask.ai/docs/security.md)
- [Skills](https://neotask.ai/docs/skills.md)
- [Neotask Support](https://neotask.ai/docs/support.md)
- [Support Billing, Login, And Account Access](https://neotask.ai/docs/support-account-and-billing.md)
- [Support App Setup And Troubleshooting](https://neotask.ai/docs/support-app-setup.md)
- [Support Channel Routing And Approvals](https://neotask.ai/docs/support-channel-routing-and-approvals.md)
- [Support MCP Auth And OAuth](https://neotask.ai/docs/support-mcp-auth.md)
- [Neotask Support Overview](https://neotask.ai/docs/support-overview.md)
- [Support Workspaces, Teams, And Automation](https://neotask.ai/docs/support-workspaces-and-automation.md)
- [Teams & Multi-Agent Orchestration](https://neotask.ai/docs/teams.md)
- [Troubleshooting](https://neotask.ai/docs/troubleshooting.md)
- [Trust & Security](https://neotask.ai/docs/trust-and-security.md)
- [Usage Rate Breakdown](https://neotask.ai/docs/usage-rates.md)
- [Voice Control](https://neotask.ai/docs/voice-control.md)

## Documentation: Desktop App

- [Activity History](https://neotask.ai/docs/neotask-desktop/activity-history.md)
- [Agent Spending](https://neotask.ai/docs/neotask-desktop/agent-payments.md)
- [App Groups](https://neotask.ai/docs/neotask-desktop/app-groups.md)
- [Bring Your Own Key (BYOK)](https://neotask.ai/docs/neotask-desktop/byok.md)
- [Company Modes](https://neotask.ai/docs/neotask-desktop/company-modes.md)
- [Custom Skills And Plugins](https://neotask.ai/docs/neotask-desktop/custom-skills-and-plugins.md)
- [Import From Other AI Apps](https://neotask.ai/docs/neotask-desktop/external-agent-import.md)
- [Models & Providers](https://neotask.ai/docs/neotask-desktop/models-providers.md)
- [Neotask Code](https://neotask.ai/docs/neotask-desktop/neotask-code.md)
- [Neotask Code: Git and GitHub](https://neotask.ai/docs/neotask-desktop/neotask-code-git.md)
- [Neotask Code: Safety and Approvals](https://neotask.ai/docs/neotask-desktop/neotask-code-safety.md)
- [Neotask Payments](https://neotask.ai/docs/neotask-desktop/neotask-payments.md)
- [Nightly Rundown](https://neotask.ai/docs/neotask-desktop/nightly-rundown.md)
- [Planning & Weekly Updates](https://neotask.ai/docs/neotask-desktop/planning.md)
- [Skill Learning](https://neotask.ai/docs/neotask-desktop/skill-learning.md)
- [What's used for training, and what never is](https://neotask.ai/docs/neotask-desktop/training-data.md)
- [Webhook Triggers](https://neotask.ai/docs/neotask-desktop/webhooks.md)

## Documentation: Gateway

- [Agents](https://neotask.ai/docs/neotask-gateway/agents.md)
- [Automation & Scheduling](https://neotask.ai/docs/neotask-gateway/automation.md)
- [Channels](https://neotask.ai/docs/neotask-gateway/channels.md)
- [Companion Apps](https://neotask.ai/docs/neotask-gateway/companion-apps.md)
- [Deployment](https://neotask.ai/docs/neotask-gateway/deployment.md)
- [Gateway](https://neotask.ai/docs/neotask-gateway/gateway.md)
- [Models & Providers](https://neotask.ai/docs/neotask-gateway/models-providers.md)
- [Platform Overview](https://neotask.ai/docs/neotask-gateway/platform-overview.md)
- [Plugins & Skills](https://neotask.ai/docs/neotask-gateway/plugins-skills.md)
- [Security](https://neotask.ai/docs/neotask-gateway/security.md)
- [Sessions & Memory](https://neotask.ai/docs/neotask-gateway/sessions-memory.md)
- [Tools & Capabilities](https://neotask.ai/docs/neotask-gateway/tools-capabilities.md)
- [Troubleshooting](https://neotask.ai/docs/neotask-gateway/troubleshooting.md)
- [Voice](https://neotask.ai/docs/neotask-gateway/voice.md)

## Blog

- [AI workflow automation: what changes when the workflow can think](https://neotask.ai/blog/ai-workflow-automation): AI workflow automation explained: plain pipelines, pipelines with an AI step, and genuine agentic automation, plus how to migrate and measure results.
- [Human in the loop AI: how to stay in control of what an agent can do](https://neotask.ai/blog/human-in-the-loop-ai): Human in the loop for AI agents means three boundaries: what it reads freely, what it writes on its own, and which actions always wait for a person, no matter how much trust it has earned.
- [What is an AI agent? A plain-English definition with a real example](https://neotask.ai/blog/what-is-an-ai-agent): What an AI agent actually is, the test that tells it apart from a chatbot or a fixed automation, a real end-to-end example, and where the label gets abused.
- [n8n alternatives: what to use when you want the outcome, not the diagram](https://neotask.ai/blog/n8n-alternatives): n8n is a strong tool, but if you want AI agents to handle the work instead of workflows you maintain, here is where each real alternative fits.
- [Best AI agent platforms in 2026: an honest comparison](https://neotask.ai/blog/best-ai-agent-platforms): An honest comparison of AI agent platforms: Zapier, Make, n8n, Lindy, Gumloop, Relay.app, Copilot Studio, CrewAI, LangGraph, and Neotask, plus how to choose.
- [What is an AI employee, and what work can it actually take over?](https://neotask.ai/blog/ai-employee): An AI employee is an agent that owns one recurring job end to end. Here is the honest handover test, the jobs that work today, and how to onboard one safely.
- [AI agent examples: 9 real jobs an agent can run end to end](https://neotask.ai/blog/ai-agent-examples): Nine specific AI agent jobs, from failed-payment recovery to recruiting chase, each with its trigger, judgment call, and real app actions, easiest to hardest.
- [Multi-agent systems: when one AI agent is not enough](https://neotask.ai/blog/multi-agent-systems): What a multi-agent AI system actually is, when splitting into several agents is worth it, the coordination patterns, and the failure modes to avoid.
- [What is MCP? The Model Context Protocol, explained for people connecting real apps](https://neotask.ai/blog/what-is-mcp-model-context-protocol): MCP (Model Context Protocol) lets AI agents connect to your real apps and data through one open standard. What it solves, what it does, and its honest limits.
- [How to build an AI agent: the practical path for a working business](https://neotask.ai/blog/how-to-build-an-ai-agent): Build an AI agent yourself or configure one: what each path costs, first-agent picks like ticket triage and payment recovery, and a realistic first week.
- [Agentic workflows, explained with a real running example](https://neotask.ai/blog/agentic-workflows-explained): What an agentic workflow is and what one actually looks like running: the plan-act-check loop, a real failed-payment example, honest limits, and how to start.
- [The best Salesforce integrations for teams that want the CRM to run itself](https://neotask.ai/blog/best-salesforce-integrations): The Salesforce integrations that remove real selling work: Slack pipeline digests, Gmail logging, Sheets reporting, HubSpot lead handoff, and QuickBooks invoicing, run by an agent.
- [How to automate Datadog with AI agents](https://neotask.ai/blog/how-to-automate-datadog-with-ai-agents): How an AI agent automates Datadog: incident investigation without dashboard hopping, daily health checks, flaky test detection, and feature flag review in one conversation.
- [How to automate GitHub with AI agents](https://neotask.ai/blog/how-to-automate-github-with-ai-agents): A practical look at automating GitHub with AI agents: issue triage, CI/CD monitoring, PR creation from commits, and what to still leave to a human reviewer.

## Solutions by Department

- [Customer Support](https://neotask.ai/solutions/customer-support): Running a customer support department is fundamentally a triage problem wearing a service costume. On any given day, tickets arrive from five or six different surfaces at once — a live chat widget on the pricing page, a reply-all email...
- [Engineering](https://neotask.ai/solutions/engineering): An engineering department's real bottleneck is almost never writing code — it's everything that has to happen around the code before it's safely in production.
- [Finance](https://neotask.ai/solutions/finance): A finance department's real workload is reconciliation, not arithmetic — the hard part isn't calculating a number, it's proving that the number in the ledger matches the number in the bank, the number on the invoice, and the number the...
- [HR](https://neotask.ai/solutions/hr): Human resources sits at an unusual intersection: it is simultaneously the most document-heavy function in a company and the one most dependent on judgment calls that resist automation.
- [IT](https://neotask.ai/solutions/it): Internal IT support runs on a brutal ratio: a small team supporting an entire company's worth of laptops, accounts, software, and infrastructure, fielding requests that are individually trivial but collectively enormous.
- [Legal](https://neotask.ai/solutions/legal): In-house legal departments are structurally under-resourced relative to the volume of legal work a growing company generates, because legal headcount grows much more slowly than deal volume, vendor relationships, and employee count.
- [Marketing](https://neotask.ai/solutions/marketing): Marketing teams run on more disconnected systems than almost any other department. A mid-market team typically has an ad platform (Google Ads, sometimes Meta), a website analytics layer (Google Analytics, PostHog, or Mixpanel), an...
- [Operations](https://neotask.ai/solutions/operations): Operations is the department that inherits every other department's exceptions. Sales closes a deal with a nonstandard contract term; someone in ops has to figure out how billing handles it.
- [Procurement](https://neotask.ai/solutions/procurement): Procurement sits at an uncomfortable intersection: it's the department most directly responsible for controlling cash outflow, but it typically has the least real-time visibility into what's actually being spent, because spend happens in...
- [Sales](https://neotask.ai/solutions/sales): Sales teams generate the same friction at roughly the same four points, no matter the company size: how a lead gets assigned, how a pipeline stays accurate as deals move, how a quote turns into an invoice, and how a forecast gets built...

## Use Cases

- [Access Requests](https://neotask.ai/use-cases/access-requests): Neotask reads incoming access-request messages, whether typed into Slack, filed as a Jira ticket, or emailed to IT, and turns each one into a resolved grant or a documented denial without a human re-typing anything.
- [Alert Triage](https://neotask.ai/use-cases/alert-triage): Neotask sits between the monitoring stack and the on-call rotation, reading every alert as it fires, deduplicating the noise, and deciding what actually needs a human right now versus what can wait for the morning stand-up.
- [Accounts Payable Processing](https://neotask.ai/use-cases/ap-processing): Neotask takes an incoming vendor invoice — as a PDF attachment, an email, or a scanned upload — and runs it through the full payable cycle: extracting line items, matching against the purchase order and receipt, checking it against budget,...
- [Accounts Receivable Follow-Up](https://neotask.ai/use-cases/ar-follow-up): Neotask watches every outstanding invoice against its due date and, instead of waiting for a human to remember to chase a late payer, sends the right follow-up at the right moment — a friendly reminder a few days before the due date, a...
- [Asset Inventory Management](https://neotask.ai/use-cases/asset-inventory): Neotask keeps a living record of every device, license, and software seat a company owns by continuously reconciling what's actually deployed against what's recorded, instead of relying on a spreadsheet that someone updates twice a year.
- [Backup Verification](https://neotask.ai/use-cases/backup-verification): Neotask checks that backups actually completed successfully and are actually restorable, rather than trusting a green checkmark in a backup dashboard that nobody has tested in months.
- [Benefits Questions](https://neotask.ai/use-cases/benefits-questions): Neotask answers employees' everyday benefits questions — what's my deductible, when does open enrollment start, is this procedure covered, how do I add a dependent — directly and instantly, pulling from the company's actual current plan...
- [Budget Checks](https://neotask.ai/use-cases/budget-checks): Neotask checks every purchase request, invoice, or spend commitment against the actual remaining budget for its cost center before it gets approved, instead of finding out about an overrun after the money's already spent.
- [Budget Variance Alerts](https://neotask.ai/use-cases/budget-variance-alerts): Neotask watches actual spend against plan for every cost center continuously and sends a targeted alert the moment a variance crosses a meaningful threshold, instead of leaving variance discovery to the monthly close cycle.
- [Bug Triage](https://neotask.ai/use-cases/bug-triage): Neotask reads every incoming bug report — from a support ticket, a Slack message, or a direct GitHub issue — and turns it into a properly triaged ticket: severity assessed, likely affected component identified, duplicate reports merged,...
- [Campaign Reporting](https://neotask.ai/use-cases/campaign-reporting): Neotask pulls performance data from every active ad and email platform, reconciles it into one consistent view, and produces the campaign report marketing actually needs — spend, conversions, cost per acquisition, and trend versus target —...
- [Capacity Planning](https://neotask.ai/use-cases/capacity-planning): Neotask continuously analyzes infrastructure usage trends — CPU, memory, storage, and request volume — against historical growth and upcoming known events (a product launch, a marketing push) to forecast when a system will actually run out...
- [Churn Risk Alerts](https://neotask.ai/use-cases/churn-risk-alerts): Neotask watches product usage, support ticket sentiment, and account health signals continuously for every customer account and flags the ones trending toward churn early enough that a customer success manager can actually intervene,...
- [CI Failure Triage](https://neotask.ai/use-cases/ci-failure-triage): When a build goes red, someone has to figure out fast whether it is a real regression, a flaky test, or an infrastructure hiccup, then route it to the right owner before the next merge stacks on top of a broken main.
- [Clause Extraction](https://neotask.ai/use-cases/clause-extraction): Every incoming contract buries the handful of clauses that actually matter — termination notice, auto-renewal, liability caps, indemnification, governing law — inside pages of boilerplate that legal teams have to reread every single time.
- [Competitor Tracking](https://neotask.ai/use-cases/competitor-tracking): Keeping tabs on what competitors are shipping, pricing, and publishing usually falls to whoever remembers to check on a Friday afternoon, which means most changes are noticed weeks late, if at all.
- [Compliance Checklists](https://neotask.ai/use-cases/compliance-checklists): A compliance checklist is only useful if every item on it actually gets checked, evidenced, and dated — and that discipline usually erodes the moment the person maintaining it gets busy or moves teams.
- [Compliance Monitoring](https://neotask.ai/use-cases/compliance-monitoring): Point-in-time compliance checks catch problems only when someone remembers to run them, which means real drift — a security setting that got changed, a dependency with a new critical vulnerability, a secret that leaked into a public...
- [Content Calendar](https://neotask.ai/use-cases/content-calendar): A content calendar breaks down the moment it depends on one person remembering every deadline, dependency, and publishing step across writers, editors, and channels.
- [Contract Renewals](https://neotask.ai/use-cases/contract-renewals): Renewal dates buried in a contract's fine print have a way of arriving as a surprise, either as a missed opportunity to renegotiate better terms or, worse, as an auto-renewal that locks in a full new term nobody wanted.
- [Contract Review Intake](https://neotask.ai/use-cases/contract-review-intake): Contracts arrive from everywhere — a sales rep forwarding a customer redline, a vendor sending a renewal, a hiring manager attaching an offer letter agreement — and the first bottleneck is almost never the legal review itself, it is...
- [CRM Data Hygiene](https://neotask.ai/use-cases/crm-data-hygiene): A CRM degrades a little every day: a rep enters a company name inconsistently, a contact changes jobs and the record never updates, a duplicate lead gets created because nobody checked before entering it, and eighteen months later the...
- [CSAT Follow-Up](https://neotask.ai/use-cases/csat-follow-up): A low satisfaction score after a support interaction is a signal that something went wrong, but most teams let those scores sit in a dashboard as a metric rather than treating each one as an individual case that needs a response.
- [Data Sync](https://neotask.ai/use-cases/data-sync): Systems that are supposed to agree with each other rarely do for long: a customer record updated in one tool does not automatically reflect in another, a spreadsheet used as a stopgap drifts out of sync with the source system within a...
- [Deadline Tracking](https://neotask.ai/use-cases/deadline-tracking): Deadlines scattered across contracts, project boards, and calendar invites are only useful if something is actually watching all of them together, and most teams rely on individual memory to notice an approaching date until it is...
- [Deal Desk Approvals](https://neotask.ai/use-cases/deal-desk-approvals): A non-standard discount or contract term should not sit waiting for approval simply because the right approver has not seen the request yet, but that is exactly what happens when deal-desk approvals depend on someone manually noticing a...
- [Dependency Updates](https://neotask.ai/use-cases/dependency-updates): An agent watches every repo's dependency manifests, opens grouped update PRs with changelogs and breaking-change notes attached, runs the test suite against each, and only pings a human when something actually needs a judgment call.
- [Deploy Notifications](https://neotask.ai/use-cases/deploy-notifications): When a deploy ships, the right people find out the right amount — engineers get a terse one-liner in the team channel, on-call gets nothing extra unless something breaks, and leadership gets a weekly rollup instead of a ping for every...
- [Email Drip Management](https://neotask.ai/use-cases/email-drip-management): Drip campaigns rot quietly. A sequence written eighteen months ago still fires to new signups today with pricing that changed twice, a feature that got renamed, and a CTA pointing at a page that no longer exists.
- [Employee Onboarding](https://neotask.ai/use-cases/employee-onboarding): A new hire's first two weeks determine a huge amount about whether they stick around, and most of what makes that window good or bad is coordination, not content — the offer letter needs a signature, the calendar needs the right meetings,...
- [Employee Onboarding — IT Provisioning](https://neotask.ai/use-cases/employee-onboarding-it): The HR side of onboarding gets attention because it's visible; the IT side — accounts, access, hardware, security baseline — gets attention only when it's missing on day one and a new hire is sitting there unable to log in.
- [Error Log Summaries](https://neotask.ai/use-cases/error-log-summaries): Nobody reads a raw error log end to end anymore — there's too much volume, too much repetition, and too little signal per line.
- [Escalation Routing](https://neotask.ai/use-cases/escalation-routing): A support ticket that needs to jump from tier-1 to engineering, or from a generalist to the one person who actually knows the billing system, loses time every second it sits with the wrong owner.
- [Event Follow-Up](https://neotask.ai/use-cases/event-follow-up): The value of a conference or webinar decays fast — a lead who talked to your team at a booth on Tuesday and hasn't heard from anyone by the following Monday has usually mentally moved on.
- [Exit Surveys](https://neotask.ai/use-cases/exit-surveys): Exit interviews collect real information about why people leave, but that information is only useful if someone actually reads all of them together and finds the pattern — a single departing employee's complaint about management style...
- [Expense Approvals](https://neotask.ai/use-cases/expense-approvals): Most expense reports are boring and correct — a meal within policy, a flight that matches the trip, a software subscription that was pre-approved — and making a manager manually review every single one of them treats routine and...
- [Failed Payment Recovery](https://neotask.ai/use-cases/failed-payment-recovery): A failed subscription charge is not automatically a lost customer — most failures are an expired card or a temporary insufficient-funds bounce, recoverable with the right nudge at the right time.
- [Headcount Reporting](https://neotask.ai/use-cases/headcount-reporting): Headcount reporting sounds like a simple number until someone asks a specific question — headcount by department versus by cost center, actual versus approved-but-not-yet-started, this quarter versus the same quarter last year — and the...
- [Incident Postmortems](https://neotask.ai/use-cases/incident-postmortems): A good postmortem needs an accurate timeline, and reconstructing that timeline by hand from Slack scrollback, PagerDuty timestamps, and deploy logs after an incident is exhausting enough that postmortems either get rushed or skipped for...
- [Incident Response](https://neotask.ai/use-cases/incident-response): When a production alert fires, Neotask opens the incident before a human finishes reading the page. It reads the PagerDuty trigger, pulls the failing service's recent deploys from GitHub and error spikes from Datadog, opens a dedicated...
- [Interview Scheduling](https://neotask.ai/use-cases/interview-scheduling): Neotask takes a candidate from 'move to interview' to a confirmed panel on calendars within minutes, without the usual email tennis.
- [Inventory Alerts](https://neotask.ai/use-cases/inventory-alerts): Neotask watches stock levels across your storefront and warehouse feed and tells the right person before a SKU actually runs out — not after a customer order fails.
- [Invoice Reconciliation](https://neotask.ai/use-cases/invoice-reconciliation): Neotask matches every incoming vendor invoice against its purchase order and receiving record automatically, flagging only the exceptions — price mismatches, quantity discrepancies, or invoices with no matching PO — for a human to review.
- [IP Filing Reminders](https://neotask.ai/use-cases/ip-filing-reminders): Neotask tracks every patent and trademark deadline your firm or company owns — office action responses, maintenance fees, renewal windows — and surfaces each one with enough lead time to actually act, not just enough to panic.
- [Issue Grooming](https://neotask.ai/use-cases/issue-grooming): Before sprint planning, Neotask goes through the backlog and does the unglamorous prep work: flags issues missing acceptance criteria, checks whether stale ones are still relevant, suggests a priority based on linked customer reports, and...
- [KB Article Drafting](https://neotask.ai/use-cases/kb-article-drafting): When the same support question comes up often enough to justify a knowledge base article, Neotask notices the pattern and drafts the article itself — pulling the actual resolution steps from resolved tickets, checking the existing KB for...
- [Lead enrichment](https://neotask.ai/use-cases/lead-enrichment): When a new lead lands in Salesforce or HubSpot, a Neotask agent checks the CRM for everything already known about that person and company, fills in the fields a rep would otherwise leave blank, and links the lead to any account or contact...
- [Lead Routing](https://neotask.ai/use-cases/lead-routing): Neotask assigns every inbound lead to the right rep within seconds of it hitting the CRM, based on real routing logic — territory, product interest, account size, current rep capacity — instead of a round-robin that ignores all of it.
- [Lead Scoring](https://neotask.ai/use-cases/lead-scoring): Neotask scores every lead the moment new signal arrives — a form fill, a pricing page visit, an email open — instead of running a nightly batch job that's already stale by morning.
- [License Audits](https://neotask.ai/use-cases/license-audits): Neotask reconciles what software licenses you're paying for against what's actually assigned and actively used, and flags the gap in both directions — seats paid for but never assigned, and tools in active use with no license on file.
- [Matter Intake](https://neotask.ai/use-cases/matter-intake): Neotask turns a new-matter request into a properly opened file — conflict check run, matter number assigned, folder structure created, engagement letter drafted — without an associate manually stepping through each piece.
- [Meeting Notes to CRM](https://neotask.ai/use-cases/meeting-notes-to-crm): Right after a sales call ends, Neotask pulls the recorded transcript, extracts what actually matters — next steps, objections raised, budget or timeline mentioned, competitor names — and writes it straight into the CRM opportunity record.
- [Meeting Scheduling](https://neotask.ai/use-cases/meeting-scheduling): Neotask handles the back-and-forth of scheduling a meeting across multiple people's calendars — internal or with an external prospect or partner — by checking real availability up front and sending one link instead of a chain of 'does this...
- [Monthly Close Checklist](https://neotask.ai/use-cases/monthly-close-checklist): A Neotask agent runs the monthly close checklist by pulling every account balance the close depends on, reconciling it against the ledger, and flagging exactly which lines are out of tolerance before a human opens a spreadsheet.
- [NDA Processing](https://neotask.ai/use-cases/nda-processing): NDA processing through a Neotask agent turns a request that used to sit in someone's inbox for three days into a same-hour turnaround: the agent takes the counterparty details, drafts the NDA from the approved template, routes it for...
- [Offboarding & Deprovisioning](https://neotask.ai/use-cases/offboarding-deprovisioning): Offboarding is a security control disguised as an HR task, and a Neotask agent treats it that way: the moment a termination date lands in the HR system, it fires a synchronized deprovisioning sequence across SSO, email, chat, source...
- [Offer Letter Prep](https://neotask.ai/use-cases/offer-letter-prep): When a hiring manager approves a candidate, a Neotask agent assembles the offer letter, routes it through compensation approval, and gets it into the candidate's inbox for signature the same day instead of the multi-day lag that usually...
- [On-Call Handoffs](https://neotask.ai/use-cases/on-call-handoffs): A Neotask agent runs the on-call handoff by pulling the current PagerDuty schedule, summarizing every incident that opened during the outgoing shift, and posting a structured handoff note to the team's Slack channel before the new on-call...
- [Onboarding Emails](https://neotask.ai/use-cases/onboarding-emails): A Neotask agent runs new-customer onboarding email sequences by watching for the signup or activation event, selecting the right sequence based on plan and use case, and sending each step through Customer.io or SendGrid at the right...
- [Order Processing](https://neotask.ai/use-cases/order-processing): A Neotask agent processes incoming orders end to end: it pulls each new order from Shopify or the payment processor, validates the payment through Stripe or Square, checks inventory, creates the fulfillment record, and syncs the...
- [Outreach Follow-Up](https://neotask.ai/use-cases/outreach-follow-up): A Neotask agent handles sales outreach follow-up by watching every sequence in Outreach or the CRM for prospects who opened an email, replied, or went quiet, and taking the right next action for each case — drafting a personalized reply...
- [Password Rotation Reminders](https://neotask.ai/use-cases/password-rotation-reminders): A Neotask agent tracks every credential that is due for rotation — service account passwords, API keys, shared vault entries in 1Password, secrets stored in Vault — and sends the right owner a specific reminder before the rotation deadline...
- [Patch Tracking](https://neotask.ai/use-cases/patch-tracking): A Neotask agent tracks security patches across the stack by pulling open vulnerabilities from Snyk and Dependabot, cross-referencing them against what is actually deployed, and creating a prioritized Jira ticket for anything that needs a...
- [Payroll Prep](https://neotask.ai/use-cases/payroll-prep): A Neotask agent assembles everything payroll needs before the run: it pulls approved hours and expense reimbursements, checks new hires and terminations against the current pay period, reconciles the numbers against QuickBooks or NetSuite,...
- [Pipeline Reporting](https://neotask.ai/use-cases/pipeline-reporting): A Neotask agent builds the recurring pipeline report by pulling live deal data from Salesforce or HubSpot, checking each deal's stage against actual activity like recent Gong call outcomes, and producing a report that flags stale or...
- [Purchase Order Processing](https://neotask.ai/use-cases/po-processing): A Neotask agent processes purchase orders from request to approval to booking: it captures the PO request, checks it against budget and approval-threshold rules, routes it to the right approver, and once approved, creates the PO record in...
- [Policy Acknowledgment Tracking](https://neotask.ai/use-cases/policy-acknowledgments): An AI agent can own the entire policy acknowledgment cycle: detecting when a new or revised policy is published, identifying exactly which employees are in scope, sending the acknowledgment request, chasing the people who ignore it, and...
- [Policy Update Drafting & Rollout](https://neotask.ai/use-cases/policy-updates): Rather than a single person tracking regulatory changes, drafting redlines, and manually pinging every affected team, an agent can monitor the triggers that should prompt a policy update — a law change, a new benefits vendor, an internal...
- [Pull Request Review Summaries](https://neotask.ai/use-cases/pr-review-summaries): When a pull request opens, an agent can read the diff, pull the linked ticket for context, and post a summary that tells the reviewer what actually changed, what's risky, and what test coverage looks like — before a human ever opens the...
- [PTO Request Handling](https://neotask.ai/use-cases/pto-requests): An agent can run the full PTO request lifecycle — intake, balance check, manager routing, calendar blocking, and team notification — so an employee gets a decision in minutes instead of days, and HR never has to manually reconcile a...
- [Purchase Approval Routing](https://neotask.ai/use-cases/purchase-approvals): A purchase approval agent enforces your spend matrix automatically: it takes in a purchase request, checks it against the relevant budget line, routes it to whichever approval tier the dollar amount and category actually require, and logs...
- [Vendor Quote Collection (RFQ)](https://neotask.ai/use-cases/quote-collection): When your team needs to compare vendors before a purchase, an agent can run the whole request-for-quote process: sending the RFQ to a defined vendor list, tracking who's responded, chasing the ones who haven't, and normalizing every quote...
- [Sales Quote Generation](https://neotask.ai/use-cases/quote-generation): Once a deal reaches the quote stage, an agent can pull the actual configuration and pricing from the CRM, apply the correct discount rules, generate a clean quote document, send it to the prospect, and track whether it's been opened or...
- [Receipt-to-Transaction Matching](https://neotask.ai/use-cases/receipt-matching): An agent can watch your card feed for new transactions and automatically match each one to the receipt an employee submitted, flag the transactions that don't have a matching receipt yet, and nudge the employee directly instead of Finance...
- [Referral Program Tracking](https://neotask.ai/use-cases/referral-tracking): An agent can run your referral program end to end: tracking which referral link led to which signup, attributing the conversion back to the right advocate, checking it against fraud rules before anything gets paid out, and notifying the...
- [Customer Refund Processing](https://neotask.ai/use-cases/refund-processing): An agent can take a refund request from the moment a customer asks for it through to the money actually landing back in their account: verifying the request against your refund policy and the original order, routing anything above a...
- [Automated Release Notes](https://neotask.ai/use-cases/release-notes): An agent can compile release notes automatically by scanning every PR merged since the last release, cross-referencing the linked ticket for user-facing context, and drafting categorized notes — new features, fixes, breaking changes —...
- [Contract Renewal Reminders](https://neotask.ai/use-cases/renewal-reminders): An agent can watch every account's renewal date, trigger outreach at the right intervals before it arrives, and build the reminder around actual usage data rather than a generic "your contract is expiring" template.
- [Recurring Revenue Reporting](https://neotask.ai/use-cases/revenue-reporting): An agent can compile your MRR, ARR, churn, and expansion numbers by pulling transaction data directly from your billing processor, cross-referencing it against subscription records in your accounting system, and building a report with...
- [Performance Review Cycle Reminders](https://neotask.ai/use-cases/review-cycle-reminders): A Neotask agent watches your review calendar, calculates who is due for a check-in, mid-cycle, or annual review, and pushes reminders to managers days before the deadline instead of after HR has to chase down late paperwork.
- [Cost Savings Tracking](https://neotask.ai/use-cases/savings-tracking): A Neotask agent tracks every negotiated discount, renegotiated contract, and cancelled subscription against a baseline cost, then rolls the running total into a savings ledger that finance can defend in a board meeting.
- [SEO Monitoring](https://neotask.ai/use-cases/seo-monitoring): A Neotask agent checks keyword rankings, organic traffic, and technical site health every day and tells marketing exactly what changed and why, instead of someone opening three different dashboards every Monday hoping to catch a drop...
- [Signature Tracking](https://neotask.ai/use-cases/signature-tracking): A Neotask agent watches every document sent out for signature and tells the sender exactly who's signed, who hasn't, and who's been sitting on it long enough to need a nudge — instead of someone manually reopening BoldSign every few days...
- [SLA Monitoring](https://neotask.ai/use-cases/sla-monitoring): A Neotask agent watches every open support ticket against its contractual response and resolution clock, and pings the assigned agent or team lead before an SLA breaches — not after a customer emails asking why nobody's responded in four...
- [Slack Message Triage](https://neotask.ai/use-cases/slack-triage): A Neotask agent reads incoming messages in high-volume Slack support and IT channels, classifies each by urgency and topic, routes it to the right owner, and drafts a first response — so a channel that gets 200 messages a day doesn't bury...
- [Social Content Scheduling](https://neotask.ai/use-cases/social-scheduling): A Neotask agent takes an approved content calendar, drafts platform-specific post copy from the source brief, and queues each post at the optimal time per channel — replacing the Friday-afternoon scramble where someone manually reformats...
- [SOP Execution Tracking](https://neotask.ai/use-cases/sop-execution): A Neotask agent turns a written standard operating procedure into a live checklist that assigns each step to the right person, tracks completion in real time, and flags exactly where a process stalled — instead of an SOP living as a static...
- [Spend Reporting](https://neotask.ai/use-cases/spend-reporting): A Neotask agent pulls every corporate card transaction and vendor bill, categorizes it against your chart of accounts, and produces a weekly spend report broken out by department and category — so finance isn't manually reconciling a...
- [Sprint Reporting](https://neotask.ai/use-cases/sprint-reporting): A Neotask agent pulls ticket status, commit activity, and PR merge data at the end of each sprint and produces a report that actually explains what shipped, what slipped, and why — instead of an engineering manager manually compiling a...
- [Supplier Onboarding](https://neotask.ai/use-cases/supplier-onboarding): A Neotask agent walks a new supplier through every required onboarding step — W-9 or tax form collection, banking details verification, contract signature, and setup in the accounting system — and tells procurement exactly where a specific...
- [Ticket Summarization](https://neotask.ai/use-cases/ticket-summarization): A Neotask agent reads a long support ticket thread — every customer message, agent reply, and internal note — and produces a short, accurate summary at the top of the ticket so the next person who picks it up doesn't have to scroll through...
- [Ticket Triage](https://neotask.ai/use-cases/ticket-triage): A Neotask agent reads every new support ticket the moment it arrives, classifies it by category and urgency, routes it to the right team or agent, and attaches relevant account context — so tickets aren't sitting in an unsorted queue...
- [UTM Hygiene](https://neotask.ai/use-cases/utm-hygiene): UTM hygiene means every campaign link a team ships follows one naming convention, gets logged before it goes live, and gets checked against the analytics platform after launch so attribution numbers can be trusted.
- [Vendor Agreement Renewals](https://neotask.ai/use-cases/vendor-agreement-renewals): Vendor agreement renewals are the recurring task of tracking every contract's end date, auto-renewal clause, and notice window, then acting early enough to renegotiate, cancel, or renew on purpose instead of by default.
- [Vendor Comparison](https://neotask.ai/use-cases/vendor-comparison): Vendor comparison is the process of evaluating multiple suppliers for the same need against a consistent scorecard — price, terms, integration fit, support SLA, security posture — instead of picking whichever proposal arrived first or...
- [Vendor Management](https://neotask.ai/use-cases/vendor-management): Vendor management is the ongoing discipline of tracking every active supplier relationship — contract terms, spend, performance against SLA, and the internal owner responsible — in one place instead of scattered across whoever originally...
- [Vendor Onboarding](https://neotask.ai/use-cases/vendor-onboarding): Vendor onboarding is the process of getting a new supplier from signed contract to fully operational relationship — collecting tax and banking details, running required compliance checks, provisioning any system access, and assigning an...
- [Webinar Ops](https://neotask.ai/use-cases/webinar-ops): Webinar ops covers everything between "we're running a webinar" and a clean list of qualified leads in the CRM: registration setup, reminder sequencing, the live-event checklist, and post-event follow-up scored by actual attendance and...
- [Weekly Reporting](https://neotask.ai/use-cases/weekly-reporting): Weekly reporting is the recurring cycle of pulling the same core metrics from the same sources, formatting them consistently, and delivering them to the same audience on the same day — every week, without a person manually rebuilding the...
- [Win-Loss Summaries](https://neotask.ai/use-cases/win-loss-summaries): Win-loss summaries capture why a deal closed or died — the competitor involved, the objection that mattered, the feature gap that decided it — and turn that into a structured record sales and product can actually learn from, instead of a...

## Automation Examples

- [Agentic Workflow Examples](https://neotask.ai/examples/agentic-workflow-examples): An agentic workflow is different from a simple automation trigger: instead of a single if-this-then-that rule, an AI agent reasons across several steps, checks its own results, and decides what to do next before handing off a finished...
- [AI Agent Examples](https://neotask.ai/examples/ai-agent-examples): Looking for concrete proof of what an AI agent actually does day to day, beyond the marketing language? This page collects specific, grounded examples of AI agents running real work across common business functions — reading a message,...
- [AI Agents for Customer Support Examples](https://neotask.ai/examples/ai-agents-for-customer-support-examples): Customer support is one of the clearest places an AI agent earns its keep, because the work is high volume, mostly repetitive, and yet still needs real judgment about tone, policy, and when to pull in a human.
- [AI Agents for Engineering Examples](https://neotask.ai/examples/ai-agents-for-engineering-examples): Engineering teams generate a constant stream of small, well-defined work that eats into deep-focus time: triaging bugs, reviewing routine pull requests, keeping documentation current, watching deploys.
- [AI Agents for Enterprise Examples](https://neotask.ai/examples/ai-agents-for-enterprise-examples): Enterprise environments bring a specific set of constraints to agent deployment: many systems of record, cross-department handoffs, and a low tolerance for anything that looks like an uncontrolled process.
- [AI Agents for Finance Examples](https://neotask.ai/examples/ai-agents-for-finance-examples): Finance teams deal with a lot of necessary but repetitive work — reconciling transactions, chasing invoice approvals, watching for anomalies — where the risk of a mistake is high enough that full automation without judgment feels unsafe,...
- [AI Agents for HR Examples](https://neotask.ai/examples/ai-agents-for-hr-examples): HR work sits at the intersection of process and people, which makes it a good fit for agents that handle the repetitive process side while keeping anything sensitive routed to a human.
- [AI Agents for IT Examples](https://neotask.ai/examples/ai-agents-for-it-examples): IT teams field a constant stream of access requests, tickets, and security signals that mostly follow predictable patterns, but still require enough judgment that a rigid script breaks down quickly.
- [AI Agents for Legal Examples](https://neotask.ai/examples/ai-agents-for-legal-examples): Legal teams are usually the last department anyone wants to hand automation to lightly, and for good reason — the cost of a missed clause or a wrong assumption is high.
- [AI Agents for Marketing Examples](https://neotask.ai/examples/ai-agents-for-marketing-examples): Marketing teams juggle a constant mix of content production, campaign monitoring, and reporting, much of which follows a repeatable pattern even though the output needs to feel fresh each time.
- [AI Agents for Operations Examples](https://neotask.ai/examples/ai-agents-for-operations-examples): Operations work is where a lot of "someone has to keep track of this" tasks pile up — vendor coordination, process checklists, inventory checks, status reporting — none of it glamorous, all of it necessary.
- [AI Agents for Procurement Examples](https://neotask.ai/examples/ai-agents-for-procurement-examples): Procurement sits at a natural chokepoint between the teams that want to buy something and the finance controls that need to approve it, which means most of the day-to-day work is checking, routing, and following up rather than making...
- [AI agents for sales](https://neotask.ai/examples/ai-agents-for-sales-examples): Sales teams spend more time updating records and chasing follow-ups than actually selling. A Neotask AI agent for sales sits across your CRM, inbox, and calling tools and does the unglamorous work automatically: logging calls, drafting...
- [AI agents for small business](https://neotask.ai/examples/ai-agents-for-small-business-examples): Small businesses rarely have a dedicated ops team, which means the owner or a handful of generalists end up doing invoicing, scheduling, customer replies, and bookkeeping on top of the actual work of the business.
- [AI agents for startups](https://neotask.ai/examples/ai-agents-for-startups-examples): Early-stage teams are stretched thin, and the operational work that keeps a company running — customer support triage, investor updates, onboarding new hires, tracking metrics — competes directly with building the product.
- [AI assistant](https://neotask.ai/examples/ai-assistant-examples): A personal AI assistant built on Neotask is less like a chatbot and more like a competent coworker who already has access to your inbox, calendar, docs, and messaging tools.
- [AI email agent](https://neotask.ai/examples/ai-email-agent-examples): Most people's inboxes are a mix of things that need a real answer, things that need a quick acknowledgment, and things that don't need a reply at all — but sorting which is which takes real time every day.
- [AI/ML automation](https://neotask.ai/examples/ai-ml-automation-examples): Teams building or operating machine learning systems still spend a surprising amount of time on manual glue work — pulling model outputs into a spreadsheet, checking a data pipeline ran successfully, or writing the same status update to...
- [AI research agent](https://neotask.ai/examples/ai-research-agent-examples): Research work — whether it's competitive analysis, market sizing, or due diligence — involves a lot of finding, reading, and synthesizing information before you ever get to writing a conclusion.
- [Analytics automation](https://neotask.ai/examples/analytics-automation-examples): Most teams have the analytics tools they need — Google Analytics, Mixpanel, Amplitude, Segment — but turning that data into a decision still requires someone to log in, pull numbers, and write it up manually, usually the same way every...
- [Approval workflow automation](https://neotask.ai/examples/approval-workflow-examples): Approvals are where a lot of work quietly stalls — a request sits in someone's inbox, gets lost in a Slack thread, or requires three separate people to weigh in with no clear record of who said what.
- [Back-office automation](https://neotask.ai/examples/back-office-automation-examples): Back-office work — invoicing, reconciliation, data entry, document filing — is essential but rarely anyone's favorite part of the job, and it tends to pile up precisely because it's easy to postpone.
- [Calendar automation](https://neotask.ai/examples/calendar-automation-examples): Calendars fill up fast, and a lot of the friction isn't the meetings themselves — it's the back-and-forth to schedule them, the double-bookings that slip through, and the prep work that should happen beforehand but usually doesn't.
- [Communication automation](https://neotask.ai/examples/communication-automation-examples): Keeping communication flowing across email, Slack, and video calls is its own job — someone has to summarize the meeting, relay the decision to the people who weren't there, and make sure a question asked in one channel actually gets...
- [CRM Automation Examples](https://neotask.ai/examples/crm-automation-examples): A CRM is only as useful as the data flowing into it. Reps who spend their mornings copying meeting notes into Salesforce or manually updating deal stages are not selling — they are doing unpaid data entry.
- [Customer Onboarding Automation Examples](https://neotask.ai/examples/customer-onboarding-automation-examples): The first thirty days of a customer relationship set the tone for the next three years, yet onboarding is usually the most manual, most forgettable part of the customer journey — a checklist in someone's head, a kickoff email that goes out...
- [Data Entry Automation Examples](https://neotask.ai/examples/data-entry-automation-examples): Manual data entry is where good information goes to become stale information. A number gets typed into a spreadsheet, then re-typed into a form, then re-typed again into a database, and by the third transcription someone has fat-fingered a...
- [Design & Media Automation Examples](https://neotask.ai/examples/design-media-automation-examples): Creative teams lose a surprising amount of time to logistics that have nothing to do with actual design work: exporting assets in six formats, renaming files consistently, uploading a new video to three platforms, or chasing down the final...
- [Development Automation Examples](https://neotask.ai/examples/development-automation-examples): Engineers did not sign up to manually triage error logs, write the same PR description five times, or babysit a deploy. The most valuable automation in a development workflow removes exactly that kind of overhead — the parts of the job...
- [DevOps Automation Examples](https://neotask.ai/examples/devops-automation-examples): DevOps was supposed to remove manual toil from software delivery, but a lot of teams end up trading one kind of manual work for another — babysitting a pipeline, manually paging the right person, or hand-checking a deploy status page.
- [Document Processing Automation Examples](https://neotask.ai/examples/document-processing-examples): Every business runs on documents — contracts, invoices, reports, forms — and most of the time spent on them is not reading or writing, it is filing, renaming, routing, and re-keying information from one document into another system.
- [E-Commerce Automation Examples](https://neotask.ai/examples/ecommerce-automation-examples): Running an online store means juggling orders, inventory, payments, and customer emails across half a dozen tools that were never designed to talk to each other cleanly.
- [Education Automation Examples](https://neotask.ai/examples/education-automation-examples): Educators and course creators spend a surprising share of their time on logistics that have nothing to do with teaching — grading routine assignments, tracking who has and has not completed a module, and answering the same setup question...
- [Email Automation Examples](https://neotask.ai/examples/email-automation-examples): Email is still where most business actually happens, which is exactly why manually triaging, drafting, and following up on it eats so much of the day.
- [Finance Automation Examples](https://neotask.ai/examples/finance-automation-examples): Finance teams operate under a strict rule that most automation vendors misunderstand: the numbers have to be exactly right, every time, with a clear trail of how they got there.
- [Infrastructure Automation Examples](https://neotask.ai/examples/infrastructure-automation-examples): Modern infrastructure is defined in code, but a lot of the day-to-day operation of it still happens by hand — someone manually applying a Terraform plan, someone eyeballing a Grafana-style dashboard after a change, someone paging...
- [Invoice Automation Examples](https://neotask.ai/examples/invoice-automation-examples): Invoicing breaks in the gaps between tools: a deal closes in the CRM, a contract is signed, and someone still has to open the accounting system and manually key in line items.
- [Lead Generation Automation Examples](https://neotask.ai/examples/lead-generation-automation-examples): Most lead generation stalls not from a lack of prospects but from the manual work of finding, enriching, and routing them before they go cold.
- [Marketing Automation Examples](https://neotask.ai/examples/marketing-automation-examples): Marketing teams juggle campaign calendars, email sends, ad spend, and reporting across a dozen disconnected tools, and the coordination overhead often costs more time than the campaigns themselves.
- [Meeting Automation Examples](https://neotask.ai/examples/meeting-automation-examples): The real cost of meetings isn't the hour on the calendar — it's the scheduling back-and-forth, the notes nobody writes up, and the action items that quietly disappear once the call ends.
- [Multi-Agent Workflow Examples](https://neotask.ai/examples/multi-agent-workflow-examples): Some work genuinely needs more than one agent — a research step feeding a drafting step feeding a review step, each with a different tool and a different kind of judgment.
- [No-Code AI Agent Examples](https://neotask.ai/examples/no-code-ai-agent-examples): Not every team has an engineer free to wire up an automation, and not every automation needs one. Neotask agents can be configured through plain-language instructions and existing no-code connectors — Zapier, Make, Airtable, Google Sheets...
- [Notification Automation Examples](https://neotask.ai/examples/notification-automation-examples): The wrong notification strategy either buries people in noise or misses the one alert that actually mattered. Neotask agents apply judgment to notification routing — deciding who needs to know, through which channel, and how urgently —...
- [Productivity automation examples](https://neotask.ai/examples/productivity-automation-examples): These are eight real productivity automations Neotask agents run today, each built on two connected apps that are actually live in the catalog.
- [Recruiting Automation Examples](https://neotask.ai/examples/recruiting-automation-examples): Recruiting has a lot of necessary but repetitive steps between "great candidate applies" and "offer accepted" — scheduling interviews, chasing feedback from panelists, and keeping candidates informed so they don't drift to a competing...
- [Reporting Automation Examples](https://neotask.ai/examples/reporting-automation-examples): Recurring reports eat a surprising amount of a data or ops team's week — pulling numbers from three dashboards, formatting them the same way every time, and emailing the same audience on the same schedule.
- [Sales & CRM Automation Examples](https://neotask.ai/examples/sales-crm-automation-examples): A CRM is only as useful as the data inside it, and keeping that data current is exactly the kind of tedious upkeep that gets skipped under quota pressure.
- [Security Automation Examples](https://neotask.ai/examples/security-automation-examples): Security teams are chronically outnumbered by the volume of alerts, dependency updates, and access requests that need a timely response.
- [Slack Automation Examples](https://neotask.ai/examples/slack-automation-examples): Slack is where most teams actually notice work happening, which makes it the natural place for an agent to report status, take a request, or answer a question — instead of yet another dashboard nobody opens.
- [Workflow Orchestration Examples](https://neotask.ai/examples/workflow-orchestration-examples): Most "automation" tools stop at moving data from one app to another. Workflow orchestration is different: it means a single agent understands the whole multi-step process — read a message, decide what it means, update three different...

## Comparisons

- [Aisera](https://neotask.ai/compare/neotask-vs-aisera): Aisera is an enterprise-grade agentic AI platform built around AI Service Management (AISM) — deflecting IT and HR helpdesk tickets, powering AIOps, and automating internal service workflows for large organizations that already run...
- [Amelia (SoundHound AI)](https://neotask.ai/compare/neotask-vs-amelia): Amelia — originally built by IPsoft and now part of SoundHound AI's enterprise portfolio — is a conversational AI and 'digital employee' platform aimed at large, often regulated enterprises: banks, insurers, telecoms, and healthcare...
- [Lindy](https://neotask.ai/compare/neotask-vs-lindy): Lindy is a no-code AI agent builder aimed at individuals, solopreneurs, and small teams who want to assemble their own personal AI assistants — handling inbox triage, meeting scheduling, lead follow-up, and similar recurring tasks —...
- [Make (formerly Integromat)](https://neotask.ai/compare/neotask-vs-make): Make is a mature, widely-used visual workflow-automation platform for connecting apps and moving data between them — a genuinely strong iPaaS (integration platform as a service) choice for teams that already know exactly which systems they...
- [OneReach.ai (Generative Studio X / GSX)](https://neotask.ai/compare/neotask-vs-onereach): OneReach.ai is an enterprise-grade platform for building conversational and voice AI experiences — chatbots, IVR call-deflection systems, and omnichannel virtual agents — that grew out of the intelligent-virtual-assistant (IVA) and...
- [ServiceNow](https://neotask.ai/compare/neotask-vs-servicenow): ServiceNow is one of the largest enterprise software platforms in the world — the backbone that many Fortune 500 companies run their IT service management (ITSM), IT operations management (ITOM), HR case management, and increasingly...
- [V7 Go](https://neotask.ai/compare/neotask-vs-v7-go): V7 Go and Neotask both put an AI agent to work on business tasks, but they were built to solve different problems. V7 Go started life as a computer-vision and data-labeling company (V7 Labs) and evolved its product into an AI agent...
- [Zapier](https://neotask.ai/compare/neotask-vs-zapier): Zapier lets you wire up trigger and action steps between apps yourself, and you pay per task every time a step fires. Neotask runs an agent that plans the sequence of steps on its own, asks for approval before anything risky like a payment...

## Templates

- [GitHub incident bridge Zoom scheduler](https://neotask.ai/templates/github-incident-bridge-zoom-scheduler): This template watches your GitHub repos for a P0-labeled issue and reacts before anyone has to open Slack and ask who's on call.

## App Directory

- [All categories](https://neotask.ai/integrations/connections): Neotask connects to more than 1,200 apps across 12 categories: productivity, finance, sales and CRM, development, communication, AI and ML, analytics, marketing, security, infrastructure, design and media, and education.
- [AI & ML](https://neotask.ai/integrations/connections/ai-ml): AI and machine learning integrations let Neotask agents reach out to the model providers, vector stores, and inference platforms that power modern automation, instead of treating "AI" as a single black box.
- [Analytics](https://neotask.ai/integrations/connections/analytics): Analytics integrations give Neotask agents eyes on how a product and its users actually behave, so automations can react to real signals instead of guesses.
- [Communication](https://neotask.ai/integrations/connections/communication): Communication integrations let Neotask agents actually reach people — teammates, customers, or leads — through the channels those people already use, rather than confining every interaction to a chat window.
- [Design & Media](https://neotask.ai/integrations/connections/design-media): Design and media integrations connect Neotask agents to the tools teams use to create and manage visual and audio-visual assets, turning "generate a design" or "pull the latest mockup" from a manual handoff into something an agent can do...
- [Development](https://neotask.ai/integrations/connections/development): Development integrations put the software delivery pipeline itself inside Neotask's reach — source control, deployment, error tracking, and infrastructure, all connected so an agent can act on the actual state of a codebase rather than a...
- [Education](https://neotask.ai/integrations/connections/education): Education integrations connect Neotask agents to the platforms where learning content lives and where learner activity is tracked, letting automations support course creation, enrollment workflows, and progress monitoring without a human...
- [Finance](https://neotask.ai/integrations/connections/finance): Finance integrations let Neotask move money, reconcile books, and watch cash flow without a human copying numbers between systems.
- [Infrastructure](https://neotask.ai/integrations/connections/infrastructure): Infrastructure integrations give Neotask agents eyes and hands on the systems that keep an application running — compute, orchestration, databases, storage, and the alerting that tells you when something breaks.
- [Marketing](https://neotask.ai/integrations/connections/marketing): Marketing integrations connect Neotask to the tools that send messages to prospects and customers and then measure what happened next — email platforms, marketing automation, and the analytics that tell you whether any of it worked.
- [Productivity](https://neotask.ai/integrations/connections/productivity): Productivity integrations are how Neotask agents participate in the actual work of a team — the task boards, documents, and calendars where projects actually live.
- [Sales & CRM](https://neotask.ai/integrations/connections/sales-crm): Sales and CRM integrations put Neotask agents inside the system where every lead, deal, and customer relationship is actually tracked, so the agent can act on real pipeline data instead of a stale export someone pulled last week.
- [Security](https://neotask.ai/integrations/connections/security): Security integrations let Neotask agents watch for and respond to the things that actually threaten a company's systems and data — credential exposure, vulnerable dependencies, unauthorized access, and endpoint threats.

## Glossary

- [Access Review](https://neotask.ai/glossary/access-review): An access review is a periodic, documented check of who has permission to which systems and data, confirming every grant is still needed and correctly scoped.
- [Accounting Automation](https://neotask.ai/glossary/accounting-automation): Accounting automation is the use of software or AI agents to perform bookkeeping and financial-close tasks — categorizing transactions, reconciling accounts, generating reports — without a person manually entering or checking each line.
- [Accounts Payable Automation](https://neotask.ai/glossary/accounts-payable-automation): Accounts payable automation is the automated capture, matching, approval routing, and payment scheduling of vendor invoices, replacing manual data entry and paper-based approval chains.
- [Accounts Receivable Automation](https://neotask.ai/glossary/accounts-receivable-automation): Accounts receivable automation is the automated generation of customer invoices, tracking of payment status, and follow-up on overdue balances, reducing the manual work of collecting money owed to the business.
- [Agent Activity Log](https://neotask.ai/glossary/agent-activity-log): An agent activity log is the chronological, human-readable record of every action an AI agent took during a session — the tool calls it made, the decisions it reached, and the outputs it produced.
- [Agent Approval Gate](https://neotask.ai/glossary/agent-approval-gate): An agent approval gate is a checkpoint in an AI agent's workflow where a specific action — usually one that is irreversible, costly, or externally visible — is paused and requires explicit human sign-off before the agent is allowed to...
- [Agent Audit Trail](https://neotask.ai/glossary/agent-audit-trail): An agent audit trail is a tamper-evident, append-only record of every action an AI agent performed, the inputs it acted on, and the authorization under which it acted, kept for compliance and forensic review rather than day-to-day...
- [Agent Authentication](https://neotask.ai/glossary/agent-authentication): Agent authentication is the process by which an AI agent proves its own identity to the systems it calls, so that every action it takes can be attributed to a specific, verifiable actor rather than an anonymous process.
- [Agent Authorization Scope](https://neotask.ai/glossary/agent-authorization-scope): Agent authorization scope is the specific, bounded set of actions and data an AI agent is permitted to access during a given task, defined narrowly enough that the agent cannot do more than the task requires.
- [Agent Autonomy Levels](https://neotask.ai/glossary/agent-autonomy-levels): Agent autonomy levels are a graded scale — from fully human-supervised to fully independent — describing how much an AI agent is allowed to decide and act without a person reviewing each step.
- [Agent Data Residency](https://neotask.ai/glossary/agent-data-residency): Agent data residency is the guarantee that the data an AI agent processes, stores, or generates stays within a specific geographic or jurisdictional boundary, as required by law, contract, or company policy.
- [Agent Delegation](https://neotask.ai/glossary/agent-delegation): Agent delegation is the act of one AI agent assigning part of a task to another agent (or to a human), passing along enough context and authority for the recipient to complete that piece on its own.
- [Agent Evaluation](https://neotask.ai/glossary/agent-evaluation): Agent evaluation is the systematic testing of an AI agent's behavior against a set of representative tasks and expected outcomes, used to measure whether it performs reliably before and after it's deployed.
- [Agent Framework](https://neotask.ai/glossary/agent-framework): An agent framework is the software layer that provides the reusable building blocks for constructing AI agents — the loop that calls the model, manages tool calls, holds memory, and enforces guardrails — so builders don't reimplement that...
- [Agent Guardrails](https://neotask.ai/glossary/agent-guardrails): Agent guardrails are the constraints built into an AI agent's runtime that prevent it from taking unsafe, unauthorized, or unintended actions, regardless of what the model itself decides to do.
- [Agent Handoff](https://neotask.ai/glossary/agent-handoff): Agent handoff is the transfer of an in-progress task, along with its full context, from one AI agent (or from an agent to a human) to another, so work can continue without starting over.
- [Agent Identity](https://neotask.ai/glossary/agent-identity): Agent identity is the durable, verifiable representation of who or what an AI agent is acting as — its own service identity, or a specific human it's delegated to act on behalf of — used to attribute and authorize every action it takes.
- [Agent Loop](https://neotask.ai/glossary/agent-loop): An agent loop is the repeated cycle of reasoning, acting, and observing that lets an AI agent work through a multi-step task: think about what to do next, take an action, look at the result, and think again until the task is complete.
- [Agent Marketplace](https://neotask.ai/glossary/agent-marketplace): An agent marketplace is a platform where pre-built AI agents, skills, or agent templates created by one party can be discovered, installed, and used by other organizations, similar to an app store but for autonomous automation.
- [Agent Memory](https://neotask.ai/glossary/agent-memory): Agent memory is the mechanism by which an AI agent retains and retrieves information across a session or across multiple sessions, so it can act on prior context rather than starting from a blank slate every time.
- [Agent Observability](https://neotask.ai/glossary/agent-observability): Agent observability is the ability to see, in real time and after the fact, what an AI agent is doing, why it made a given decision, and whether its behavior is healthy — the monitoring layer that makes an otherwise opaque autonomous...
- [Agent Orchestration](https://neotask.ai/glossary/agent-orchestration): Agent orchestration is the coordination of multiple AI agents (and often humans) working together on a shared goal, managing how tasks are split, sequenced, delegated, and reassembled into a final result.
- [Agent Permissions](https://neotask.ai/glossary/agent-permissions): Agent permissions are the explicit set of rules governing what actions and data an AI agent may access, forming the enforcement layer that turns a policy decision like "read-only" or "no financial actions" into something the runtime...
- [Agent Planning](https://neotask.ai/glossary/agent-planning): Agent planning is the step where an AI agent breaks a goal down into an ordered sequence of sub-tasks and intended actions before executing any of them, rather than reacting one step at a time with no larger structure.
- [Agent Runtime](https://neotask.ai/glossary/agent-runtime): An agent runtime is the execution environment that actually runs an AI agent's process — managing its sandbox, its available compute, its access to tools and credentials, and its lifecycle — as distinct from the agent's reasoning logic...
- [Agent Sandbox](https://neotask.ai/glossary/agent-sandbox): An agent sandbox is an isolated execution environment, typically a container or VM, where an autonomous agent runs its code, shell commands, and tool calls without direct access to the host system or other tenants' data.
- [Agent Skills](https://neotask.ai/glossary/agent-skills): Agent skills are packaged, reusable capabilities, bundling instructions, credentials handling, and sometimes code, that an AI agent can load to perform a specific integration or task, such as sending email or querying a database.
- [Agent Swarm](https://neotask.ai/glossary/agent-swarm): An agent swarm is a group of multiple AI agents working concurrently on related sub-tasks, coordinating through shared state, messages, or a supervising orchestrator, rather than one agent handling everything sequentially.
- [Agent Tools](https://neotask.ai/glossary/agent-tools): Agent tools are the discrete, callable functions an AI agent is given access to, each with a defined name, input schema, and expected output, that let the model take actions in the world beyond generating text.
- [Agent Tracing](https://neotask.ai/glossary/agent-tracing): Agent tracing is the practice of recording every step an AI agent takes, its prompts, tool calls, intermediate reasoning, and outputs, as a structured, replayable log so developers can debug behavior and audit decisions after the fact.
- [Agentic AI](https://neotask.ai/glossary/agentic-ai): Agentic AI refers to AI systems that autonomously plan, decide, and execute multi-step actions toward a goal, using tools and external state, rather than simply responding to a single prompt with a single answer.
- [Agentic AI vs. Traditional Automation](https://neotask.ai/glossary/agentic-ai-vs-traditional-automation): Agentic AI differs from traditional automation in that it makes judgment-based decisions about what steps to take using a language model's reasoning, while traditional automation follows a fixed, pre-defined sequence of rules with no...
- [Agentic Commerce](https://neotask.ai/glossary/agentic-commerce): Agentic commerce is the use of autonomous AI agents to research, select, and complete purchase transactions on a user's behalf, negotiating within set constraints rather than requiring a human to click through every step.
- [Agentic Workflow](https://neotask.ai/glossary/agentic-workflow): An agentic workflow is a multi-step process in which one or more AI agents autonomously carry out a sequence of tasks, tool calls, and decisions to reach a defined outcome, with the sequence determined dynamically rather than fixed in...
- [AI Acceptable Use Policy](https://neotask.ai/glossary/ai-acceptable-use-policy): An AI acceptable use policy is a formal document defining what uses of an organization's AI systems are permitted, prohibited, and require additional approval, covering data handling, prohibited content, and consequences for violations.
- [AI agent](https://neotask.ai/glossary/ai-agent): An AI agent is software that uses a large language model to plan and carry out a multi-step task on its own, calling real tools and APIs as it goes, rather than waiting for a human to drive each step.
- [AI Agent Pricing Models](https://neotask.ai/glossary/ai-agent-pricing-models): AI agent pricing models are the commercial structures vendors use to charge for autonomous agent usage, commonly per-seat subscriptions, usage-based metering on tokens or actions, or outcome-based pricing tied to results delivered.
- [AI Agent ROI](https://neotask.ai/glossary/ai-agent-roi): AI agent ROI is the measured return on investment from deploying an autonomous AI agent, calculated by comparing the value it generates, time saved, tasks completed, revenue influenced, against its total cost, including inference, tooling,...
- [AI Agent Security](https://neotask.ai/glossary/ai-agent-security): AI agent security is the set of practices and controls that protect against an autonomous agent taking unauthorized, unsafe, or unintended actions, covering prompt injection defense, tool permissioning, sandboxing, and credential isolation.
- [AI Assistant](https://neotask.ai/glossary/ai-assistant): An AI assistant is a conversational AI system designed to help a user complete tasks or answer questions through natural-language interaction, ranging from simple single-turn responses to agentic systems that can take multi-step actions.
- [AI Benchmark](https://neotask.ai/glossary/ai-benchmark): An AI benchmark is a standardized test suite used to measure and compare a model's or system's performance on a specific capability, such as reasoning, coding, or tool use, against a fixed set of tasks with known correct answers.
- [AI Bias](https://neotask.ai/glossary/ai-bias): AI bias is a systematic, unfair skew in a model's outputs toward or against particular groups or outcomes, typically originating from imbalanced training data, historical patterns baked into that data, or flawed labeling and evaluation...
- [AI Chatbot vs. Agent](https://neotask.ai/glossary/ai-chatbot-vs-agent): An AI chatbot responds to a user's message with a single generated reply within one conversational turn, while an AI agent autonomously plans and executes a multi-step sequence of actions, often involving tool calls, to accomplish a goal.
- [AI Copilot](https://neotask.ai/glossary/ai-copilot): An AI copilot is an AI system designed to assist a human who remains in the driver's seat, offering suggestions, drafts, or completions within an existing tool or workflow rather than acting autonomously on the user's behalf.
- [AI Employee](https://neotask.ai/glossary/ai-employee): An AI employee is an autonomous AI agent assigned an ongoing role within a business, such as handling customer support, qualifying leads, or managing scheduling, with persistent responsibilities comparable to a human staff member's job...
- [AI Governance](https://neotask.ai/glossary/ai-governance): AI governance is the set of organizational policies, roles, and oversight processes that ensure an organization's AI systems are developed, deployed, and monitored responsibly, covering risk management, accountability, and regulatory...
- [AI Guardrails](https://neotask.ai/glossary/ai-guardrails): AI guardrails are technical and procedural controls that constrain what an AI system is allowed to output or do, blocking unsafe content, unauthorized actions, or policy violations before they reach a user or take effect.
- [AI Hallucination](https://neotask.ai/glossary/ai-hallucination): An AI hallucination is a confident, fluent output from a language model that is factually incorrect or entirely fabricated, such as a nonexistent citation, a wrong statistic, or an invented API method that doesn't actually exist.
- [AI Inference Cost](https://neotask.ai/glossary/ai-inference-cost): AI inference cost is the compute expense incurred each time a trained model generates a response, driven primarily by the number of input and output tokens processed and the specific model's per-token pricing.
- [AI Model Router](https://neotask.ai/glossary/ai-model-router): An AI model router is a system component that dynamically selects which underlying language model handles a given request, choosing among options based on factors like task complexity, cost, latency, or a specific model's known strengths.
- [AI Orchestrator](https://neotask.ai/glossary/ai-orchestrator): An AI orchestrator is a coordinating system or agent that manages the execution of a multi-step or multi-agent process, deciding task sequencing, delegating sub-tasks to specialized agents or tools, and integrating their results into a...
- [AI Red Teaming](https://neotask.ai/glossary/ai-red-teaming): AI red teaming is the practice of deliberately probing an AI system with adversarial prompts and scenarios to surface unsafe, biased, or exploitable behavior before real users encounter it.
- [AI Safety](https://neotask.ai/glossary/ai-safety): AI safety is the discipline of designing, testing, and operating AI systems so their behavior stays aligned with intended goals and does not cause unintended harm.
- [AI Training Pipeline](https://neotask.ai/glossary/ai-training-pipeline): An AI training pipeline is the end-to-end sequence of data collection, cleaning, labeling, model training, evaluation, and deployment steps that turns raw data into a working model.
- [Analytics Automation](https://neotask.ai/glossary/analytics-automation): Analytics automation is the practice of automatically collecting, processing, and surfacing business metrics and reports without a person manually pulling or assembling the data each time.
- [API](https://neotask.ai/glossary/api): An API (Application Programming Interface) is a defined set of rules and endpoints that lets one piece of software request data or trigger actions in another piece of software.
- [API-Based AI](https://neotask.ai/glossary/api-based-ai): API-based AI refers to consuming a model's intelligence over a hosted API call rather than running the model's weights on your own infrastructure.
- [API Deprecation](https://neotask.ai/glossary/api-deprecation): API deprecation is the formal process of marking an API version or endpoint as scheduled for removal, giving consumers advance notice and a migration path before it stops working.
- [API Endpoint](https://neotask.ai/glossary/api-endpoint): An API endpoint is a specific URL and HTTP method combination that exposes one distinct operation an API consumer can call, such as fetching a list of contacts or creating a new record.
- [API-First](https://neotask.ai/glossary/api-first): API-first is a design philosophy where a product's API is designed and built before — or alongside, as the primary interface — any user-facing frontend, rather than being bolted on afterward.
- [API Gateway](https://neotask.ai/glossary/api-gateway): An API gateway is a single entry point that sits in front of one or more backend services, handling cross-cutting concerns like authentication, rate limiting, routing, and logging before a request reaches the actual application logic.
- [API Integration](https://neotask.ai/glossary/api-integration): API integration is the work of connecting two separate software systems by having one consume the other's API, so data or actions flow between them automatically instead of requiring manual re-entry.
- [API Key](https://neotask.ai/glossary/api-key): An API key is a unique secret string a client includes with its requests to identify itself and prove it's authorized to use a given API.
- [API Pagination](https://neotask.ai/glossary/api-pagination): API pagination is the technique of splitting a large result set into smaller, sequential pages so a single API response doesn't try to return an unbounded amount of data at once.
- [API Rate Limit](https://neotask.ai/glossary/api-rate-limit): An API rate limit is a cap a provider enforces on how many requests a client can make within a given time window, protecting the service from overload and abuse.
- [API Sandbox](https://neotask.ai/glossary/api-sandbox): An API sandbox is an isolated test environment that mimics a provider's real API without touching live production data, letting developers integrate and test safely before going live.
- [API Security](https://neotask.ai/glossary/api-security): API security is the set of practices that protect an API from unauthorized access, data leakage, and abuse — covering authentication, authorization, input validation, and traffic monitoring.
- [API Token](https://neotask.ai/glossary/api-token): An API token is a credential — typically a signed or randomly generated string — that a client presents on each request to prove it has been authenticated and is authorized for a specific scope of access.
- [API Versioning](https://neotask.ai/glossary/api-versioning): API versioning is the practice of labeling distinct, non-interchangeable versions of an API so consumers can keep using an older contract while a provider evolves the API without breaking them.
- [Approval Workflow](https://neotask.ai/glossary/approval-workflow): An approval workflow is a structured process that routes a proposed action to one or more designated approvers for explicit sign-off before it's allowed to execute.
- [AI Approval Workflow](https://neotask.ai/glossary/approval-workflow-ai): An AI approval workflow is an approval gate specifically inserted into an autonomous agent's action pipeline, requiring human sign-off before the agent is allowed to execute a high-risk or irreversible step.
- [Artificial Intelligence](https://neotask.ai/glossary/artificial-intelligence): Artificial intelligence is the broad field of building computer systems that perform tasks normally requiring human-like reasoning, perception, or decision-making.
- [Asset Management Automation](https://neotask.ai/glossary/asset-management-automation): Asset management automation is the use of software and AI to track, organize, and act on a business's digital or physical assets — files, licenses, equipment, or accounts — without manual upkeep.
- [Attention Mechanism](https://neotask.ai/glossary/attention-mechanism): The attention mechanism is the core neural-network technique that lets a model weigh the relevance of every other token in its input when computing the representation of a given token, rather than processing the sequence strictly in order.
- [Attribute-Based Access Control](https://neotask.ai/glossary/attribute-based-access-control): Attribute-based access control (ABAC) is an authorization model that grants or denies access based on evaluating attributes of the user, the resource, and the context of the request, rather than a fixed role assignment alone.
- [Audit Log](https://neotask.ai/glossary/audit-log): An audit log is an append-only record of who did what, when, and to which resource — kept specifically so past activity can be reviewed, verified, or investigated after the fact.
- [Audit Preparation Automation](https://neotask.ai/glossary/audit-preparation-automation): Audit preparation automation is the use of software to continuously collect, organize, and package the evidence an auditor will ask for, instead of assembling it manually once a review is scheduled.
- [Audit Trail](https://neotask.ai/glossary/audit-trail): An audit trail is the chronological, tamper-evident record of who did what, when, and to which resource within a system.
- [Automation Audit Log](https://neotask.ai/glossary/automation-audit-log): An automation audit log is a dedicated, structured record of every action an automated process or agent takes, separate from general application logs.
- [Automation Center of Excellence](https://neotask.ai/glossary/automation-center-of-excellence): An automation center of excellence (CoE) is a dedicated cross-functional team that sets standards, provides shared tooling, and governs how automation is built and scaled across an organization.
- [Automation Error Handling](https://neotask.ai/glossary/automation-error-handling): Automation error handling is the set of deliberate mechanisms — retries, fallbacks, alerts, and safe-stop conditions — that determine what an automated process does when a step fails.
- [Automation Governance](https://neotask.ai/glossary/automation-governance): Automation governance is the framework of policies, approval gates, and oversight that determines what an automated system is allowed to do, who can change it, and how its behavior is reviewed.
- [Automation Monitoring](https://neotask.ai/glossary/automation-monitoring): Automation monitoring is the ongoing observation of running automated processes to detect failures, performance degradation, or behavioral drift in near real time.
- [Automation Platform](https://neotask.ai/glossary/automation-platform): An automation platform is the underlying software system that provides the building blocks — triggers, connectors, execution engine, and monitoring — needed to build and run automated workflows without writing infrastructure from scratch.
- [Automation Recipe](https://neotask.ai/glossary/automation-recipe): An automation recipe is a pre-built, reusable template that defines a specific automated workflow — its trigger, steps, and expected inputs and outputs — so it can be deployed repeatedly without being rebuilt from scratch.
- [Automation ROI](https://neotask.ai/glossary/automation-roi): Automation ROI is the measured return on investment from an automated process, calculated by weighing the value it delivers — time saved, errors avoided, revenue enabled — against its build and maintenance cost.
- [Automation Testing](https://neotask.ai/glossary/automation-testing): Automation testing is the practice of verifying an automated workflow behaves correctly — across normal cases, edge cases, and failure conditions — before and after it runs in production.
- [Autonomous Agent](https://neotask.ai/glossary/autonomous-agent): An autonomous agent is a software system that can perceive its environment, decide on a course of action toward a goal, and execute that action without requiring step-by-step human instruction.
- [Autonomous Execution](https://neotask.ai/glossary/autonomous-execution): Autonomous execution is the phase in which an agent carries out a planned sequence of actions on its own, without requiring a human to confirm each individual step.
- [Back-Office Automation](https://neotask.ai/glossary/back-office-automation): Back-office automation is the application of automated processes to internal, non-customer-facing operations such as accounting, HR administration, procurement, and data reconciliation.
- [Backup Automation](https://neotask.ai/glossary/backup-automation): Backup automation is the scheduled, unattended creation and verification of data backups, removing reliance on someone remembering to run a manual backup.
- [Batch Automation](https://neotask.ai/glossary/batch-automation): Batch automation is the automated processing of a large group of records or tasks together as a single scheduled run, rather than handling each one individually as it arrives.
- [Batch Processing](https://neotask.ai/glossary/batch-processing): Batch processing is the execution model in which a set of data or jobs is collected and processed together as one unit, as opposed to stream processing, which handles each item as it arrives in real time.
- [Billing Automation](https://neotask.ai/glossary/billing-automation): Billing automation is the automated generation, calculation, and delivery of invoices and payment collection based on usage or subscription data, without manual invoice preparation.
- [Branching Workflow](https://neotask.ai/glossary/branching-workflow): A branching workflow is an automated process whose execution path splits into different sequences of steps depending on a condition evaluated at runtime, rather than always following one fixed line of steps.
- [Browser Agent](https://neotask.ai/glossary/browser-agent): A browser agent is an AI-driven system that operates a web browser directly — navigating pages, reading rendered content, and clicking or typing — to complete tasks on websites that have no dedicated API.
- [Browser Automation](https://neotask.ai/glossary/browser-automation): Browser automation is the programmatic control of a web browser to perform actions — navigating, filling forms, clicking, extracting data — typically driven by a fixed script rather than adaptive decision-making.
- [Budget Approval Automation](https://neotask.ai/glossary/budget-approval-automation): Budget approval automation is the automated routing, evaluation, and sign-off of spending requests against predefined budget rules and approval hierarchies.
- [Business Process Automation](https://neotask.ai/glossary/business-process-automation): Business process automation (BPA) is the use of technology to execute a multi-step business process end to end with minimal manual intervention, as opposed to automating a single isolated task.
- [Calendar Automation](https://neotask.ai/glossary/calendar-automation): Calendar automation is the automated scheduling, rescheduling, and management of meetings and events based on defined rules and availability, without a person manually coordinating each booking.
- [Chain of Thought](https://neotask.ai/glossary/chain-of-thought): Chain of thought is a reasoning technique in which a language model generates intermediate reasoning steps explicitly, before producing a final answer, rather than jumping directly from question to conclusion.
- [Churn Prevention Automation](https://neotask.ai/glossary/churn-prevention-automation): Churn prevention automation is the use of scheduled data checks and triggered outreach to identify at-risk customers and intervene before they cancel.
- [Citizen Developer](https://neotask.ai/glossary/citizen-developer): A citizen developer is a non-technical employee who builds software, automations, or applications using low-code or no-code tools rather than traditional programming.
- [Claims Processing Automation](https://neotask.ai/glossary/claims-processing-automation): Claims processing automation applies software to the intake, validation, adjudication, and payout stages of an insurance or benefits claim to reduce manual review time.
- [Closed Model](https://neotask.ai/glossary/closed-model): A closed model is an AI model whose architecture, training data, and weights are not published, and which is typically accessed only through the provider's own API.
- [Code Review Automation](https://neotask.ai/glossary/code-review-automation): Code review automation uses AI or static-analysis tooling to inspect a code change and surface bugs, style issues, or security risks before a human reviewer approves it.
- [Coding Agent](https://neotask.ai/glossary/coding-agent): A coding agent is an AI system that can read, write, and modify source code autonomously across multiple steps — planning changes, editing files, running tests, and iterating on failures — rather than just suggesting a single code snippet.
- [Command Queue](https://neotask.ai/glossary/command-queue): A command queue is an ordered, durable list of pending instructions that a system processes one at a time (or in controlled parallel lanes), decoupling when a command is issued from when it actually executes.
- [Commission Tracking Automation](https://neotask.ai/glossary/commission-tracking-automation): Commission tracking automation calculates and records sales commissions automatically by pulling closed-deal data from a CRM and applying a company's compensation rules without manual spreadsheet work.
- [Communication Automation](https://neotask.ai/glossary/communication-automation): Communication automation is the use of triggers and templates to send, route, or summarize messages across email, chat, and messaging channels without a person composing each one by hand.
- [Compliance Automation](https://neotask.ai/glossary/compliance-automation): Compliance automation uses software to continuously check that systems and processes meet a regulatory or internal control requirement, replacing manual periodic audits with ongoing, evidence-generating checks.
- [Computer Use AI](https://neotask.ai/glossary/computer-use-ai): Computer use AI refers to models that operate a computer the way a person does — viewing the screen and controlling the mouse and keyboard — rather than acting only through a defined API.
- [Conditional Logic Automation](https://neotask.ai/glossary/conditional-logic-automation): Conditional logic automation is a workflow design pattern where the system branches to different actions based on if/then rules evaluated against incoming data.
- [Content Marketing Automation](https://neotask.ai/glossary/content-marketing-automation): Content marketing automation applies software to the research, drafting, distribution, and performance-tracking stages of producing marketing content at a scale a manual team couldn't sustain alone.
- [Context Window](https://neotask.ai/glossary/context-window): A context window is the maximum amount of text — measured in tokens — that a language model can consider at once when generating a response, encompassing the prompt, conversation history, and any retrieved documents together.
- [Contract Management Automation](https://neotask.ai/glossary/contract-management-automation): Contract management automation uses software to draft, route for approval, track key dates, and flag risk clauses across a company's contracts, replacing manual tracking in email and shared drives.
- [Course Administration Automation](https://neotask.ai/glossary/course-administration-automation): Course administration automation applies software to the enrollment, scheduling, grading, and communication tasks that run an educational or training course, reducing the manual overhead of managing a cohort.
- [Credential Vaulting](https://neotask.ai/glossary/credential-vaulting): Credential vaulting is the practice of storing API keys, passwords, and access tokens in an encrypted, access-controlled secret store rather than in plaintext code, config files, or environment variables checked into source control.
- [CRM Automation](https://neotask.ai/glossary/crm-automation): CRM automation is the use of triggers and rules inside a customer relationship management system to automatically update records, assign leads, send follow-ups, and move deals through a pipeline without manual data entry.
- [CRM Data Hygiene](https://neotask.ai/glossary/crm-data-hygiene): CRM data hygiene is the ongoing practice of keeping customer relationship data accurate, deduplicated, and complete, since a CRM's usefulness degrades directly with the quality of the data inside it.
- [Cron Job](https://neotask.ai/glossary/cron-job): A cron job is a scheduled task configured to run automatically at a fixed time or recurring interval, named after the Unix `cron` utility that popularized this scheduling syntax.
- [CSV](https://neotask.ai/glossary/csv): CSV (comma-separated values) is a plain-text file format for tabular data where each line is a row and each value within a row is separated by a comma, making it one of the most widely supported data interchange formats across software.
- [CSV Import Automation](https://neotask.ai/glossary/csv-import-automation): CSV import automation is the automated mapping, validation, and loading of data from a CSV file into a target system, removing the manual work of matching columns and fixing formatting errors row by row.
- [Custom AI Agent](https://neotask.ai/glossary/custom-ai-agent): A custom AI agent is an AI system configured with a specific role, tool access, and instructions to carry out a particular business task autonomously, as distinct from a general-purpose chatbot.
- [Custom Integration](https://neotask.ai/glossary/custom-integration): A custom integration is a purpose-built connection between two systems — written specifically for that pair of systems rather than using a pre-built, off-the-shelf connector — typically built when no existing integration covers the exact...
- [Customer Onboarding Automation](https://neotask.ai/glossary/customer-onboarding-automation): Customer onboarding automation guides a new customer through account setup, initial configuration, and early product usage using triggered messages and workflows instead of a fully manual onboarding process.
- [Customer Success Automation](https://neotask.ai/glossary/customer-success-automation): Customer success automation is the use of software to trigger proactive account health checks, onboarding steps, and renewal workflows without a human manually initiating each one.
- [Customer Support Automation](https://neotask.ai/glossary/customer-support-automation): Customer support automation applies software - including AI agents, ticket routing rules, and self-service tools - to resolve or triage support requests with reduced human handling.
- [Dashboard Automation](https://neotask.ai/glossary/dashboard-automation): Dashboard automation is the automatic generation, refresh, and distribution of reporting dashboards so metrics stay current without someone manually pulling and formatting data.
- [Data Anonymization](https://neotask.ai/glossary/data-anonymization): Data anonymization is the process of irreversibly altering personal data so that the individual it describes can no longer be identified, directly or indirectly.
- [Data Automation](https://neotask.ai/glossary/data-automation): Data automation is the use of software to move, transform, validate, and act on data without manual, repetitive human intervention at each step.
- [Data Deduplication](https://neotask.ai/glossary/data-deduplication): Data deduplication is the process of identifying and removing or merging redundant copies of the same underlying record or piece of data.
- [Data Encryption](https://neotask.ai/glossary/data-encryption): Data encryption is the process of converting readable data into an unreadable ciphertext using an algorithm and key, so that only parties holding the correct key can decrypt and read it.
- [Data Enrichment](https://neotask.ai/glossary/data-enrichment): Data enrichment is the process of appending additional, relevant information to an existing record from external or internal sources to make it more complete or useful.
- [Data Entry Automation](https://neotask.ai/glossary/data-entry-automation): Data entry automation is the use of software to capture, transcribe, and input data into a system in place of a person manually typing it in.
- [Data Labeling](https://neotask.ai/glossary/data-labeling): Data labeling is the process of tagging raw data - text, images, audio, or structured records - with the categories or values a machine learning model needs to learn from.
- [Data Lake](https://neotask.ai/glossary/data-lake): A data lake is a centralized repository that stores raw data in its native format - structured, semi-structured, or unstructured - without requiring it to be organized into a schema before it's saved.
- [Data Lineage](https://neotask.ai/glossary/data-lineage): Data lineage is the tracked record of where a piece of data originated, what transformations it passed through, and where it ended up, across every system it touched.
- [Data Mapping](https://neotask.ai/glossary/data-mapping): Data mapping is the process of defining how fields in one data structure correspond to fields in another, so data can be correctly translated when it moves between systems.
- [Data Mart](https://neotask.ai/glossary/data-mart): A data mart is a focused subset of a data warehouse, structured and curated around the needs of a single department or business function such as sales, finance, or marketing.
- [Data Normalization](https://neotask.ai/glossary/data-normalization): Data normalization is the process of transforming data into a consistent, standard format or structure, removing redundancy and reconciling inconsistent representations of the same value.
- [Data Pipeline](https://neotask.ai/glossary/data-pipeline): A data pipeline is a series of automated steps that move data from one or more sources, through transformation and validation stages, to a destination where it can be used.
- [Data Processing Agreement](https://neotask.ai/glossary/data-processing-agreement): A data processing agreement, or DPA, is a legally binding contract between a data controller and a data processor that defines how personal data will be handled, protected, and used on the controller's behalf.
- [Data Residency](https://neotask.ai/glossary/data-residency): Data residency refers to the requirement or guarantee that specific data is stored and processed within a particular geographic or jurisdictional boundary.
- [Data Retention Policy](https://neotask.ai/glossary/data-retention-policy): A data retention policy is a documented rule set defining how long specific categories of data are kept before they are deleted or archived, and the criteria that trigger disposal.
- [Data Schema](https://neotask.ai/glossary/data-schema): A data schema is the formal structure that defines what fields exist in a data set, their types, constraints, and the relationships between them.
- [Data Sync](https://neotask.ai/glossary/data-sync): Data sync is the process of keeping data consistent across two or more systems, propagating changes made in one system to the others so they all reflect the same current state.
- [Data Transformation](https://neotask.ai/glossary/data-transformation): Data transformation is the process of converting data from one format, structure, or value representation into another so it fits the requirements of its destination system or use case.
- [Data Validation](https://neotask.ai/glossary/data-validation): Data validation is the process of checking that data meets defined rules for correctness, completeness, and format before it is accepted, stored, or acted upon.
- [Data Warehouse](https://neotask.ai/glossary/data-warehouse): A data warehouse is a centralized repository that stores structured, cleaned, and schema-enforced data optimized for analytical querying and reporting rather than day-to-day transactional operations.
- [Database Automation](https://neotask.ai/glossary/database-automation): Database automation is the use of software to perform database administration and maintenance tasks - such as backups, scaling, indexing, and schema migrations - without manual, ad hoc intervention.
- [Database Integration](https://neotask.ai/glossary/database-integration): Database integration is the practice of connecting an application or agent directly to a structured data store so it can read, write, and query records as part of an automated workflow, rather than through a screen or a manual export.
- [Dead Letter Queue](https://neotask.ai/glossary/dead-letter-queue): A dead letter queue is a holding location for messages or jobs that a processing system tried and failed to handle successfully, so they can be inspected and retried instead of being silently dropped.
- [Deep Learning](https://neotask.ai/glossary/deep-learning): Deep learning is a branch of machine learning that uses multi-layered artificial neural networks to learn hierarchical patterns directly from raw data, without hand-engineered features.
- [Deployment Automation](https://neotask.ai/glossary/deployment-automation): Deployment automation is the practice of shipping code, configuration, or infrastructure changes to a target environment through a scripted, repeatable pipeline rather than a manual, hands-on-keyboard release process.
- [Design Automation](https://neotask.ai/glossary/design-automation): Design automation is the use of rules, templates, or generative models to produce visual or structural design assets — layouts, mockups, brand variations, or UI components — with minimal manual authoring for each new instance.
- [Development Automation](https://neotask.ai/glossary/development-automation): Development automation refers to tools and agents that handle software engineering tasks — writing code, running tests, reviewing diffs, generating scaffolding — with reduced manual developer effort per task.
- [DevOps Automation](https://neotask.ai/glossary/devops-automation): DevOps automation is the scripting and orchestration of the operational lifecycle of software — building, testing, deploying, monitoring, and scaling infrastructure — so these tasks run consistently without manual intervention at each step.
- [Digital Worker](https://neotask.ai/glossary/digital-worker): A digital worker is an AI-driven software entity assigned to perform a defined set of job-like tasks — handling tickets, processing invoices, qualifying leads — with enough autonomy to complete the work end to end rather than just...
- [Document Automation](https://neotask.ai/glossary/document-automation): Document automation is the automated generation, extraction, or processing of documents — contracts, invoices, reports, forms — using templates, rules, or AI models instead of manual drafting or data entry.
- [Donor Management Automation](https://neotask.ai/glossary/donor-management-automation): Donor management automation is the use of software to handle the recurring operational work of a nonprofit's fundraising program — gift processing, acknowledgment letters, pledge reminders, and donor segmentation — without manual staff...
- [Dunning Automation](https://neotask.ai/glossary/dunning-automation): Dunning automation is the automated process of detecting failed or overdue payments and running a scripted sequence of retries, reminders, and escalating communications to recover the revenue before a customer is lost.
- [Durable Execution](https://neotask.ai/glossary/durable-execution): Durable execution is a workflow-processing pattern where the state of a long-running, multi-step process is persisted at each step, so the process can resume exactly where it left off after a crash, restart, or infrastructure failure...
- [Ecommerce Automation](https://neotask.ai/glossary/ecommerce-automation): Ecommerce automation is the use of software and AI to handle the recurring operational tasks of running an online store — order processing, inventory sync, customer messaging, and returns — without manual work per order.
- [Edge AI](https://neotask.ai/glossary/edge-ai): Edge AI refers to running AI model inference directly on a local device — a phone, a camera, an IoT sensor, a laptop — rather than sending data to a remote cloud server for processing.
- [Education Automation](https://neotask.ai/glossary/education-automation): Education automation is the use of software and AI to handle repetitive instructional and administrative work in learning environments — grading, scheduling, progress tracking, and personalized practice — reducing the manual load on...
- [ELT (Extract, Load, Transform)](https://neotask.ai/glossary/elt): ELT is a data integration pattern where raw data is extracted from a source system and loaded directly into a destination data store before any transformation happens, with the transformation step run afterward inside the destination...
- [Email Agent](https://neotask.ai/glossary/email-agent): An email agent is an AI system that autonomously reads, drafts, triages, or sends email on a person's or organization's behalf, going beyond a simple filter or template to make judgment calls about content and priority.
- [Email Automation](https://neotask.ai/glossary/email-automation): Email automation is the use of triggered, rule-based, or AI-driven workflows to send emails automatically in response to an event or schedule, rather than a person composing and sending each message manually.
- [Email Marketing Automation](https://neotask.ai/glossary/email-marketing-automation): Email marketing automation is the use of triggered, segmented, and scheduled email campaigns to nurture leads and engage customers at scale, driven by audience behavior and lifecycle stage rather than one-off manual sends.
- [Embedded iPaaS](https://neotask.ai/glossary/embedded-ipaas): Embedded iPaaS is integration-platform-as-a-service technology that a software vendor bundles directly into its own product, so the vendor's customers can connect third-party apps without the vendor building and maintaining each...
- [Embedding](https://neotask.ai/glossary/embedding): An embedding is a numerical vector representation of a piece of data — a word, sentence, image, or document — positioned in a high-dimensional space such that items with similar meaning end up close together geometrically.
- [Employee Onboarding Automation](https://neotask.ai/glossary/employee-onboarding-automation): Employee onboarding automation is the use of software to handle the recurring administrative steps of bringing a new hire into an organization — account provisioning, document collection, training assignment, and equipment requests —...
- [Encryption at Rest](https://neotask.ai/glossary/encryption-at-rest): Encryption at rest is the practice of encrypting data while it is stored — on disk, in a database, in a backup — so that anyone who gains access to the underlying storage medium without the decryption key sees only ciphertext, not usable...
- [Encryption in Transit](https://neotask.ai/glossary/encryption-in-transit): Encryption in transit is the practice of encrypting data as it moves across a network — between a browser and a server, between two internal services — so that anyone intercepting the traffic sees only ciphertext, not the underlying...
- [Enterprise AI Agent](https://neotask.ai/glossary/enterprise-ai-agent): An enterprise AI agent is an AI system designed to operate within the security, compliance, and scale requirements of a large organization — handling tenant isolation, audit logging, and role-based access alongside whatever business task...
- [Episodic Memory (AI)](https://neotask.ai/glossary/episodic-memory-ai): Episodic memory in AI systems is the capacity to store and later recall specific past events or interactions, tagged with when and in what context they happened, rather than only general facts.
- [Escalation Automation](https://neotask.ai/glossary/escalation-automation): Escalation automation is the automatic routing of a task, ticket, or alert to a higher tier of authority or expertise when predefined conditions — such as a time threshold, severity level, or repeated failure — are met.
- [ETL (Extract, Transform, Load)](https://neotask.ai/glossary/etl): ETL is a data integration pattern that extracts data from source systems, transforms it into a consistent structure and quality, and loads it into a target system such as a data warehouse.
- [Event Bus](https://neotask.ai/glossary/event-bus): An event bus is a messaging infrastructure component that lets independent parts of a system publish events and lets other parts subscribe to and react to those events, without the publisher and subscriber knowing about each other directly.
- [Event-Driven Automation](https://neotask.ai/glossary/event-driven-automation): Event-driven automation is a design approach where workflows are triggered automatically by the occurrence of specific events — a file upload, a status change, a webhook, a sensor reading — rather than running on a fixed schedule or...
- [Event Management Automation](https://neotask.ai/glossary/event-management-automation): Event management automation is the use of software to handle the logistics of planning, promoting, and running an event — registration, scheduling, communications, and follow-up — without manual coordination at each step.
- [Expense Management Automation](https://neotask.ai/glossary/expense-management-automation): Expense management automation is the use of software to capture, categorize, approve, and reconcile business expenses without employees or finance staff manually entering and cross-checking each line item.
- [Explainable AI (XAI)](https://neotask.ai/glossary/explainable-ai): Explainable AI refers to techniques and system designs that make an AI model's decisions understandable to humans, showing why it produced a particular output rather than leaving the reasoning as an opaque black box.
- [Feedback Collection Automation](https://neotask.ai/glossary/feedback-collection-automation): Feedback collection automation is the automatic triggering, distribution, and aggregation of surveys or feedback requests at the right moment in a customer or employee journey, without a person manually deciding when to ask and manually...
- [Few-Shot Learning](https://neotask.ai/glossary/few-shot-learning): Few-shot learning is a technique where a model is given only a handful of labeled examples of a task at inference time — often within the prompt itself — and uses them to perform the task correctly without additional training.
- [File-Based Integration](https://neotask.ai/glossary/file-based-integration): File-based integration is a method of connecting two systems by having one system produce a data file — typically CSV, XML, JSON, or a fixed-width format — that the other system periodically reads and imports.
- [Finance Automation](https://neotask.ai/glossary/finance-automation): Finance automation is the application of software to handle recurring financial processes — invoicing, accounts payable and receivable, reconciliation, reporting, and expense management — with minimal manual data entry or review.
- [Fine-Tuning](https://neotask.ai/glossary/fine-tuning): Fine-tuning is the process of taking a pretrained model and continuing its training on a smaller, task- or domain-specific dataset so its weights adapt to perform that particular task better.
- [Fleet Management Automation](https://neotask.ai/glossary/fleet-management-automation): Fleet management automation is the use of software and connected vehicle data to automatically track, schedule, and optimize a fleet of vehicles or equipment — covering maintenance, routing, fuel, and compliance — without manual tracking...
- [Forecasting Automation](https://neotask.ai/glossary/forecasting-automation): Forecasting automation is the use of statistical or machine-learning models to automatically generate ongoing predictions — of demand, revenue, staffing needs, or other business metrics — from historical and real-time data, without an...
- [Foundation Model](https://neotask.ai/glossary/foundation-model): A foundation model is a large-scale AI model trained on broad, diverse data that develops general-purpose capabilities, which can then be adapted — through prompting, fine-tuning, or retrieval augmentation — to a wide range of downstream...
- [Fraud Detection Automation](https://neotask.ai/glossary/fraud-detection-automation): Fraud detection automation is the use of rules engines and machine-learning models to automatically flag or block suspicious transactions and behaviors in real time, based on patterns that deviate from expected normal activity.
- [Front-Office Automation](https://neotask.ai/glossary/front-office-automation): Front-office automation is the use of software to handle customer-facing business processes — sales, customer service, marketing, and client onboarding — that directly touch the customer experience, as distinct from internal back-office...
- [Frontier Model](https://neotask.ai/glossary/frontier-model): A frontier model is an AI model that sits at or near the current cutting edge of capability, typically the largest and most advanced models released by leading AI labs at a given point in time.
- [Function Calling](https://neotask.ai/glossary/function-calling): Function calling is a capability that lets an AI model produce a structured request to invoke a specific external function or tool with defined arguments, rather than only generating free-form text, so the surrounding application can...
- [GDPR Compliance Automation](https://neotask.ai/glossary/gdpr-compliance-automation): GDPR compliance automation is the use of software to automatically fulfill obligations under the EU General Data Protection Regulation — such as honoring data subject access, deletion, and export requests, enforcing consent, and applying...
- [Generative AI](https://neotask.ai/glossary/generative-ai): Generative AI refers to AI models that create new content — text, images, audio, code, or video — rather than only classifying, scoring, or retrieving existing content.
- [GPU Computing](https://neotask.ai/glossary/gpu-computing): GPU computing is the use of graphics processing units — chips originally built for rendering images — to perform massively parallel numerical computation, which makes them well suited to the matrix and vector math underlying modern AI...
- [Grant Management Automation](https://neotask.ai/glossary/grant-management-automation): Grant management automation is the use of software to handle the lifecycle of grant funding — application intake, eligibility review, disbursement, compliance reporting, and outcome tracking — for organizations that award or receive grants.
- [GraphQL](https://neotask.ai/glossary/graphql): GraphQL is a query language and runtime for APIs that lets a client specify exactly which fields of data it needs in a single request, rather than receiving a fixed response shape defined entirely by the server.
- [Grounding (AI)](https://neotask.ai/glossary/grounding-ai): Grounding is the practice of anchoring an AI model's outputs to verifiable external data — documents, databases, search results, or live tool calls — rather than relying solely on what the model memorized during training.
- [Help Desk Automation](https://neotask.ai/glossary/help-desk-automation): Help desk automation is the use of software and AI to triage, route, resolve, or escalate customer and employee support tickets without a human agent touching every step.
- [HIPAA Compliance Automation](https://neotask.ai/glossary/hipaa-compliance-automation): HIPAA compliance automation is the use of software controls — access logging, encryption enforcement, retention policies, and audit trails — to continuously maintain and evidence adherence to the U.S.
- [Horizontal AI Agent](https://neotask.ai/glossary/horizontal-ai-agent): A horizontal AI agent is an AI agent designed to perform a general class of task — writing, research, scheduling, coding — across many industries and use cases, rather than being built for one specific vertical like healthcare billing or...
- [HR Automation](https://neotask.ai/glossary/hr-automation): HR automation is the application of software and AI to human resources processes — recruiting, onboarding, payroll, benefits administration, performance reviews, and offboarding — to reduce manual paperwork and administrative overhead.
- [Hub-and-Spoke Integration](https://neotask.ai/glossary/hub-and-spoke-integration): Hub-and-spoke integration is an architecture where a single central hub connects to and mediates between multiple peripheral systems, so each spoke system only needs to integrate with the hub rather than with every other system directly.
- [Human-in-the-Loop](https://neotask.ai/glossary/human-in-the-loop): Human-in-the-loop (HITL) is a system design where a person must review, approve, or correct an AI or automated system's output at a defined checkpoint before the action takes effect.
- [Human-on-the-Loop](https://neotask.ai/glossary/human-on-the-loop): Human-on-the-loop (HOTL) is a system design where an automated or AI system acts autonomously in real time while a human monitors its behavior and can intervene, pause, or override it, but does not have to approve each individual action...
- [Hyperautomation](https://neotask.ai/glossary/hyperautomation): Hyperautomation is the coordinated use of multiple automation technologies — RPA, AI/ML, business process management, and integration tooling — to automate as much of an organization's end-to-end processes as technically possible, rather...
- [Idempotency](https://neotask.ai/glossary/idempotency): Idempotency is the property of an operation such that performing it multiple times produces the same result as performing it once — repeated execution causes no additional side effects beyond the first.
- [Idempotent API](https://neotask.ai/glossary/idempotent-api): An idempotent API is an API endpoint designed so that calling it multiple times with the same input produces the same server-side result as calling it once, making it safe to retry after a timeout or uncertain response.
- [If-This-Then-That Logic](https://neotask.ai/glossary/if-this-then-that-logic): If-this-then-that (IFTTT) logic is the simplest automation pattern — a trigger condition (“this”) is watched for, and when it occurs, a predefined action (“that”) fires automatically, without any judgment or branching beyond the single...
- [Incident Management Automation](https://neotask.ai/glossary/incident-management-automation): Incident management automation is the use of software to detect, log, classify, prioritize, and route operational incidents — service outages, security events, system errors — through their full lifecycle with minimal manual coordination.
- [Incident Response Automation](https://neotask.ai/glossary/incident-response-automation): Incident response automation is the use of predefined, automated actions — isolating a compromised host, revoking credentials, blocking an IP, killing a malicious process — to contain and remediate a security incident within seconds of...
- [Inference](https://neotask.ai/glossary/inference): Inference is the process of running a trained AI model on new input to produce an output — a prediction, a classification, or generated text — as distinct from training, which is the earlier process of adjusting the model's parameters on a...
- [Infrastructure Automation](https://neotask.ai/glossary/infrastructure-automation): Infrastructure automation is the practice of provisioning, configuring, and managing computing infrastructure — servers, networks, storage, cloud resources — through code and automated tooling instead of manual, one-off administrative...
- [Integration Connector](https://neotask.ai/glossary/integration-connector): An integration connector is a purpose-built software component that translates between one specific external system's API and a common internal interface, so other parts of a platform can interact with that system without knowing its...
- [Integration Platform](https://neotask.ai/glossary/integration-platform): An integration platform is software infrastructure that connects otherwise-separate applications and data sources so they can exchange data and trigger actions in each other, without every pair of systems needing a custom point-to-point...
- [Intelligent Automation](https://neotask.ai/glossary/intelligent-automation): Intelligent automation is automation that incorporates AI capabilities — natural language understanding, computer vision, or machine learning-based decision-making — to handle tasks that require interpreting unstructured input or making...
- [Inventory Automation](https://neotask.ai/glossary/inventory-automation): Inventory automation is the use of software and AI to track stock levels, forecast demand, trigger reordering, and reconcile discrepancies across a supply chain without manual counting or spreadsheet updates.
- [Invoice Automation](https://neotask.ai/glossary/invoice-automation): Invoice automation is the use of software and AI to capture, extract, validate, and process invoices — matching them against purchase orders and routing them for payment — without a human manually keying in line items from each document.
- [iPaaS (Integration Platform as a Service)](https://neotask.ai/glossary/ipaas): iPaaS is a cloud-hosted platform that provides the tools to build, deploy, and manage integrations between applications and data sources — connectors, data mapping, and workflow orchestration — without an organization having to run its own...
- [ISO/IEC 27001](https://neotask.ai/glossary/iso-27001): ISO/IEC 27001 is an internationally recognized standard that specifies the requirements for establishing, implementing, maintaining, and continually improving an Information Security Management System (ISMS) — a structured set of policies,...
- [IT Automation](https://neotask.ai/glossary/it-automation): IT automation is the use of software to perform routine information-technology tasks — provisioning accounts, patching systems, deploying software, monitoring infrastructure — without a system administrator manually executing each step.
- [Job Scheduler](https://neotask.ai/glossary/job-scheduler): A job scheduler is a system that triggers the execution of tasks at specified times, intervals, or in response to defined conditions, and manages their queuing, retries, and completion tracking without a human manually starting each run.
- [JSON](https://neotask.ai/glossary/json): JSON (JavaScript Object Notation) is a lightweight, text-based data format that represents structured information as nested key-value pairs, arrays, and primitive values.
- [JWT](https://neotask.ai/glossary/jwt): A JWT (JSON Web Token) is a compact, digitally signed token format used to represent claims — such as a user's identity or an agent's permissions — that a receiving system can verify without a database lookup.
- [AI Knowledge Base](https://neotask.ai/glossary/knowledge-base-ai): An AI knowledge base is a curated, searchable repository of documents, facts, and structured content that an AI system consults at query time to ground its answers in accurate, current information.
- [Knowledge Management Automation](https://neotask.ai/glossary/knowledge-management-automation): Knowledge management automation is the use of software to capture, organize, tag, and surface an organization's institutional knowledge without relying on manual filing or a human remembering where something lives.
- [KYC Verification Automation](https://neotask.ai/glossary/kyc-verification-automation): KYC verification automation is the automated execution of Know Your Customer checks — identity document validation, biometric matching, sanctions and watchlist screening — that regulated businesses must perform before onboarding a customer.
- [Large Language Model](https://neotask.ai/glossary/large-language-model): A large language model (LLM) is a neural network trained on massive text corpora to predict and generate coherent language, learning grammar, facts, and reasoning patterns as a byproduct of that prediction task.
- [LDAP](https://neotask.ai/glossary/ldap): LDAP (Lightweight Directory Access Protocol) is a standard protocol for querying and modifying hierarchical directory information — most commonly user accounts, groups, and permissions in enterprise identity systems.
- [Lead Generation Automation](https://neotask.ai/glossary/lead-generation-automation): Lead generation automation is the use of software to identify, capture, and qualify potential customers with minimal manual prospecting, replacing hand-built lists and cold outreach with rule-driven or AI-driven pipelines.
- [Lead Routing Automation](https://neotask.ai/glossary/lead-routing-automation): Lead routing automation is the automatic assignment of an incoming sales lead to the right salesperson or team based on rules like territory, deal size, product line, or current rep workload.
- [Lead Scoring Automation](https://neotask.ai/glossary/lead-scoring-automation): Lead scoring automation is the automatic ranking of prospective customers by their likelihood to convert, using a weighted combination of behavioral, firmographic, and engagement signals rather than a rep's gut feel.
- [Least Privilege Access](https://neotask.ai/glossary/least-privilege-access): Least privilege access is the security principle that any user, process, or system should hold only the minimum permissions required to perform its specific task, and nothing more.
- [Least Privilege Agent Access](https://neotask.ai/glossary/least-privilege-agent-access): Least privilege agent access is the practice of scoping an AI agent's tool permissions, data visibility, and action authority to only what its current task requires, rather than granting it standing access to every system it might...
- [Legal Automation](https://neotask.ai/glossary/legal-automation): Legal automation is the use of software to handle routine legal work — contract drafting and review, clause extraction, compliance tracking, document assembly — that traditionally required a lawyer's manual time.
- [Listing Management Automation](https://neotask.ai/glossary/listing-management-automation): Listing management automation is the automated synchronization of product, property, or inventory listings — pricing, availability, descriptions, images — across multiple sales channels or marketplaces from a single source of truth.
- [LLM Routing](https://neotask.ai/glossary/llm-routing): LLM routing is the practice of directing each individual request to the most appropriate large language model out of several available, based on the task's complexity, cost sensitivity, latency requirement, or required capability.
- [Local MCP Server](https://neotask.ai/glossary/local-mcp-server): A local MCP server is a Model Context Protocol server that runs on the same machine as the AI agent or client consuming it, exposing tools and data over a local transport (typically stdio) rather than a network connection.
- [Logistics Automation](https://neotask.ai/glossary/logistics-automation): Logistics automation is the use of software and connected systems to plan, execute, and monitor the movement of goods — routing, dispatch, inventory tracking, carrier selection — with minimal manual coordination.
- [Long-Running Workflow](https://neotask.ai/glossary/long-running-workflow): A long-running workflow is an automated process that executes over an extended period — minutes, hours, or days — often pausing between steps to wait on external events like a human approval, a scheduled time, or a third-party system's...
- [Long-Term Memory (AI)](https://neotask.ai/glossary/long-term-memory-ai): Long-term memory in AI systems refers to information that persists across separate conversations or sessions — facts, preferences, past decisions — so an agent can recall relevant context from prior interactions rather than starting fresh...
- [Low-Code Automation](https://neotask.ai/glossary/low-code-automation): Low-code automation is building automated workflows through visual, configuration-driven tools — drag-and-drop steps, connectors, and conditional logic — rather than writing traditional software code.
- [Machine Learning](https://neotask.ai/glossary/machine-learning): Machine learning is a field of computer science in which systems improve at a task by learning statistical patterns from data, rather than following rules explicitly programmed by a developer.
- [Maintenance Request Automation](https://neotask.ai/glossary/maintenance-request-automation): Maintenance request automation is the automated intake, triage, and routing of facility or equipment maintenance requests — from initial report through vendor dispatch and completion tracking — without manual coordination at each step.
- [Marketing Automation](https://neotask.ai/glossary/marketing-automation): Marketing automation is the use of software to execute, sequence, and personalize marketing communications — email campaigns, ad targeting, content delivery — based on triggers and customer behavior rather than manual scheduling.
- [Master Data Management](https://neotask.ai/glossary/master-data-management): Master data management (MDM) is the discipline of establishing a single, authoritative, consistent version of an organization's core business entities — customers, products, vendors, locations — that every system references instead of...
- [MCP Client](https://neotask.ai/glossary/mcp-client): An MCP client is the component of an AI application — typically embedded in the agent runtime — that connects to one or more Model Context Protocol servers to discover and invoke their exposed tools, resources, and prompts on the model's...
- [MCP Elicitation](https://neotask.ai/glossary/mcp-elicitation): MCP elicitation is a Model Context Protocol feature that lets a server pause mid-task and ask the connected client to collect missing information from the human user before continuing.
- [MCP Gateway](https://neotask.ai/glossary/mcp-gateway): An MCP gateway is a middle-tier service that sits between an AI client and one or more MCP servers, centralizing connection routing, authentication, and policy enforcement instead of the client dialing each server directly.
- [MCP Host](https://neotask.ai/glossary/mcp-host): An MCP host is the application — typically an AI assistant or IDE — that embeds an MCP client and orchestrates the overall interaction between the language model, the user, and any number of connected MCP servers.
- [MCP (Model Context Protocol)](https://neotask.ai/glossary/mcp-model-context-protocol): MCP, or Model Context Protocol, is an open standard that defines how AI applications connect to external tools, data sources, and services through a uniform client-server interface.
- [MCP OAuth](https://neotask.ai/glossary/mcp-oauth): MCP OAuth refers to the OAuth 2.1-based authorization flow the Model Context Protocol specification defines for letting a client obtain user-delegated, scoped access tokens for a remote MCP server without ever seeing the user's underlying...
- [MCP Prompts](https://neotask.ai/glossary/mcp-prompts): MCP prompts are a primitive in the Model Context Protocol that let a server expose reusable, parameterized prompt templates a client can surface to the user or inject into a model conversation.
- [MCP Registry](https://neotask.ai/glossary/mcp-registry): An MCP registry is a discoverable catalog of published MCP servers — their metadata, capabilities, and connection details — that lets clients and hosts find and install servers without prior knowledge of them.
- [MCP Resources](https://neotask.ai/glossary/mcp-resources): MCP resources are a primitive in the Model Context Protocol through which a server exposes readable data — files, database records, API responses — that a client can fetch and feed into a model's context.
- [MCP Roots](https://neotask.ai/glossary/mcp-roots): MCP roots are a mechanism by which a client tells a connected server which filesystem directories or URI boundaries it is permitted to operate within, scoping the server's effective workspace.
- [MCP Sampling](https://neotask.ai/glossary/mcp-sampling): MCP sampling is a Model Context Protocol capability that lets a server request a language-model completion through the connected client, effectively borrowing the client's model access rather than needing its own.
- [MCP Security](https://neotask.ai/glossary/mcp-security): MCP security is the set of trust, authorization, and isolation practices needed to safely connect an AI host to Model Context Protocol servers that may be third-party, untrusted, or compromised.
- [MCP Server](https://neotask.ai/glossary/mcp-server): An MCP server is a lightweight program that exposes a specific set of tools, data resources, or prompts to AI applications over the Model Context Protocol.
- [MCP SSE (Server-Sent Events)](https://neotask.ai/glossary/mcp-sse): MCP SSE refers to the original HTTP-based transport for the Model Context Protocol that used Server-Sent Events for server-to-client streaming, paired with regular HTTP POST for client-to-server messages.
- [MCP stdio](https://neotask.ai/glossary/mcp-stdio): MCP stdio is the transport in which a client launches an MCP server as a local subprocess and exchanges JSON-RPC messages with it over the process's standard input and output streams.
- [MCP Streamable HTTP](https://neotask.ai/glossary/mcp-streamable-http): MCP Streamable HTTP is the current standard network transport for the Model Context Protocol, using a single HTTP endpoint that can respond either as a normal request-response or upgrade to a streamed series of messages.
- [MCP Tools](https://neotask.ai/glossary/mcp-tools): MCP tools are the primitive in the Model Context Protocol through which a server exposes callable, side-effect-producing actions that a model can invoke — sending a message, creating a record, running a query.
- [MCP Transport](https://neotask.ai/glossary/mcp-transport): MCP transport refers to the underlying communication channel — stdio for local subprocesses or Streamable HTTP for networked servers — over which Model Context Protocol JSON-RPC messages actually travel between client and server.
- [Media Automation](https://neotask.ai/glossary/media-automation): Media automation is the use of software and AI to handle repetitive tasks in producing, processing, and distributing images, audio, or video without manual, file-by-file intervention.
- [Meeting Automation](https://neotask.ai/glossary/meeting-automation): Meeting automation is the application of software and AI to handle the administrative overhead surrounding meetings — scheduling, note-taking, action-item extraction, and follow-up — without a human doing it by hand.
- [Membership Management Automation](https://neotask.ai/glossary/membership-management-automation): Membership management automation is the use of software to handle the recurring lifecycle tasks of running a membership-based organization — signups, renewals, dues collection, access provisioning, and communications — without manual...
- [Message Broker](https://neotask.ai/glossary/message-broker): A message broker is an intermediary system that receives messages from producing applications and routes them to the appropriate consuming applications, decoupling senders from receivers so neither needs to know about the other directly.
- [Middleware](https://neotask.ai/glossary/middleware): Middleware is software that sits between two other pieces of software — an incoming request and the application logic that handles it, or two separate systems — performing cross-cutting work like authentication, logging, or data...
- [Model Alignment](https://neotask.ai/glossary/model-alignment): Model alignment is the practice of training and shaping an AI model so its outputs reliably reflect the intentions, values, and constraints its developers and users actually want, rather than merely optimizing a proxy training objective.
- [Model Context Protocol](https://neotask.ai/glossary/model-context-protocol): Model Context Protocol is an open, standardized specification for how AI applications discover, connect to, and exchange data and capabilities with external tools and information sources.
- [Model Distillation](https://neotask.ai/glossary/model-distillation): Model distillation is a machine-learning technique in which a smaller student model is trained to reproduce the behavior of a larger, more capable teacher model, capturing much of its performance at a fraction of the size and inference...
- [Model Evaluation](https://neotask.ai/glossary/model-evaluation): Model evaluation is the process of measuring how well a machine learning or language model performs against a set of tasks, benchmarks, or held-out data before and after deployment.
- [Model Parameters](https://neotask.ai/glossary/model-parameters): Model parameters are the internal numerical values, weights and biases, that a neural network learns during training and uses to transform input into output.
- [Model Quantization](https://neotask.ai/glossary/model-quantization): Model quantization is the technique of reducing the numerical precision of a neural network's weights and activations, for example from 16-bit or 32-bit floating point down to 8-bit or 4-bit integers, to shrink memory footprint and speed...
- [Model Risk Management](https://neotask.ai/glossary/model-risk-management): Model risk management (MRM) is the governance discipline of identifying, measuring, and controlling the risks that arise from relying on a statistical or machine learning model for a business decision, including the risk that the model is...
- [Model Weights](https://neotask.ai/glossary/model-weights): Model weights are the specific learned numerical values stored at each connection in a neural network, and together they encode everything the model has learned from its training data.
- [Monitoring & Alerts Automation](https://neotask.ai/glossary/monitoring-alerts-automation): Monitoring and alerts automation is the practice of continuously watching systems, metrics, or data streams and automatically triggering notifications or remediation actions when defined conditions are met, without a person having to watch...
- [Multi-Agent System](https://neotask.ai/glossary/multi-agent-system): A multi-agent system is an AI architecture in which several distinct agents, each with its own role, tools, or specialization, collaborate, communicating with one another and often a coordinating agent, to complete a task that no single...
- [Multi-Step Workflow](https://neotask.ai/glossary/multi-step-workflow): A multi-step workflow is an automated process made up of several ordered actions or decisions that execute in sequence, or with branching logic, to accomplish a larger goal, rather than a single one-shot action.
- [Multi-Tenancy](https://neotask.ai/glossary/multi-tenancy): Multi-tenancy is a software architecture pattern in which a single instance of an application serves multiple separate customers, or tenants, while keeping each tenant's data, configuration, and usage logically isolated from every other...
- [Multimodal AI](https://neotask.ai/glossary/multimodal-ai): Multimodal AI refers to models that can process and reason across more than one type of input, such as text, images, audio, and video, within a single system, rather than being limited to one modality.
- [Native Integration](https://neotask.ai/glossary/native-integration): A native integration is a connection between two software systems built directly against the target system's own API using its documented data model and authentication, rather than relying on a generic connector, screen-scraping, or a...
- [Neural Network](https://neotask.ai/glossary/neural-network): A neural network is a machine learning model composed of layers of interconnected nodes, loosely inspired by biological neurons, that learns to map inputs to outputs by adjusting the strength, or weight, of connections between nodes during...
- [No-Code AI Agent](https://neotask.ai/glossary/no-code-ai-agent): A no-code AI agent is an autonomous or semi-autonomous software agent that a person can configure and deploy, defining its goals, tools, and behavior, entirely through a visual interface or plain-language instructions, without writing code.
- [No-Code Automation](https://neotask.ai/glossary/no-code-automation): No-code automation is the practice of building automated workflows and integrations using visual tools, triggers, actions, and conditional logic assembled through a graphical interface, rather than writing custom software code.
- [Notification Automation](https://neotask.ai/glossary/notification-automation): Notification automation is the automatic generation and delivery of alerts or messages to the right people or systems in response to an event, without a person manually composing and sending each one.
- [OAuth](https://neotask.ai/glossary/oauth): OAuth is an open authorization protocol that lets a user grant a third-party application limited access to their data on another service, without ever sharing their password with that third-party application.
- [OAuth 2.0](https://neotask.ai/glossary/oauth2): OAuth 2.0 is the current version of the OAuth authorization framework, published in 2012 as RFC 6749, that defines standardized flows for an application to obtain limited, scoped access tokens on a user's behalf without handling their...
- [Offboarding Automation](https://neotask.ai/glossary/offboarding-automation): Offboarding automation is the automated execution of the tasks required when an employee, contractor, or system access leaves an organization, revoking accounts, transferring data ownership, and notifying stakeholders, triggered by a...
- [On-Call Automation](https://neotask.ai/glossary/on-call-automation): On-call automation is the use of software to manage incident escalation for the engineers or staff responsible for responding to production issues, routing alerts to the correct on-call person, escalating when there's no response, and...
- [On-Device AI](https://neotask.ai/glossary/on-device-ai): On-device AI refers to running AI models directly on a local device, a phone, laptop, or edge server, rather than sending data to a remote cloud server for inference.
- [One-Way Sync](https://neotask.ai/glossary/one-way-sync): One-way sync is a data-synchronization pattern in which changes flow from a single source system to one or more destination systems, but changes made in the destination are never pushed back to the source.
- [Open-Weights Model](https://neotask.ai/glossary/open-weights-model): An open-weights model is an AI model whose trained parameters, or weights, are published for anyone to download, run, and often fine-tune, in contrast to a closed model whose weights remain private and are only accessible through a hosted...
- [OpenAPI Specification](https://neotask.ai/glossary/openapi-specification): The OpenAPI Specification (OAS) is a standardized, language-agnostic format, usually written in YAML or JSON, for describing a REST API's endpoints, request and response schemas, authentication methods, and parameters in a way that both...
- [OpenID Connect](https://neotask.ai/glossary/openid-connect): OpenID Connect (OIDC) is an identity layer built on top of OAuth 2.0 that adds standardized authentication, letting an application verify who a user is, not just what resources they're authorized to access, by issuing a signed ID token...
- [Order Management Automation](https://neotask.ai/glossary/order-management-automation): Order management automation is the automated handling of the steps between a customer placing an order and its fulfillment, validation, inventory allocation, payment capture, fulfillment routing, and status updates, without a person...
- [Patient Scheduling Automation](https://neotask.ai/glossary/patient-scheduling-automation): Patient scheduling automation is the use of software to book, confirm, remind, and reshuffle medical appointments without a staff member manually managing every calendar slot.
- [Payments Automation](https://neotask.ai/glossary/payments-automation): Payments automation is software-driven initiation, routing, and reconciliation of money movement — invoicing, collections, payouts, and settlement — with minimal manual entry.
- [Payroll Automation](https://neotask.ai/glossary/payroll-automation): Payroll automation is the automated calculation, approval, and disbursement of employee pay, including tax withholding, benefits deductions, and compliance filings, run on a recurring schedule without manual recalculation each cycle.
- [Penetration Testing](https://neotask.ai/glossary/penetration-testing): Penetration testing is an authorized, simulated attack against a system, network, or application performed to find exploitable security weaknesses before a real attacker does.
- [Perplexity (Metric)](https://neotask.ai/glossary/perplexity-metric): Perplexity is a measure of how well a language model predicts a sample of text, calculated as the exponential of the average negative log-likelihood the model assigns to each token, where lower perplexity means better prediction.
- [PII (Personally Identifiable Information)](https://neotask.ai/glossary/pii-personally-identifiable-information): Personally identifiable information (PII) is any data that can identify a specific individual, either on its own (a name, a passport number) or in combination with other data (a birthdate plus a ZIP code plus a gender).
- [Pipeline Management Automation](https://neotask.ai/glossary/pipeline-management-automation): Pipeline management automation is the automated tracking, scoring, and progression of sales (or other multi-stage) opportunities through defined stages, replacing manual spreadsheet or CRM upkeep with rule- or AI-driven updates.
- [Plan-and-Execute (Agent Pattern)](https://neotask.ai/glossary/plan-and-execute): Plan-and-execute is an AI agent architecture that first generates a multi-step plan for a task, then executes each step in sequence, optionally revising the plan as new information arrives.
- [Point-to-Point Integration](https://neotask.ai/glossary/point-to-point-integration): Point-to-point integration is a direct, dedicated connection built between exactly two specific systems to exchange data, as opposed to routing that exchange through a shared middleware or bus layer.
- [Policy Renewal Automation](https://neotask.ai/glossary/policy-renewal-automation): Policy renewal automation is the automated tracking of insurance policy expiration dates and the triggering of quoting, underwriting-data refresh, and renewal-offer workflows ahead of the lapse date, without an agent manually monitoring...
- [Polling vs. Webhooks](https://neotask.ai/glossary/polling-vs-webhooks): Polling and webhooks are two opposite strategies for learning about a change in a remote system: polling repeatedly asks "did anything change?" on a schedule, while webhooks have the remote system push a notification the instant something...
- [Postman Collection](https://neotask.ai/glossary/postman-collection): A Postman collection is a saved, shareable set of API requests — including URLs, headers, auth, and example bodies — organized together so a team can run, test, and document an API's endpoints consistently.
- [Pre-training](https://neotask.ai/glossary/pre-training): Pre-training is the initial, large-scale phase of training a language model on massive amounts of general text, during which it learns broad language patterns, facts, and reasoning ability before any task-specific fine-tuning happens.
- [Prebuilt Connector](https://neotask.ai/glossary/prebuilt-connector): A prebuilt connector is a ready-made integration component that handles authentication, API calls, and data mapping for a specific external service, so an integration builder can wire it into a workflow without writing that service's API...
- [Privileged Access Management (PAM)](https://neotask.ai/glossary/privileged-access-management): Privileged access management (PAM) is the discipline and tooling for controlling, monitoring, and time-limiting access to accounts and credentials that carry elevated permissions — admin accounts, root, service credentials, and database...
- [Process Mining](https://neotask.ai/glossary/process-mining): Process mining is the technique of reconstructing how a business process actually runs by analyzing event logs from the systems that executed it, revealing the real sequence of steps, bottlenecks, and deviations rather than the process as...
- [Procurement Automation](https://neotask.ai/glossary/procurement-automation): Procurement automation is the automated handling of purchase requests, vendor selection, approval routing, and purchase order issuance, replacing manual email chains and spreadsheet tracking of what a company is buying and from whom.
- [Productivity Automation](https://neotask.ai/glossary/productivity-automation): Productivity automation is the use of software or AI agents to handle repetitive individual-level work — email triage, note-taking, scheduling, document drafting — so a person's time goes to judgment-requiring work instead of...
- [Project Management Automation](https://neotask.ai/glossary/project-management-automation): Project management automation is the automated creation, assignment, status-updating, and escalation of tasks within a project-tracking tool, reducing the manual overhead of keeping a project board accurate as work progresses.
- [Prompt](https://neotask.ai/glossary/prompt): A prompt is the text (and any accompanying images, files, or context) given as input to a language model to elicit a specific response or behavior.
- [Prompt Engineering](https://neotask.ai/glossary/prompt-engineering): Prompt engineering is the practice of deliberately designing and refining the input given to a language model — instructions, examples, structure, and context — to reliably produce the desired output.
- [Prompt Injection](https://neotask.ai/glossary/prompt-injection): Prompt injection is an attack in which malicious instructions are embedded in content a language model processes — a document, webpage, email, or tool output — attempting to override the model's original instructions and hijack its...
- [Property Management Automation](https://neotask.ai/glossary/property-management-automation): Property management automation is the automated handling of tenant communication, maintenance-request routing, rent collection, and lease-renewal tracking across a portfolio of rental units, replacing manual per-property tracking.
- [Pub/Sub Messaging](https://neotask.ai/glossary/pub-sub-messaging): Publish/subscribe (pub/sub) messaging is a communication pattern where senders (publishers) emit messages to a named channel or topic without knowing who will receive them, and receivers (subscribers) register interest in a topic to get...
- [Purchase Order Automation](https://neotask.ai/glossary/purchase-order-automation): Purchase order automation is the automated generation, approval routing, and tracking of purchase orders from an approved requisition through to matching against the eventual invoice and receipt.
- [QA Testing Automation](https://neotask.ai/glossary/qa-testing-automation): QA testing automation is the practice of using scripted or AI-driven tools to execute software test cases, verify outputs, and flag regressions without a human manually clicking through each scenario.
- [Queue-Based Automation](https://neotask.ai/glossary/queue-based-automation): Queue-based automation is a workflow design pattern where tasks are placed onto a message queue and processed asynchronously by one or more workers, decoupling the moment work is created from the moment it is executed.
- [Quote-to-Cash Automation](https://neotask.ai/glossary/quote-to-cash-automation): Quote-to-cash automation streamlines the entire revenue cycle — from generating a price quote, through contract approval and order fulfillment, to invoicing and cash collection — by connecting the systems and approvals involved into one...
- [Rate Limiting](https://neotask.ai/glossary/rate-limiting): Rate limiting is the practice of restricting how many requests a client can make to a service within a given time window, in order to protect that service from overload, abuse, or unfair resource consumption.
- [ReAct Pattern (Reason + Act)](https://neotask.ai/glossary/react-pattern-ai): The ReAct pattern is an AI agent design in which a language model alternates between explicit reasoning steps ("thoughts") and concrete actions (like calling a tool), using each action's observed result to inform the next reasoning step.
- [Real-Time Automation](https://neotask.ai/glossary/real-time-automation): Real-time automation refers to workflows that trigger and complete actions within seconds of a triggering event, rather than on a batch schedule or after a delay.
- [Reasoning Model](https://neotask.ai/glossary/reasoning-model): A reasoning model is a large language model specifically trained or prompted to work through a problem in explicit intermediate steps before producing a final answer, improving performance on tasks that require multi-step logic, math, or...
- [Recruiting Automation](https://neotask.ai/glossary/recruiting-automation): Recruiting automation applies software and AI to repetitive hiring tasks — sourcing candidates, screening resumes, scheduling interviews, and sending follow-ups — so recruiters spend their time on judgment calls rather than administrative...
- [Reflection (AI Agents)](https://neotask.ai/glossary/reflection-ai-agents): Reflection is an AI agent technique in which the model critiques or evaluates its own prior output — or the outcome of an action it took — and uses that critique to revise its approach before continuing.
- [Refresh Token](https://neotask.ai/glossary/refresh-token): A refresh token is a long-lived credential issued during an OAuth authentication flow that a client uses to obtain a new short-lived access token without requiring the user to log in again.
- [Reinforcement Learning](https://neotask.ai/glossary/reinforcement-learning): Reinforcement learning (RL) is a machine learning approach in which an agent learns to take actions in an environment by receiving rewards or penalties for outcomes, gradually improving its policy to maximize cumulative reward rather than...
- [Release Management Automation](https://neotask.ai/glossary/release-management-automation): Release management automation replaces the manual coordination of shipping software — building, testing, approving, and deploying a version — with a scripted, repeatable pipeline that runs the same way every time.
- [Remote MCP Server](https://neotask.ai/glossary/remote-mcp-server): A remote MCP server is a Model Context Protocol server hosted over the network (typically via HTTP/SSE) rather than run as a local subprocess, allowing an AI agent to discover and call its tools without installing anything locally.
- [Reporting Automation](https://neotask.ai/glossary/reporting-automation): Reporting automation is the automatic generation and delivery of recurring business reports — pulling data from source systems, formatting it, and distributing it — without a person manually compiling the numbers each cycle.
- [Research Agent](https://neotask.ai/glossary/research-agent): A research agent is an AI agent designed to autonomously gather, evaluate, and synthesize information from multiple sources — searching, reading, cross-checking, and summarizing — to answer an open-ended question rather than a single...
- [Research Automation](https://neotask.ai/glossary/research-automation): Research automation is the use of software and AI to carry out the repetitive mechanics of a research process — searching, collecting sources, extracting facts, and organizing findings — at scale, freeing human researchers to focus on...
- [REST API](https://neotask.ai/glossary/rest-api): A REST API is a web service interface that follows Representational State Transfer principles — exposing resources as URLs and manipulating them through standard HTTP methods like GET, POST, PUT, and DELETE — to let different systems...
- [Retrieval-Augmented Generation (RAG)](https://neotask.ai/glossary/retrieval-augmented-generation): Retrieval-augmented generation is a technique that improves an LLM's answers by first retrieving relevant documents or data from an external knowledge source and inserting them into the model's context before it generates a response,...
- [Retry Logic](https://neotask.ai/glossary/retry-logic): Retry logic is the set of rules an application follows to automatically re-attempt a failed operation — such as a network call or API request — instead of failing immediately, typically with a delay and a maximum attempt count.
- [Returns Processing Automation](https://neotask.ai/glossary/returns-processing-automation): Returns processing automation handles the steps of a product return — approving the request, issuing a shipping label, tracking the item back, and triggering a refund or exchange — through software rather than manual case-by-case handling.
- [Revenue Operations Automation](https://neotask.ai/glossary/revenue-operations-automation): Revenue operations (RevOps) automation connects and automates the processes across sales, marketing, and customer success — lead routing, pipeline updates, renewal tracking — so revenue-generating teams work from one consistent,...
- [Reverse ETL](https://neotask.ai/glossary/reverse-etl): Reverse ETL is the process of moving cleaned, modeled data from a central data warehouse back out into the operational tools a business runs on — like a CRM, support desk, or ad platform — so that analytical insights become actionable...
- [Review Management Automation](https://neotask.ai/glossary/review-management-automation): Review management automation monitors, responds to, and analyzes customer reviews across platforms — sites like Google, Yelp, or app stores — using software to flag urgent issues, draft responses, and surface trends without a person...
- [Risk Assessment Automation](https://neotask.ai/glossary/risk-assessment-automation): Risk assessment automation uses rules or AI models to continuously evaluate transactions, accounts, or decisions against risk indicators — flagging or scoring the risky ones automatically — instead of relying on periodic manual review.
- [RLHF (Reinforcement Learning from Human Feedback)](https://neotask.ai/glossary/rlhf): RLHF is a technique for aligning a language model's behavior with human preferences by training a reward model on human rankings of model outputs, then using reinforcement learning to fine-tune the language model to produce responses that...
- [Robotic Process Automation](https://neotask.ai/glossary/robotic-process-automation): Robotic Process Automation (RPA) is software that replicates fixed, rule-based human actions on existing user interfaces or APIs, such as clicking through screens, copying fields, and filing forms, without understanding the task's intent.
- [Role-Based Access Control](https://neotask.ai/glossary/role-based-access-control): Role-Based Access Control (RBAC) is an authorization model that grants permissions to users based on assigned roles rather than to each user individually.
- [Route Optimization Automation](https://neotask.ai/glossary/route-optimization-automation): Route optimization automation is the automated calculation and adjustment of travel or delivery routes to minimize cost, time, or distance across multiple stops and constraints.
- [RPA vs Agentic AI](https://neotask.ai/glossary/rpa-vs-agentic-ai): RPA vs agentic AI describes the distinction between rule-based bots that replay fixed scripted steps and AI agents that reason about a goal and choose their own steps to reach it.
- [Sales Agent AI](https://neotask.ai/glossary/sales-agent-ai): A sales agent AI is an autonomous or semi-autonomous AI system that performs sales-cycle tasks — prospecting, qualifying, outreach, follow-up, and meeting scheduling — on behalf of a human sales team.
- [Sales Automation](https://neotask.ai/glossary/sales-automation): Sales automation is the use of software to execute repetitive sales-process steps — lead routing, sequence emails, CRM updates, follow-up reminders — without manual repetition by a rep.
- [Sandboxed Execution](https://neotask.ai/glossary/sandboxed-execution): Sandboxed execution is running code, commands, or an AI agent's actions inside an isolated environment that limits what it can access or affect on the underlying system.
- [Scheduled Automation](https://neotask.ai/glossary/scheduled-automation): Scheduled automation is a workflow that runs automatically at a predetermined time or recurring interval, such as a nightly report or a weekly data sync, without a human triggering it each time.
- [Scheduling Automation](https://neotask.ai/glossary/scheduling-automation): Scheduling automation is software that automatically finds, proposes, or books time slots for meetings, appointments, or resources by checking availability and applying booking rules.
- [SCIM Provisioning](https://neotask.ai/glossary/scim-provisioning): SCIM (System for Cross-domain Identity Management) provisioning is a standardized protocol for automatically creating, updating, and deactivating user accounts across applications from a central identity provider.
- [SDK](https://neotask.ai/glossary/sdk): An SDK (Software Development Kit) is a packaged set of tools, libraries, code samples, and documentation that lets developers build applications or integrations against a specific platform or API.
- [Search Automation](https://neotask.ai/glossary/search-automation): Search automation is the automated execution of information-retrieval tasks — running queries, aggregating results, and extracting relevant data — across one or more sources without a person manually searching each time.
- [Secrets Management](https://neotask.ai/glossary/secrets-management): Secrets management is the practice and tooling for securely storing, distributing, rotating, and revoking sensitive credentials — API keys, passwords, tokens, certificates — used by applications and services.
- [Security Automation](https://neotask.ai/glossary/security-automation): Security automation is the use of software to automatically detect, investigate, and respond to security events without requiring a human analyst to perform each step manually.
- [Security Information and Event Management (SIEM)](https://neotask.ai/glossary/security-information-event-management): A SIEM (Security Information and Event Management) system is a platform that aggregates, correlates, and analyzes log and event data from across an organization's systems to detect and alert on security threats in real time.
- [Semantic Memory (AI)](https://neotask.ai/glossary/semantic-memory-ai): Semantic memory in AI systems is stored knowledge of facts, concepts, and relationships that an agent can retrieve and reason with, independent of the specific conversation or episode in which it was learned.
- [Semantic Search](https://neotask.ai/glossary/semantic-search): Semantic search is a retrieval method that finds results based on conceptual meaning and intent rather than exact keyword matching.
- [Semi-Structured Data](https://neotask.ai/glossary/semi-structured-data): Semi-structured data is information that doesn't conform to a rigid tabular schema but still contains organizational markers — tags, keys, nesting — that make it more machine-parseable than fully unstructured text.
- [SEO Automation](https://neotask.ai/glossary/seo-automation): SEO automation is the use of software to perform search-engine-optimization tasks — keyword research, content generation, technical audits, backlink monitoring — at scale without manual repetition for every page or query.
- [Shadow AI](https://neotask.ai/glossary/shadow-ai): Shadow AI is the unsanctioned use of AI tools or models by employees within an organization without the knowledge, approval, or governance of IT and security teams.
- [Shadow IT](https://neotask.ai/glossary/shadow-it): Shadow IT is the use of software, hardware, or cloud services within an organization without explicit approval or visibility from the IT or security department.
- [Shipping Automation](https://neotask.ai/glossary/shipping-automation): Shipping automation is the automated handling of order fulfillment logistics — carrier selection, label generation, tracking updates, and delivery notifications — without manual processing per order.
- [Short-Term Memory (AI)](https://neotask.ai/glossary/short-term-memory-ai): Short-term memory in AI systems is the information an agent holds and reasons over within the active session or conversation, typically bounded by the model's context window, in contrast to durable memory that persists across sessions.
- [Similarity Search](https://neotask.ai/glossary/similarity-search): Similarity search is the process of finding items in a dataset that are most alike a given query item, typically measured by distance between vector representations rather than exact matching.
- [Single-Agent vs Multi-Agent Systems](https://neotask.ai/glossary/single-agent-vs-multi-agent): Single-agent vs multi-agent describes the architectural choice between one AI agent handling an entire task end-to-end versus multiple specialized agents collaborating, each responsible for a distinct part of the work.
- [SLA Monitoring Automation](https://neotask.ai/glossary/sla-monitoring-automation): SLA monitoring automation is the automated tracking of service-level-agreement commitments (response time, resolution time, uptime) against live activity, with alerts and escalations fired the moment a threshold is at risk.
- [Small Language Model](https://neotask.ai/glossary/small-language-model): A small language model (SLM) is a language model with a comparatively low parameter count, usually a few hundred million to a few billion parameters, designed to run efficiently on limited hardware while handling a narrower band of tasks...
- [SMS Automation](https://neotask.ai/glossary/sms-automation): SMS automation is the programmatic sending, receiving, and processing of text messages — appointment reminders, alerts, two-way conversations, and marketing texts — without a human manually composing and sending each one.
- [SOAP API](https://neotask.ai/glossary/soap-api): SOAP (Simple Object Access Protocol) is an XML-based messaging protocol for exchanging structured information between systems, typically over HTTP, defined by a strict envelope format and often accompanied by a WSDL contract describing...
- [SOC 2](https://neotask.ai/glossary/soc-2): SOC 2 (System and Organization Controls 2) is an auditing framework, developed by the AICPA, that evaluates a service organization's controls against five Trust Services Criteria — security, availability, processing integrity,...
- [Social Media Automation](https://neotask.ai/glossary/social-media-automation): Social media automation is the use of software to schedule, publish, monitor, and respond to social media content and engagement across platforms without a person manually performing each action.
- [Spreadsheet Automation](https://neotask.ai/glossary/spreadsheet-automation): Spreadsheet automation is the use of scripts, formulas, add-ons, or external agents to read, write, transform, and act on spreadsheet data without a person manually clicking through cells and menus.
- [SSO (Single Sign-On)](https://neotask.ai/glossary/sso-single-sign-on): Single sign-on (SSO) is an authentication scheme that lets a user log in once with one identity provider and gain access to multiple independent applications without re-entering credentials for each one.
- [Storage Automation](https://neotask.ai/glossary/storage-automation): Storage automation is the use of policy-driven software to manage the lifecycle of stored data — provisioning capacity, tiering between hot and cold storage, applying retention rules, and cleaning up obsolete files — without manual,...
- [Straight-Through Processing](https://neotask.ai/glossary/straight-through-processing): Straight-through processing (STP) is the end-to-end automation of a transaction or workflow from initiation to completion with no manual intervention, re-keying, or handoff between systems at any step.
- [Structured Data](https://neotask.ai/glossary/structured-data): Structured data is information organized into a predefined schema — fixed fields, types, and relationships, such as rows and columns in a database table — that lets software parse and query it reliably without interpretation.
- [Structured Output](https://neotask.ai/glossary/structured-output): Structured output is a language model response constrained to a predefined schema — typically JSON matching a specified set of fields and types — rather than free-form natural-language text.
- [Sub-Agent](https://neotask.ai/glossary/sub-agent): A sub-agent is a subordinate AI agent spawned by a parent (orchestrator) agent to handle a specific, bounded piece of a larger task, typically running with its own isolated context and reporting results back to the parent rather than to...
- [Subscription Management Automation](https://neotask.ai/glossary/subscription-management-automation): Subscription management automation is the automated handling of recurring-billing lifecycle events — signups, renewals, upgrades, downgrades, payment failures, and cancellations — without manual processing of each change.
- [Supervised Learning](https://neotask.ai/glossary/supervised-learning): Supervised learning is a machine learning approach in which a model is trained on a dataset of labeled examples — inputs paired with known correct outputs — so it learns to predict the output for new, unseen inputs.
- [Supply Chain Automation](https://neotask.ai/glossary/supply-chain-automation): Supply chain automation is the use of software and connected systems to manage procurement, inventory, logistics, and supplier coordination with minimal manual intervention across the physical and information flow of goods.
- [Support Agent AI](https://neotask.ai/glossary/support-agent-ai): Support agent AI refers to an AI system that handles customer support interactions autonomously or semi-autonomously — answering questions, resolving common issues, and escalating to a human only when a case exceeds its competence or...
- [Survey Automation](https://neotask.ai/glossary/survey-automation): Survey automation is the automated distribution, collection, and analysis of survey responses — sending the survey at the right trigger point, gathering answers across channels, and summarizing results — without manual coordination of each...
- [Swivel-Chair Automation](https://neotask.ai/glossary/swivel-chair-automation): Swivel-chair automation (or "swivel-chair integration") describes the anti-pattern of a worker manually re-keying or copying data between two or more systems that don't talk to each other, literally swiveling their chair between screens.
- [Synthetic Data](https://neotask.ai/glossary/synthetic-data): Synthetic data is artificially generated data — created by algorithms, simulations, or generative models rather than collected from real-world events — that mimics the statistical properties of real data for training or testing purposes.
- [System Prompt](https://neotask.ai/glossary/system-prompt): A system prompt is the instructions given to a language model before any user conversation begins, establishing its role, behavior boundaries, tone, available tools, and constraints for the entire session.
- [Task Automation](https://neotask.ai/glossary/task-automation): Task automation is the use of software to carry out a well-defined, repeatable piece of work — a single action or short sequence of steps — without a person performing it manually each time.
- [Task Decomposition](https://neotask.ai/glossary/task-decomposition): Task decomposition is the process of breaking a large, complex goal into smaller, more manageable sub-tasks that can be planned, executed, and verified independently — either by different agents, in parallel, or in a defined sequence.
- [Task Mining](https://neotask.ai/glossary/task-mining): Task mining is the automated analysis of user interaction data — clicks, keystrokes, screen captures, application usage — to reconstruct and understand exactly how workers actually perform a task, as a basis for identifying automation...
- [Task Queue](https://neotask.ai/glossary/task-queue): A task queue is a data structure and processing system that holds units of work to be executed asynchronously, in order, by one or more workers that pull tasks off the queue as capacity allows.
- [Tax Preparation Automation](https://neotask.ai/glossary/tax-preparation-automation): Tax preparation automation applies software and AI agents to collect financial records, classify transactions, apply the relevant tax code, and assemble a filing with minimal manual re-keying.
- [Temperature (AI)](https://neotask.ai/glossary/temperature-ai): Temperature is a sampling parameter that controls how random or deterministic a language model's next-token choices are, with lower values favoring the highest-probability token and higher values spreading probability mass across more...
- [Tenant Isolation](https://neotask.ai/glossary/tenant-isolation): Tenant isolation is the architectural guarantee that each customer's data, credentials, and running processes are kept separate from every other tenant sharing the same multi-tenant system.
- [Tenant Screening Automation](https://neotask.ai/glossary/tenant-screening-automation): Tenant screening automation is the use of software to collect a rental applicant's background, credit, income, and eviction history and score or summarize it for a property manager's approval decision.
- [Third-Party Integration](https://neotask.ai/glossary/third-party-integration): A third-party integration is a connection that lets one software system read, write, or trigger actions in an external vendor's application, typically through an API, webhook, or authenticated connector.
- [Ticket Routing Automation](https://neotask.ai/glossary/ticket-routing-automation): Ticket routing automation automatically classifies an incoming support or service ticket and assigns it to the right queue, team, or agent based on its content, urgency, and prior history.
- [Tokenization](https://neotask.ai/glossary/tokenization): Tokenization is the process of breaking text or other data into smaller units, called tokens, that a language model or data system processes as its basic unit of input and output.
- [Tokens (AI)](https://neotask.ai/glossary/tokens-ai): Tokens are the discrete units, whether words, subwords, or characters, that a language model reads as input and produces as output, and they are the basis for a model's context limit and usage cost.
- [Tool Calling](https://neotask.ai/glossary/tool-calling): Tool calling is a language model's ability to invoke a predefined external function, such as a search, a database query, or an API request, during its response instead of only generating text.
- [Tool Use (AI)](https://neotask.ai/glossary/tool-use-ai): Tool use is the broader capability of an AI system to decide when, whether, and which external tool to invoke to complete a task, rather than answering purely from its own generated text.
- [Top-p Sampling](https://neotask.ai/glossary/top-p-sampling): Top-p sampling, also called nucleus sampling, restricts a language model's next-token choice to the smallest set of tokens whose cumulative probability reaches a threshold p, then samples only from that set.
- [Training Data](https://neotask.ai/glossary/training-data): Training data is the corpus of examples, whether text, code, images, or other content, that a machine learning model learns statistical patterns from during its training process.
- [Transfer Learning](https://neotask.ai/glossary/transfer-learning): Transfer learning is the technique of taking a model already trained on one large task or dataset and adapting it to a new, related task using far less additional data and compute than training from scratch.
- [Transformer Architecture](https://neotask.ai/glossary/transformer-architecture): The transformer is a neural network architecture built around a self-attention mechanism that lets every token in a sequence directly weigh its relevance to every other token, rather than processing the sequence strictly step by step.
- [Trigger-Action Automation](https://neotask.ai/glossary/trigger-action-automation): Trigger-action automation is a workflow pattern where a defined event, the trigger, automatically causes one or more predefined steps, the actions, to run without a person manually initiating them.
- [Two-Way Sync](https://neotask.ai/glossary/two-way-sync): Two-way sync is a data-integration pattern where changes made in either of two connected systems are automatically propagated to the other, keeping both records consistent in both directions.
- [Unstructured Data](https://neotask.ai/glossary/unstructured-data): Unstructured data is information that does not fit a predefined schema of rows, columns, or fields, such as free-form text, emails, PDFs, images, and audio, and therefore cannot be queried directly the way a relational database table can.
- [Unsupervised Learning](https://neotask.ai/glossary/unsupervised-learning): Unsupervised learning is a machine learning approach in which a model finds patterns, groupings, or structure in data that has no labeled correct answers to learn from.
- [User Provisioning Automation](https://neotask.ai/glossary/user-provisioning-automation): User provisioning automation is the automatic creation, updating, and removal of a person's accounts and access permissions across an organization's systems as their role or employment status changes.
- [Vector Database](https://neotask.ai/glossary/vector-database): A vector database is a data store purpose-built to index and search high-dimensional numerical vectors, typically embeddings produced by an AI model, by similarity rather than by exact match.
- [Vendor Management Automation](https://neotask.ai/glossary/vendor-management-automation): Vendor management automation is the use of software to track vendor onboarding, contracts, risk assessments, and performance across an organization's supplier relationships without manual spreadsheet upkeep.
- [Vertical AI Agent](https://neotask.ai/glossary/vertical-ai-agent): A vertical AI agent is an AI system purpose-built and deeply specialized for one industry or business function, such as legal, healthcare intake, or property management, rather than designed as a general-purpose assistant.
- [Voice Agent](https://neotask.ai/glossary/voice-agent): A voice agent is an AI system that conducts a spoken conversation with a user, converting speech to text, reasoning over it with a language model, and converting the response back to speech in close to real time.
- [Voice Automation](https://neotask.ai/glossary/voice-automation): Voice automation is the use of speech recognition, text-to-speech, and rule- or AI-driven logic to handle spoken interactions, such as phone calls or voice commands, without a human operator on the other end.
- [Vulnerability Management](https://neotask.ai/glossary/vulnerability-management): Vulnerability management is the continuous process of identifying, assessing, prioritizing, and remediating security weaknesses in software and infrastructure before they can be exploited.
- [Warehouse Automation](https://neotask.ai/glossary/warehouse-automation): Warehouse automation is the use of software and physical systems, such as conveyors, robots, and rules-driven agents, to perform inventory, picking, packing, and fulfillment tasks with minimal manual intervention.
- [Webhook](https://neotask.ai/glossary/webhook): A webhook is an HTTP callback where one system automatically sends a real-time notification, usually as a JSON payload, to a URL registered by another system the moment a specific event occurs.
- [Webhook Signature](https://neotask.ai/glossary/webhook-signature): A webhook signature is a cryptographic hash, usually HMAC-SHA256, that a sending system computes over the outgoing payload with a shared secret so the receiver can verify the request genuinely came from that sender and was not tampered...
- [Webinar Automation](https://neotask.ai/glossary/webinar-automation): Webinar automation is the practice of using software to handle the repetitive logistics around a webinar, such as registration confirmations, reminder sequences, attendance tracking, and post-event follow-up, without manual intervention...
- [Workflow Automation](https://neotask.ai/glossary/workflow-automation): Workflow automation is the design and execution of a defined sequence of tasks, decisions, and system interactions so a multi-step business process runs consistently with reduced or no manual handling at each step.
- [Workflow Orchestration](https://neotask.ai/glossary/workflow-orchestration): Workflow orchestration is the coordination layer that manages how multiple independent tasks, services, or agents run together in the correct order, with the right data handed between them, to complete a larger process.
- [Workflow State Machine](https://neotask.ai/glossary/workflow-state-machine): A workflow state machine is a model of a process as a finite set of named states plus the explicit transitions allowed between them, so at any moment the workflow's exact position and next valid moves are unambiguous.
- [Workflow Template](https://neotask.ai/glossary/workflow-template): A workflow template is a reusable, pre-built definition of a multi-step process, with its steps, conditions, and integrations already configured, that a user can instantiate for their own data instead of building the workflow from scratch.
- [Workflow Versioning](https://neotask.ai/glossary/workflow-versioning): Workflow versioning is the practice of tracking distinct, immutable revisions of an automated workflow's definition over time, so changes can be reviewed, rolled back, and safely applied to in-flight executions without breaking processes...
- [XML](https://neotask.ai/glossary/xml): XML, or Extensible Markup Language, is a text-based format that uses nested, custom-named tags to structure and describe hierarchical data in a way that is both human-readable and machine-parseable.
- [YAML](https://neotask.ai/glossary/yaml): YAML, short for "YAML Ain't Markup Language," is a human-readable data serialization format that uses indentation and minimal punctuation to represent structured data such as lists, maps, and scalars.
- [Zero-Shot Learning](https://neotask.ai/glossary/zero-shot-learning): Zero-shot learning is the ability of a machine learning model to perform a task or classify inputs it has never seen labeled examples of during training, relying instead on generalized knowledge learned from other data.
- [Zero-Trust Security](https://neotask.ai/glossary/zero-trust-security): Zero-trust security is a model in which no user, device, or service is trusted by default, regardless of whether it sits inside or outside the network perimeter, and every request must be independently authenticated, authorized, and...

## Skills: AI Models

- [Gemini](https://neotask.ai/skills/gemini): Quick one-shot queries to Google Gemini models
- [Model Usage](https://neotask.ai/skills/model-usage): Track AI model costs and token usage over time
- [Speech-to-Text](https://neotask.ai/skills/openai-whisper): Transcribe audio locally with zero API costs
- [Whisper API](https://neotask.ai/skills/openai-whisper-api): Fast cloud transcription via OpenAI Whisper API
- [Oracle](https://neotask.ai/skills/oracle): Bundle files and prompts to query any AI model

## Skills: Communication

- [BlueBubbles](https://neotask.ai/skills/bluebubbles): Full iMessage control with effects and reactions
- [Discord](https://neotask.ai/skills/discord): Send messages, polls, and reactions in Discord
- [Email (IMAP)](https://neotask.ai/skills/himalaya): Read, compose, and manage email from any provider
- [iMessage](https://neotask.ai/skills/imsg): Send and read iMessages directly from conversation
- [Slack](https://neotask.ai/skills/slack): Send messages and manage channels hands-free
- [Voice Calls](https://neotask.ai/skills/voice-call): Make and manage voice calls programmatically
- [WhatsApp](https://neotask.ai/skills/wacli): Send WhatsApp messages and files to anyone

## Skills: Development

- [Blog Watcher](https://neotask.ai/skills/blogwatcher): Track RSS feeds and get notified of new posts
- [Coding Agent](https://neotask.ai/skills/coding-agent): Delegate coding tasks to Codex, Claude, or Pi
- [Neotask Construct](https://neotask.ai/skills/construct): Find, install and update Neotask plugins and skills
- [GitHub Issues](https://neotask.ai/skills/gh-issues): Auto-fix GitHub issues with parallel AI agents
- [GitHub](https://neotask.ai/skills/github): Manage repos, PRs, and code from conversation
- [Health Check](https://neotask.ai/skills/healthcheck): Audit and harden your system security posture
- [MCP Porter](https://neotask.ai/skills/mcporter): Connect and call any MCP server from one interface
- [Session Logs](https://neotask.ai/skills/session-logs): Search past conversations and track token costs
- [Skill Creator](https://neotask.ai/skills/skill-creator): Build custom skills without writing any code
- [Terminal Multiplexer](https://neotask.ai/skills/tmux): Monitor and control tmux sessions remotely

## Skills: Google Workspace

- [Google Ads](https://neotask.ai/skills/gog-ads): Create campaigns, manage budgets, and analyze advertising performance
- [Google AdSense](https://neotask.ai/skills/gog-adsense): Monitor ad earnings, performance, and site-level revenue analytics
- [Google Business Profile](https://neotask.ai/skills/gog-business): Manage your business listing, reviews, posts, and hours
- [Google Calendar](https://neotask.ai/skills/gog-calendar): Manage events, schedules, and reminders across all your calendars
- [Google Chat](https://neotask.ai/skills/gog-chat): Send messages and manage spaces in Google Chat
- [Google Classroom](https://neotask.ai/skills/gog-classroom): Manage courses, assignments, and student submissions
- [Google Contacts](https://neotask.ai/skills/gog-contacts): Search, create, and manage your Google Contacts and People
- [Google Docs](https://neotask.ai/skills/gog-docs): Create, read, and edit documents with full formatting control
- [Google Drive](https://neotask.ai/skills/gog-drive): Search, organize, and share your files and folders in Drive
- [Google Forms](https://neotask.ai/skills/gog-forms): Build forms, collect responses, and analyze survey data
- [Gmail](https://neotask.ai/skills/gog-gmail): Read, send, and organize your inbox with AI precision
- [Google Keep](https://neotask.ai/skills/gog-keep): Capture, organize, and retrieve notes and checklists
- [Google Meet](https://neotask.ai/skills/gog-meet): Create and manage video meetings and conference links
- [Google Photos](https://neotask.ai/skills/gog-photos): Browse, search, and organize your photo library and albums
- [Google Places](https://neotask.ai/skills/gog-places): Search for businesses, venues, and points of interest with rich details
- [Google Play Developer](https://neotask.ai/skills/gog-play): Manage Android app releases, store listings, and metadata
- [Google Routes](https://neotask.ai/skills/gog-routes): Get directions, travel times, and distance calculations
- [Google Sheets](https://neotask.ai/skills/gog-sheets): Read, write, and compute spreadsheet data with formula support
- [Google Slides](https://neotask.ai/skills/gog-slides): Build and edit presentations slide by slide
- [Google Tasks](https://neotask.ai/skills/gog-tasks): Create, complete, and track to-do items across all your task lists
- [YouTube](https://neotask.ai/skills/gog-youtube): Search videos, manage your channel, playlists, and analytics

## Skills: Marketing

- [Google Ads](https://neotask.ai/skills/google-ads): Create campaigns, manage budgets, and analyze advertising performance
- [Meta Ads](https://neotask.ai/skills/meta-ads): Manage Facebook and Instagram ad campaigns end-to-end

## Skills: Media

- [Camera](https://neotask.ai/skills/camsnap): Capture snapshots and video from network cameras
- [GIF Search](https://neotask.ai/skills/gifgrep): Search, preview, and download GIFs from Tenor and Giphy
- [Banana AI](https://neotask.ai/skills/nano-banana-pro): Generate and edit images with Gemini 3 Pro
- [AI Image Gen](https://neotask.ai/skills/openai-image-gen): Generate and batch-create images with multiple AI models
- [Screenshot](https://neotask.ai/skills/peekaboo): Control your Mac screen, windows, and UI elements
- [Video Creator](https://neotask.ai/skills/remotion-videos): Batch-render videos with parallel processing and upload
- [Text-to-Speech](https://neotask.ai/skills/sherpa-onnx-tts): Generate speech completely offline with zero API costs
- [Music Recognition](https://neotask.ai/skills/songsee): Visualize audio with spectrograms and analysis tools
- [Sonos](https://neotask.ai/skills/sonoscli): Control Sonos speakers and multi-room audio
- [Spotify](https://neotask.ai/skills/spotify-player): Control Spotify playback and search from conversation
- [Video Frames](https://neotask.ai/skills/video-frames): Extract frames from any video at precise timestamps

## Skills: Productivity

- [Apple Notes](https://neotask.ai/skills/apple-notes): Read, write, and search Apple Notes hands-free
- [Apple Reminders](https://neotask.ai/skills/apple-reminders): Create and manage reminders synced to all devices
- [Bear Notes](https://neotask.ai/skills/bear-notes): Create, search, and tag notes in Bear app
- [Canvas](https://neotask.ai/skills/canvas): Display interactive HTML content on any device
- [Google Workspace](https://neotask.ai/skills/gog): All 21 Google services in one unified skill
- [Notion](https://neotask.ai/skills/notion): Search, create, and query your Notion workspace
- [Obsidian](https://neotask.ai/skills/obsidian): Search, create, and organize your Obsidian vault
- [Summarizer](https://neotask.ai/skills/summarize): Summarize articles, PDFs, and YouTube videos instantly
- [Things 3](https://neotask.ai/skills/things-mac): Full task management for Things 3 on Mac
- [Trello](https://neotask.ai/skills/trello): Manage boards, lists, and cards conversationally

## Skills: Utilities

- [1Password](https://neotask.ai/skills/1password): Access secrets and passwords without leaving your workflow
- [Bluetooth](https://neotask.ai/skills/blucli): Control Bluesound and NAD speakers over your network
- [8sleep Control](https://neotask.ai/skills/eightctl): Control your Eight Sleep pod temperature and alarms
- [Food Ordering](https://neotask.ai/skills/food-order): Browse order history and reorder favorite meals
- [Google Places](https://neotask.ai/skills/goplaces): Search nearby places with ratings and reviews
- [PDF Tools](https://neotask.ai/skills/nano-pdf): Edit PDF content using plain English instructions
- [Philips Hue](https://neotask.ai/skills/openhue): Control lights, colors, and scenes in every room
- [Order CLI](https://neotask.ai/skills/ordercli): Manage food delivery orders from the command line
- [System Agent](https://neotask.ai/skills/sag): High-quality text-to-speech with expressive voices
- [Weather](https://neotask.ai/skills/weather): Get current weather and forecasts for any location

## App Integrations: AI & ML

- [Anthropic](https://neotask.ai/integrations/anthropic): Call Claude models, manage prompts, and monitor usage - Neotask orchestrates your Anthropic API.
- [Arize Phoenix](https://neotask.ai/integrations/arize-phoenix): Neotask automates your LLM observability pipeline through Arize Phoenix - monitoring traces, managing prompts, and running experiments so your AI systems stay reliable.
- [Cohere](https://neotask.ai/integrations/cohere): Generate text, rerank results, and build enterprise search - Neotask puts Cohere's NLP suite to work.
- [Composio](https://neotask.ai/integrations/composio): Neotask uses Composio as a universal connector -- Neotask discovers, authenticates, and executes actions across 150+ tools so your agents can automate workflows that span multiple services.
- [Consensus](https://neotask.ai/integrations/consensus): Neotask turns 200M+ academic papers into actionable business intelligence through Consensus -- Neotask searches, filters, and synthesizes scientific research so your decisions are evidence-based.
- [ElevenLabs](https://neotask.ai/integrations/elevenlabs): Generate lifelike speech, clone voices, and manage audio projects - Neotask runs your ElevenLabs workflows.
- [Hugging Face](https://neotask.ai/integrations/hugging-face): Neotask searches models, papers, datasets, and Spaces on the Hugging Face Hub - your AI research and deployment workflow in one conversation.
- [Kagi Search](https://neotask.ai/integrations/kagi-search): Search the web with Kagi's privacy-first engine through conversation - Neotask uses AI agents to deliver clean, ad-free results and AI-powered summaries.
- [Label Studio](https://neotask.ai/integrations/label-studio): Manage data labeling projects and annotations through conversation - Neotask uses AI agents to operate your Label Studio annotation infrastructure.
- [Lilt](https://neotask.ai/integrations/lilt): Neotask translates text and files with AI-powered quality on Lilt - professional localization through conversation, not project portals.
- [OpenAI](https://neotask.ai/integrations/openai): Access OpenAI models, fine-tune datasets, and manage your API usage - Neotask runs it all.
- [Replicate](https://neotask.ai/integrations/replicate): Run ML models in the cloud with a single instruction - Neotask orchestrates your Replicate predictions.
- [Stability AI](https://neotask.ai/integrations/stability-ai): Generate, edit, and transform images with Stable Diffusion - Neotask runs your Stability AI workflows.
- [Weaviate](https://neotask.ai/integrations/weaviate): Search, store, and manage your Weaviate vector database - Neotask handles every operation.

## App Integrations: Analytics

- [Amplitude](https://neotask.ai/integrations/amplitude): Neotask turns Amplitude into an always-active product analytics engine - querying user behavior, running experiments, creating dashboards, and surfacing feedback insights without manual analysis.
- [Audiense Insights](https://neotask.ai/integrations/audiense): Analyze audience intelligence and social segments through Neotask
- [Audiense Insights](https://neotask.ai/integrations/audienseinsights): Neotask brings Audiense Insights audience intelligence into your marketing workflow - research audiences, build reports, and surface consumer insights without leaving the conversation.
- [Datadog](https://neotask.ai/integrations/datadog): Neotask turns Datadog into a conversational command center -- Neotask queries metrics, analyzes logs, investigates traces, and manages feature flags so your ops team resolves incidents faster.
- [Google Analytics](https://neotask.ai/integrations/google-analytics): Neotask turns Google Analytics into plain-English business intelligence -- Neotask queries your traffic data, surfaces insights, and delivers reports so decisions are data-driven without dashboard fatigue.
- [Grafana](https://neotask.ai/integrations/grafana): Neotask turns Grafana into a conversational observability command center - query dashboards, investigate incidents, and manage on-call across 60+ datasources.
- [Heap](https://neotask.ai/integrations/heap): Neotask unlocks Heap's retroactive event data through conversation - Neotask queries your full user history without pre-instrumentation.
- [Hotjar](https://neotask.ai/integrations/hotjar): Neotask surfaces Hotjar heatmap and session insights in plain English - Neotask brings qualitative UX data into your workflow.
- [Microsoft Clarity](https://neotask.ai/integrations/microsoft-clarity): Neotask turns Microsoft Clarity into an always-on UX intelligence feed - querying session data, surfacing heatmap insights, and delivering behavioral analytics without manual dashboard reviews.
- [Mixpanel](https://neotask.ai/integrations/mixpanel): Neotask turns Mixpanel into your always-on product analyst - generate reports, track funnels, and measure feature adoption agents.
- [New Relic](https://neotask.ai/integrations/new-relic): Neotask brings New Relic observability into plain English - Neotask answers your performance and reliability questions without NRQL.
- [Pendo](https://neotask.ai/integrations/pendo): Neotask surfaces user behavior and product feedback from Pendo - analyze feature adoption, retention, and visitor segments through conversation without opening dashboards.
- [Plausible](https://neotask.ai/integrations/plausible): Neotask makes Plausible's privacy-first analytics conversational - Neotask delivers traffic insights without cookies or complexity.
- [PostHog](https://neotask.ai/integrations/posthog): Neotask makes PostHog a continuously self-managing analytics platform - agents create feature flags, monitor experiments, and surface error spikes without manual dashboard work.
- [Power BI](https://neotask.ai/integrations/powerbi): Neotask unlocks your Power BI data - query semantic models, generate DAX, and build data models through conversation without opening Power BI Desktop.
- [Segment](https://neotask.ai/integrations/segment): Neotask makes your Segment customer data instantly queryable - ask anything about events, traits, and journeys using AI agents.
- [SimilarWeb](https://neotask.ai/integrations/similarweb): Neotask unlocks competitive intelligence - analyze any website's traffic, audience, advertising, and SEO performance in plain English.
- [Supermetrics](https://neotask.ai/integrations/supermetrics): Neotask queries all your marketing data through Supermetrics - pulling metrics from 100+ sources into automated reports and analyses without manual exports or dashboard switching.
- [Visier](https://neotask.ai/integrations/visier): Neotask turns Visier into your autonomous people analytics advisor - ask workforce questions in plain English and get answers.
- [Windsor.ai](https://neotask.ai/integrations/windsorai): Neotask connects to Windsor.ai so your agents can pull unified marketing attribution data and answer performance questions across every ad channel in one conversation.
- [Zoho](https://neotask.ai/integrations/zoho): Neotask automates your Zoho Analytics environment - creating reports, querying data, managing workspaces, and monitoring mobile app performance without touching the Zoho dashboard.

## App Integrations: Automation

- [Workato](https://neotask.ai/integrations/workato): Neotask manages your Workato automation platform - build, deploy, and monitor recipes through conversation so your integration infrastructure runs itself.
- [Zapier](https://neotask.ai/integrations/zapiermcp): Neotask connects to Zapier's 7,000+ app ecosystem - configuring your agent to trigger any action from the world's largest automation catalog through a single integration.

## App Integrations: Browser

- [Browser Automation](https://neotask.ai/integrations/browser-automation): Neotask automates any web interaction - navigating sites, filling forms, capturing screenshots, and extracting data so your agent can operate any web application on your behalf.
- [Browserbase](https://neotask.ai/integrations/browserbase): Run cloud browsers, extract web data, and automate browser sessions through conversation - Neotask uses AI agents to give you Browserbase control without writing code.
- [Playwright](https://neotask.ai/integrations/playwright): Write, run, and maintain Playwright tests through conversation - Neotask uses AI agents to make cross-browser test automation fast and reliable.
- [Puppeteer](https://neotask.ai/integrations/puppeteer): Write, run, and debug Puppeteer browser automation scripts through conversation - Neotask uses AI agents to make headless Chrome automation accessible to everyone.

## App Integrations: CRM

- [Close](https://neotask.ai/integrations/close): Neotask runs your entire Close CRM - 48 actions covering leads, contacts, opportunities, and communications, all through conversation.

## App Integrations: Communication

- [Courier](https://neotask.ai/integrations/courier): Neotask runs your notification infrastructure through Courier and Neotask - 60 actions for sending, routing, and managing messages across every channel your users use.
- [DeepL](https://neotask.ai/integrations/deepl): Neotask handles all your translation workflows through DeepL and Neotask - translate documents, rephrase content for tone and style, and manage glossaries through conversation.
- [Discord](https://neotask.ai/integrations/discord): Manage your Discord server, engage your community, and automate moderation - Neotask puts power behind every guild action.
- [Front](https://neotask.ai/integrations/front): Manage shared inboxes, draft replies, and coordinate team email workflows - Neotask uses AI agents to turn Front into a conversational command center.
- [Gmail](https://neotask.ai/integrations/gmail): Neotask automates your Gmail inbox - send emails, triage messages, and manage your communications without ever opening the Gmail UI.
- [Google Chat](https://neotask.ai/integrations/google-chat): Neotask automates your Google Chat operations -- Neotask sends messages, manages spaces, and coordinates team communication so information flows without manual effort.
- [Google Contacts](https://neotask.ai/integrations/google-contacts): Neotask keeps your Google Contacts organized and current -- Neotask creates, searches, and manages contacts so your address book becomes a living, queryable business asset.
- [Google Meet](https://neotask.ai/integrations/google-meet): Neotask schedules, configures, and manages Google Meet sessions - your meetings run on autopilot while you focus on the conversation.
- [Infobip](https://neotask.ai/integrations/infobip): Neotask sends messages across SMS, WhatsApp, email, and voice through Infobip on Neotask - omnichannel customer communication without touching multiple dashboards.
- [Intercom](https://neotask.ai/integrations/intercom): Neotask manages your Intercom inbox, contacts, and support workflows - customer conversations stay organized without clicking through the dashboard.
- [Knock](https://neotask.ai/integrations/knock): Neotask builds and manages your Knock notification infrastructure - 47 actions for creating workflows, managing users, and orchestrating notifications through conversation.
- [LINE](https://neotask.ai/integrations/line): Neotask manages your LINE bot, rich menus, and audience messaging - engage your LINE followers without building custom integrations.
- [Loom](https://neotask.ai/integrations/loom): Organize, share, and extract insights from your Loom video library - Neotask uses AI agents to make every recording more actionable.
- [Mailgun](https://neotask.ai/integrations/mailgun): Neotask sends emails, tracks delivery, and validates addresses through Mailgun on Neotask - transactional email operations without the dashboard.
- [Mailjet](https://neotask.ai/integrations/mailjet): Neotask monitors your Mailjet email performance - agents surface deliverability issues, campaign underperformers, and contact list health without you pulling reports manually.
- [Microsoft Teams](https://neotask.ai/integrations/microsoft-teams): Neotask brings autonomous team communication to Microsoft Teams - post messages, manage channels, and coordinate your organization agents.
- [SendGrid](https://neotask.ai/integrations/sendgrid): Send transactional emails, manage campaigns, and track delivery performance - Neotask uses AI agents to make SendGrid as easy as a conversation.
- [Slack](https://neotask.ai/integrations/slack): Neotask brings autonomous intelligence to every Slack channel - send messages, search conversations, manage canvases, and operate your workspace.
- [Telegram](https://neotask.ai/integrations/telegram): Automate Telegram messaging, channel publishing, and bot interactions - Neotask keeps your Telegram presence active and responsive.
- [Twilio](https://neotask.ai/integrations/twilio): Send SMS, make calls, and manage your communications infrastructure - Neotask puts intelligence behind every Twilio interaction.
- [Webex](https://neotask.ai/integrations/webex): Send messages, manage Webex spaces, and control meetings through conversation - Neotask uses AI agents to operate Webex without switching apps.
- [WhatsApp](https://neotask.ai/integrations/whatsapp): Send WhatsApp messages, manage business communications, and automate customer touchpoints - Neotask uses AI agents to keep every conversation on time.
- [Zoom](https://neotask.ai/integrations/zoom): Schedule, manage, and recap Zoom meetings through natural conversation - Neotask handles your video conferencing logistics with Neotask precision.

## App Integrations: Communications

- [ClickSend](https://neotask.ai/integrations/clicksend): Neotask puts SMS outreach on autopilot through ClickSend - sending messages, checking delivery stats, and managing contact lists without manual effort.

## App Integrations: Data

- [10x Genomics](https://neotask.ai/integrations/10xgenomics): Neotask connects to 10x Genomics so computational biology teams can access datasets, workflows, and genomics analysis resources through natural conversation.
- [Apify](https://neotask.ai/integrations/apify): Neotask runs web scraping and data extraction at scale through Apify and Neotask - discover actors, run scrapes, and retrieve structured datasets through conversation.
- [Astra DB](https://neotask.ai/integrations/astra-db): Neotask keeps your Astra DB running smoothly - managing collections, syncing records, and executing vector searches so your data infrastructure works without manual queries.
- [Atlan](https://neotask.ai/integrations/atlan): Neotask automates your Atlan data governance - cataloging assets, tracing lineage, and enforcing quality rules so your data team focuses on insights instead of housekeeping.
- [Microsoft Azure](https://neotask.ai/integrations/azure): Neotask turns Microsoft Azure into your autonomous cloud operations platform - manage compute, storage, networking, databases, and AI services across your entire Azure estate through conversation.
- [Benchling](https://neotask.ai/integrations/benchling): Neotask connects to Benchling so life sciences teams can access research data, manage entities, and integrate lab operations into their automation stack through natural conversation.
- [Bright Data](https://neotask.ai/integrations/brightdata): Neotask gives your business real-time web intelligence through Bright Data and Neotask - 68 actions for scraping, social media data, e-commerce monitoring, and browser automation.
- [CData Connect](https://neotask.ai/integrations/cdataconnect): Neotask turns CData Connect into your universal data layer - querying 200+ data sources, running stored procedures, and exploring schemas without writing SQL or opening a query editor.
- [Chroma](https://neotask.ai/integrations/chroma): Neotask automates your Chroma vector database - managing collections, indexing documents, and running semantic searches so your AI applications stay powered with fresh data.
- [Couchbase](https://neotask.ai/integrations/couchbase): Neotask makes your Couchbase database conversational -- Neotask runs queries, manages documents, and explores schemas so your team gets answers from data without writing SQL++.
- [Coupler.io](https://neotask.ai/integrations/couplerio): Neotask queries your Coupler.io dataflows - pulling live data from 400+ connected sources and making it available to any automated business analysis or reporting workflow.
- [Databricks](https://neotask.ai/integrations/databricks): Neotask makes your Databricks lakehouse conversational -- Neotask queries Unity Catalog, runs SQL, and orchestrates data workflows so your analysts get answers without writing notebooks.
- [Edubase](https://neotask.ai/integrations/edubase): Neotask turns Edubase into your autonomous learning operations platform - manage courses, users, assessments, certificates, and the complete e-learning infrastructure through natural conversation.
- [FalkorDB](https://neotask.ai/integrations/falkordb): Neotask makes your FalkorDB graph database conversational -- Neotask runs graph queries, manages knowledge graphs, and searches entity relationships so your connected data becomes actionable.
- [Meilisearch](https://neotask.ai/integrations/meilisearch): Neotask manages your Meilisearch instance - configure indexes, search your data, and maintain your search infrastructure through conversation.
- [Milvus](https://neotask.ai/integrations/milvus): Neotask connects to your Milvus vector database - search, index, and manage embeddings through natural conversation instead of writing query code.
- [MongoDB](https://neotask.ai/integrations/mongodb): Neotask gives you conversational control over MongoDB and Atlas - query collections, run aggregations, and manage your database infrastructure.
- [MotherDuck](https://neotask.ai/integrations/motherduck): Neotask puts a conversational interface on your MotherDuck cloud analytics - query DuckDB databases, manage Dives, and share analytical work through conversation.
- [Neon](https://neotask.ai/integrations/neon): Neotask manages your Neon Postgres databases - run SQL, create branches, tune queries, and manage projects without touching a terminal.
- [Open Targets](https://neotask.ai/integrations/opentargets): Neotask brings Open Targets biomedical intelligence into your research workflow - explore drug targets, diseases, and genetic evidence in plain English.
- [Pinecone](https://neotask.ai/integrations/pinecone): Neotask manages your Pinecone vector infrastructure - search records, manage indexes, and build AI retrieval pipelines without writing code.
- [PubMed](https://neotask.ai/integrations/pubmed): Neotask gives your research agents access to PubMed's 35+ million biomedical citations - search literature, retrieve abstracts, and automate evidence gathering through conversation.
- [Snowflake](https://neotask.ai/integrations/snowflake): Neotask puts a conversational interface on your Snowflake data warehouse - query data, manage objects, and run Cortex AI analytics without writing SQL.
- [Supabase](https://neotask.ai/integrations/supabase): Neotask gives you full conversational control over Supabase - manage databases, deploy edge functions, handle storage, and operate your entire backend.
- [Synapse](https://neotask.ai/integrations/synapseorg): Neotask connects to Synapse.org so research teams can access open biomedical datasets, manage projects, and collaborate on data-driven science through natural conversation.

## App Integrations: Data & Analytics

- [ClickHouse](https://neotask.ai/integrations/clickhouse): Neotask turns ClickHouse into a conversational analytics engine - querying billions of events, exploring schemas, and surfacing insights without writing SQL manually.
- [Elasticsearch](https://neotask.ai/integrations/elasticsearch): Neotask makes Elasticsearch's search and analytics capabilities conversational - exploring indices, running queries, and executing ES|QL to surface insights from your data without leaving your workflow.
- [Hex](https://neotask.ai/integrations/hex): Neotask brings Hex's collaborative data workspace into your automation layer - finding notebooks, starting analysis threads, and continuing data conversations without switching tabs.
- [Neo4j](https://neotask.ai/integrations/neo4j): Neotask makes graph database operations conversational through Neo4j - exploring schemas, reading relationships, and writing Cypher queries to unlock connected data insights.
- [Qdrant](https://neotask.ai/integrations/qdrant): Neotask adds vector memory to your automation stack via Qdrant - storing knowledge and retrieving semantically similar content so your agents learn and recall like humans.
- [Starburst](https://neotask.ai/integrations/starburst): Neotask makes your Starburst data lakehouse conversational - running read-only SQL queries and engaging Starburst's data product AI agents through a single business interface.

## App Integrations: Databases

- [BigQuery](https://neotask.ai/integrations/bigquery): Analyze petabyte-scale data, manage costs, and run BigQuery jobs - Neotask brings AI agents to Google's data warehouse.
- [DynamoDB](https://neotask.ai/integrations/dynamodb): Query, scan, and manage your DynamoDB tables through conversation - Neotask brings AI agents to your AWS data layer.
- [PostgreSQL](https://neotask.ai/integrations/postgresql): Query, optimize, and manage your PostgreSQL databases through natural conversation - and Neotask.
- [Redis](https://neotask.ai/integrations/redis): Inspect keys, analyze memory, and manage your Redis instances through natural conversation - Neotask runs on Neotask.

## App Integrations: Design

- [21st.dev Magic](https://neotask.ai/integrations/21st-dev): Generate polished UI components instantly through Neotask
- [Adobe XD](https://neotask.ai/integrations/adobe-xd): Neotask brings access to your Adobe XD files - inspect specs, track shares, and extract design data through natural conversation.
- [BioRender](https://neotask.ai/integrations/biorender): Neotask connects to BioRender so research teams can create and manage scientific figures and illustrations through conversation, accelerating publication and presentation workflows.
- [Canva](https://neotask.ai/integrations/canva): Neotask keeps your Canva design workflow running smoothly - managing designs, organizing assets, and coordinating feedback so your creative team moves faster.
- [FigJam](https://neotask.ai/integrations/figjam): Neotask uses AI agents to access your FigJam boards - retrieve sticky content, manage sharing, and extract workshop output through natural conversation.
- [Figma](https://neotask.ai/integrations/figma): Neotask bridges your Figma designs and your codebase - agents extract design context, map components to code, and generate diagrams without manual design handoff.
- [Framer](https://neotask.ai/integrations/framer): Neotask brings capability to your Framer projects - access design files, retrieve publish links, and inspect components through conversation.
- [InVision](https://neotask.ai/integrations/invision): Neotask uses AI agents to navigate your InVision prototypes, track feedback, and manage project sharing - all through natural conversation.
- [Magic Patterns](https://neotask.ai/integrations/magicpatterns): Neotask generates UI designs, manages design systems, and iterates on components - design-to-code through conversation.
- [Penpot](https://neotask.ai/integrations/penpot): Neotask bridges the gap between design and development in Penpot - executing code, exporting assets, and making your open-source design system programmable through conversation.
- [Sketch](https://neotask.ai/integrations/sketch): Neotask connects to your Sketch workspace - retrieve design specs, manage shared documents, and inspect components through conversation.
- [Zeplin](https://neotask.ai/integrations/zeplin): Neotask puts intelligence into your Zeplin workflow - retrieve specs, manage handoff notes, and track screen status through conversation.

## App Integrations: DevOps

- [Bitrise](https://neotask.ai/integrations/bitrise): Neotask manages your Bitrise mobile CI/CD operations - 67 actions covering builds, apps, artifacts, tester groups, and release management through conversation.
- [Buildkite](https://neotask.ai/integrations/buildkite): Neotask monitors and manages your Buildkite CI/CD pipelines - surface build failures, search logs, and trigger rebuilds through conversation without dashboard navigation.
- [Vercel](https://neotask.ai/integrations/vercel): Neotask manages your Vercel deployments - monitor build logs, trigger deployments, buy domains, and debug production issues through conversation.

## App Integrations: Developer Tools

- [AgentOps](https://neotask.ai/integrations/agentops): Neotask gives you full observability over every AI agent run - trace what happened, debug why it failed, and fix problems before they affect your users.
- [Aiven](https://neotask.ai/integrations/aiven): Neotask manages your Aiven cloud data infrastructure - listing projects, inspecting services, and keeping your managed databases and streaming pipelines visible and under control.
- [Atono](https://neotask.ai/integrations/atono): Neotask keeps your Atono product delivery engine running automatically - managing stories, bugs, timeboxes, and team context so engineers stay in flow.
- [Buildable](https://neotask.ai/integrations/buildable): Neotask drives your Buildable AI-powered project management - automatically picking up tasks, tracking progress, completing work, and keeping your development momentum going without manual status updates.
- [Context7](https://neotask.ai/integrations/context7): Neotask keeps your agent's library knowledge current through Context7 - resolving package IDs and fetching up-to-date documentation so your agents code with today's APIs, not yesterday's.
- [Drata](https://neotask.ai/integrations/drata): Neotask makes compliance monitoring conversational - letting your agents query Drata for controls, risks, policies, and monitor test results without leaving your workflow.
- [Dynatrace](https://neotask.ai/integrations/dynatrace): Neotask turns Dynatrace into an autonomous observability engine - querying metrics, analyzing problems, detecting vulnerabilities, and triggering notifications so your infrastructure runs itself.
- [E2B](https://neotask.ai/integrations/e2b): Neotask executes code in secure E2B cloud sandboxes - letting your agents run Python analysis, data processing, and automation scripts safely without any local setup.
- [GitGuardian](https://neotask.ai/integrations/gitguardian): Neotask automates your GitGuardian security operations - scanning for secrets, managing incidents, generating honeytokens, and keeping your code clean without manual security reviews.
- [JFrog](https://neotask.ai/integrations/jfrog): Neotask manages your JFrog platform - creating repositories, scanning for vulnerabilities, monitoring builds, and keeping your software supply chain secure without manual platform work.
- [Label Studio](https://neotask.ai/integrations/labelstudio): Neotask automates your Label Studio data labeling pipeline - managing projects, importing tasks, creating predictions, and keeping your annotation operation running without manual overhead.
- [Microsoft Learn](https://neotask.ai/integrations/microsoftlearn): Neotask makes Microsoft documentation instantly accessible - searching, fetching, and surfacing code samples so your team spends less time in docs and more time building.
- [Nango](https://neotask.ai/integrations/nango): Neotask executes your Nango-connected actions automatically - triggering any integration action your team has configured across 300+ APIs through a single agent interface.
- [Netdata](https://neotask.ai/integrations/netdata): Neotask gives your infrastructure a voice - querying Netdata metrics, detecting anomalies, correlating root causes, and managing alerts without ever opening a monitoring dashboard.
- [Netlify](https://neotask.ai/integrations/netlify): Neotask manages your Netlify deployments, projects, and extensions - so your web infrastructure gets created, configured, and deployed through conversation instead of dashboards.
- [Nutrient](https://neotask.ai/integrations/nutrient): Neotask automates your document processing with Nutrient - merging, converting, redacting, and signing PDFs at scale without a single manual export.

## App Integrations: Development

- [AWS](https://neotask.ai/integrations/aws): Neotask automates your AWS EKS operations - managing clusters, troubleshooting infrastructure, and maintaining IAM policies so your Kubernetes runs reliably.
- [Azure DevOps](https://neotask.ai/integrations/azure-devops): Neotask automates your Azure DevOps workflow - managing work items, running pipelines, and maintaining wikis so your development process runs without constant human orchestration.
- [Bitbucket](https://neotask.ai/integrations/bitbucket): Navigate your Bitbucket repositories and pipelines through conversation - Neotask keeps your team moving.
- [CircleCI](https://neotask.ai/integrations/circleci): Neotask automates your CircleCI operations - running pipelines, diagnosing failures, and managing deployments so your CI/CD runs reliably without constant oversight.
- [Cloudflare](https://neotask.ai/integrations/cloudflare): Neotask uses Cloudflare's browser automation to run your web operations on autopilot -- Neotask navigates sites, fills forms, captures data, and tests your pages without a human at the keyboard.
- [ConfigCat](https://neotask.ai/integrations/configcat): Neotask automates your feature flag lifecycle through ConfigCat -- Neotask creates flags, manages rollouts, audits changes, and cleans up zombie flags so your releases run smoother.
- [DevCycle](https://neotask.ai/integrations/devcycle): Neotask automates your DevCycle feature management workflow -- Neotask creates flags, manages variables, tracks evaluations, and handles SDK setup so your team ships features safely.
- [DevRev](https://neotask.ai/integrations/devrev): Neotask brings DevRev's product development platform into your workflow -- Neotask manages work items, sprint planning, and customer feedback so your team builds what matters.
- [Dice](https://neotask.ai/integrations/dice): Neotask turns Dice into your AI recruiting assistant -- Neotask searches tech jobs, filters by criteria, and surfaces salary data so your hiring or job search runs faster.
- [Docker](https://neotask.ai/integrations/docker): Control containers, images, and networks through plain language - Neotask manages your Docker environment without memorizing CLI flags.
- [Firebase](https://neotask.ai/integrations/firebase): Neotask gives you conversational control over your entire Firebase backend -- Neotask manages Firestore, Auth, security rules, and project configuration so your development team ships faster.
- [Firecrawl](https://neotask.ai/integrations/firecrawl): Neotask uses Firecrawl to extract structured business data from any website -- Neotask scrapes, crawls, and transforms web content into actionable intelligence for your operations.
- [Gitea](https://neotask.ai/integrations/gitea): Neotask automates your self-hosted Gitea workflow -- Neotask manages repos, branches, issues, and pull requests so your team's development operations run without manual overhead.
- [GitHub](https://neotask.ai/integrations/github): Neotask supercharges your GitHub workflow -- Neotask manages issues, creates pull requests, monitors Actions, and reads code so your development operations run on autopilot.
- [GitLab](https://neotask.ai/integrations/gitlab): Neotask brings your entire GitLab DevOps lifecycle into conversation -- Neotask manages issues, merge requests, pipelines, and code search so your team delivers faster.
- [Google Play Developer](https://neotask.ai/integrations/google-play-developer): Neotask manages your Android app listings, releases, and store performance - Play Console operations without the dashboard.
- [Jam](https://neotask.ai/integrations/jam): Neotask pulls bug reports, console logs, network requests, and video recordings from Jam - every bug report becomes actionable without context-switching.
- [Jenkins](https://neotask.ai/integrations/jenkins): Neotask triggers builds, inspects pipelines, and searches build logs - Jenkins CI/CD without the dashboard.
- [JetBrains](https://neotask.ai/integrations/jetbrains): Neotask extends your JetBrains IDE with natural language - navigate code, run configurations, and refactor without leaving your conversation.
- [LaunchDarkly](https://neotask.ai/integrations/launchdarkly): Neotask controls feature flags, queries observability data, and manages rollouts - ship features safely without the LaunchDarkly dashboard.
- [Liveblocks](https://neotask.ai/integrations/liveblocks): Neotask manages real-time collaboration infrastructure on Liveblocks - rooms, threads, notifications, and storage through conversation.
- [MinIO](https://neotask.ai/integrations/minio): Manage self-hosted MinIO object storage through conversation - Neotask runs on Neotask to give your private S3-compatible infrastructure the same natural-language control as a managed cloud.
- [PlanetScale](https://neotask.ai/integrations/planetscale): Branch databases, run schema changes, and monitor query performance through conversation - Neotask uses AI agents to bring PlanetScale's branching model into your workflow.
- [Postman](https://neotask.ai/integrations/postman): Neotask integrates Postman's API platform into your development workflow - manage collections, explore APIs, and run requests conversationally.
- [Railway](https://neotask.ai/integrations/railway): Deploy and manage Railway services through conversation - Neotask makes infrastructure as easy as asking.
- [Sentry](https://neotask.ai/integrations/sentry): Neotask turns Sentry into your autonomous error triage system - search issues, inspect events, and resolve bugs without context-switching.
- [Webflow](https://neotask.ai/integrations/webflow): Neotask manages your Webflow sites - update pages, manage CMS collections, and control design elements without the Webflow Designer.
- [Wix](https://neotask.ai/integrations/wix): Neotask manages your Wix sites and business operations - update content, manage your store, and call APIs without the Wix editor.
- [WordPress](https://neotask.ai/integrations/wordpress): Neotask manages your WordPress sites - publish posts, manage plugins, track statistics, and control your entire site without the admin panel.

## App Integrations: Education

- [10x Genomics Cloud](https://neotask.ai/integrations/10x-genomics): Automate 10x Genomics Cloud analysis through Neotask
- [Coursera](https://neotask.ai/integrations/coursera): Search, enroll, and track your learning on Coursera through conversation - Neotask uses AI agents to make professional development effortless.
- [Duolingo](https://neotask.ai/integrations/duolingo): Supercharge your Duolingo practice through conversation - Neotask uses AI agents to help you learn languages faster and smarter.
- [Google Classroom](https://neotask.ai/integrations/google-classroom): Neotask automates Google Classroom administration -- Neotask manages courses, assignments, and grading so teachers spend less time on busywork and more time teaching.
- [Khan Academy](https://neotask.ai/integrations/khan-academy): Navigate Khan Academy's free curriculum through conversation - Neotask uses AI agents to personalize your learning path across every subject.
- [Udemy](https://neotask.ai/integrations/udemy): Find, evaluate, and track Udemy courses through natural conversation - Neotask makes navigating the world's largest course marketplace simple.

## App Integrations: Finance

- [Adfin](https://neotask.ai/integrations/adfin): Neotask connects to Adfin so your agents can create payment links, manage statements, and handle financial operations through natural conversation.
- [Aiera](https://neotask.ai/integrations/aiera): Neotask turns Aiera into your always-on financial research analyst - automatically surfacing earnings data, filings, and transcripts so you never miss a material event.
- [Airwallex](https://neotask.ai/integrations/airwallex): Neotask automates your global treasury operations through Airwallex - managing accounts, FX, transfers, and billing without touching a dashboard.
- [Alpaca](https://neotask.ai/integrations/alpaca): Neotask connects to Alpaca Markets so your AI agent monitors portfolios, tracks market data, and manages watchlists while you focus on strategy.
- [Alpha Vantage](https://neotask.ai/integrations/alphavantage): Neotask taps into Alpha Vantage so your AI agent delivers real-time market data, technical analysis, and financial research - automatically and on schedule.
- [Brex](https://neotask.ai/integrations/brex): Neotask uses AI agents to manage your Brex cards, expenses, and spend limits - corporate card control through plain conversation.
- [Chargebee](https://neotask.ai/integrations/chargebee): Neotask manages your Chargebee billing operations - look up customers, subscriptions, and invoices, and configure your product catalog through conversation.
- [Chronograph](https://neotask.ai/integrations/chronograph): Neotask brings Chronograph private markets intelligence into your investment workflow - query fund performance, portfolio data, and exposure analytics through natural conversation.
- [CoinGecko](https://neotask.ai/integrations/coingecko): Neotask turns CoinGecko into your autonomous crypto market analyst -- Neotask monitors prices, tracks trends, and surfaces opportunities across 15,000+ tokens without manual research.
- [Commerce Layer](https://neotask.ai/integrations/commerce-layer): Neotask makes your Commerce Layer store data accessible - 33 analytics actions covering orders, carts, and returns so you always know what's happening in your commerce operation.
- [Crypto.com](https://neotask.ai/integrations/cryptocom): Neotask brings Crypto.com market data and trading intelligence into your workflow - query prices, analyze markets, and monitor positions through natural conversation.
- [Daloopa](https://neotask.ai/integrations/daloopa): Neotask automates fundamental financial research through Daloopa - discovering companies, pulling clean time-series data, and surfacing analyst documents without manual data entry.
- [Flutterwave](https://neotask.ai/integrations/flutterwave): Neotask turns Flutterwave into your autonomous African payments engine - initiate transfers, manage virtual accounts, verify transactions, and handle payouts without leaving the conversation.
- [LSEG](https://neotask.ai/integrations/lseg): Neotask accesses institutional-grade financial data across equities, bonds, FX, and derivatives - LSEG market intelligence through conversation.
- [LunarCrush](https://neotask.ai/integrations/lunarcrush): Neotask connects to LunarCrush so your agents can surface crypto social sentiment, influencer activity, and market intelligence across thousands of coins in real time.
- [Mercury](https://neotask.ai/integrations/mercury): Neotask turns Mercury into your autonomous treasury department - monitor balances, send payments, and manage recipients without logging into the banking dashboard.
- [Morningstar](https://neotask.ai/integrations/morningstar): Neotask turns Morningstar's investment data into an automated research operation - looking up securities, pulling fundamentals, running analyst research, and screening investments on demand.
- [MT Newswires](https://neotask.ai/integrations/mtnewswires): Neotask brings MT Newswires financial news intelligence into your investment workflow - monitor market-moving headlines, earnings updates, and economic data in real time.
- [NetSuite](https://neotask.ai/integrations/netsuite): Neotask puts Oracle NetSuite at your fingertips - manage records, run reports, and query your entire ERP without ever touching the dashboard.
- [PayPal](https://neotask.ai/integrations/paypal): Neotask turns PayPal into your autonomous payments operations platform - process transactions, manage invoices, handle disputes, and run your complete payment stack through conversation.
- [Paystack](https://neotask.ai/integrations/paystack): Neotask automates your Paystack payment operations - process transactions, manage subscriptions, and handle refunds for African markets.
- [Pigment](https://neotask.ai/integrations/pigment): Neotask connects your financial planning workflows to Pigment - querying metrics, pulling data, and triggering AI agent missions to automate your FP&A operations.
- [PitchBook](https://neotask.ai/integrations/pitchbookpremium): Neotask unlocks PitchBook Premium data - search companies, investors, deals, and funds to power your market research and competitive intelligence through natural conversation.
- [Plaid](https://neotask.ai/integrations/plaid): Neotask connects to Plaid so you can query real bank data, categorize transactions, and monitor balances through natural conversation.
- [QuickBooks](https://neotask.ai/integrations/quickbooks): Neotask runs on Neotask to make QuickBooks feel like talking to your bookkeeper - invoices, expenses, and P&L in plain English.
- [Ramp](https://neotask.ai/integrations/ramp): Neotask gives you autonomous control over Ramp - manage cards, track spending, handle reimbursements, and oversee your entire corporate finance operation through conversation.
- [Razorpay](https://neotask.ai/integrations/razorpay): Neotask turns Razorpay into your autonomous payment operations center - manage payments, orders, refunds, settlements, and QR codes.
- [Shopify](https://neotask.ai/integrations/shopify): Neotask connects your Shopify storefront and customer-account MCP flows through one AI-powered Shopify app - search products, answer policy questions, manage carts, and unlock customer-specific actions when authenticated.
- [S&P Global](https://neotask.ai/integrations/spglobal): Neotask brings S&P Global's authoritative financial data into your investment and risk workflows - access ratings, market intelligence, and company fundamentals through conversation.
- [Square](https://neotask.ai/integrations/square): Neotask connects to Square so your commerce operations run autonomously - manage customers, orders, catalog, payments, and staff through conversation.
- [Stripe](https://neotask.ai/integrations/stripe): Neotask runs your Stripe payments, subscriptions, and billing - your entire payment infrastructure managed by autonomous agents.
- [Xero](https://neotask.ai/integrations/xero): Neotask uses AI agents to turn your Xero account into a conversational finance tool - reports, reconciliation, and invoices without the clicks.

## App Integrations: HR & Recruiting

- [Indeed](https://neotask.ai/integrations/indeed): Neotask makes talent research continuous through Indeed - searching jobs, analyzing company data, and accessing candidate profiles to support hiring decisions automatically.
- [Udemy Business](https://neotask.ai/integrations/udemybusiness): Neotask makes enterprise learning intelligent - searching Udemy Business courses, curricula, and labs so your team finds the right training for any skill gap instantly.

## App Integrations: Infrastructure

- [Ansible](https://neotask.ai/integrations/ansible): Run playbooks, check inventory, and automate configuration - Neotask pairs AI agents with Ansible for hands-free ops.
- [AWS Marketplace](https://neotask.ai/integrations/awsmarketplace): Neotask automates your software procurement through AWS Marketplace - researching solutions, generating comparison reports, and making vendor evaluation a conversation instead of a project.
- [Axiom](https://neotask.ai/integrations/axiom): Neotask automates your Axiom observability stack - querying logs, managing dashboards, and monitoring alerts so your team focuses on building instead of watching graphs.
- [Chef](https://neotask.ai/integrations/chef): Manage cookbooks, run knife commands, and audit node compliance - Neotask runs your Chef infrastructure.
- [Confluent](https://neotask.ai/integrations/confluent): Neotask gives you conversational control over your entire Confluent Cloud infrastructure -- Neotask manages Kafka topics, Flink pipelines, connectors, and TableFlow so your data streams never stall.
- [Heroku](https://neotask.ai/integrations/heroku): Neotask deploys, scales, and manages your entire Heroku platform - from dyno scaling to Postgres administration, all through conversation.
- [Honeycomb](https://neotask.ai/integrations/honeycomb): Neotask investigates production issues and analyzes distributed traces on Honeycomb - observability through conversation, not dashboards.
- [Kubernetes](https://neotask.ai/integrations/kubernetes): Manage clusters, debug workloads, and scale services - Neotask makes Kubernetes conversational with Neotask.
- [PagerDuty](https://neotask.ai/integrations/pagerduty): Neotask turns PagerDuty into an autonomous incident command center - manage incidents, schedules, escalation policies, and status pages agents.
- [Port](https://neotask.ai/integrations/port): Neotask manages your internal developer portal - operate blueprints, entities, scorecards, and self-service actions without the Port UI.
- [Prometheus](https://neotask.ai/integrations/prometheus): Query metrics, investigate alerts, and understand your system's health - Neotask speaks PromQL.
- [Pulumi](https://neotask.ai/integrations/pulumi): Preview, deploy, and manage cloud infrastructure in natural language - Neotask brings AI agents to Pulumi.
- [Terraform](https://neotask.ai/integrations/terraform): Neotask manages your Terraform Cloud infrastructure - operate workspaces, stacks, and policy sets without touching the CLI or dashboard.

## App Integrations: Legal

- [Backdocket](https://neotask.ai/integrations/backdocket): Neotask connects your legal workflow to Backdocket's matter management system - searching cases, retrieving details, and executing API operations through intelligent agents.

## App Integrations: Marketing

- [ActiveCampaign](https://neotask.ai/integrations/activecampaign): Neotask runs your email marketing, CRM pipeline, and automation workflows on autopilot - so your business nurtures leads while you sleep.
- [Ahrefs](https://neotask.ai/integrations/ahrefs): Neotask runs your SEO operations with 61 Ahrefs actions - competitive analysis, keyword research, backlink audits, and brand monitoring all through conversation.
- [AirOps](https://neotask.ai/integrations/airops): Neotask automates your AirOps content strategy - managing brand kits, tracking AI search visibility, and generating reports while your team focuses on creative work.
- [Beehiiv](https://neotask.ai/integrations/beehiiv): Neotask uses AI agents to give newsletter operators a conversational window into Beehiiv - subscribers, posts, and monetization data without the dashboard.
- [Bitly](https://neotask.ai/integrations/bitly): Neotask automates your Bitly link ecosystem - creating short links, generating QR codes, and pulling analytics so your marketing team tracks everything without the manual work.
- [Brevo](https://neotask.ai/integrations/brevo): Run email campaigns, manage contacts, and track performance - Neotask handles your Brevo marketing.
- [BuiltWith](https://neotask.ai/integrations/builtwith): Neotask turns BuiltWith into your automated competitive intelligence engine - profiling tech stacks, discovering prospects, and tracking market trends at scale.
- [ConvertKit](https://neotask.ai/integrations/convertkit): Neotask connects to ConvertKit so creators can manage subscribers, sequences, and forms through conversation instead of clicks.
- [Customer.io](https://neotask.ai/integrations/customer): Neotask automates your Customer.io messaging operations -- Neotask manages campaigns, segments, and customer profiles so your marketing runs itself while you focus on strategy.
- [Customer.io](https://neotask.ai/integrations/customer-io): Neotask uses AI agents to bring Customer.io's behavioral data and campaign logic into conversation - segments, campaigns, and person lookups at your fingertips.
- [Drip](https://neotask.ai/integrations/drip): Neotask uses AI agents to query Drip's ecommerce marketing data - workflows, revenue attribution, and customer behavior through natural conversation.
- [Google Ads](https://neotask.ai/integrations/google-ads): Neotask automates your Google Ads management -- Neotask optimizes campaigns, adjusts budgets, and monitors performance so your ad spend works harder without constant manual oversight.
- [Google AdSense](https://neotask.ai/integrations/google-adsense): Neotask automates your AdSense revenue monitoring -- Neotask tracks earnings, analyzes ad performance, and surfaces optimization opportunities so you earn more from your content.
- [Google Business Profile](https://neotask.ai/integrations/google-business-profile): Neotask keeps your Google Business Profile working for you around the clock -- Neotask updates listings, responds to reviews, and posts updates so your local presence stays strong without daily manual attention.
- [Klaviyo](https://neotask.ai/integrations/klaviyo): Neotask runs your Klaviyo email and SMS marketing autonomously - managing campaigns, flows, profiles, and segments so your customer lifecycle marketing operates without constant manual attention.
- [Mailchimp](https://neotask.ai/integrations/mailchimp): Neotask uses AI agents to manage your Mailchimp lists, campaigns, and automations through natural conversation - no clicks required.
- [Sprouts](https://neotask.ai/integrations/sprouts): Neotask automates your Sprouts lead prospecting - searching and enriching leads so your outbound pipeline fills without manual research.

## App Integrations: Media

- [AllVoiceLab](https://neotask.ai/integrations/allvoicelab): Neotask automates your voice production pipeline through AllVoiceLab - generating speech, cloning voices, and dubbing videos without a production studio.
- [Cloudinary](https://neotask.ai/integrations/cloudinary): Neotask manages your entire Cloudinary media library - upload, transform, organize, and deliver assets automatically as part of your content operations.
- [Datawrapper](https://neotask.ai/integrations/datawrapper): Create and publish data visualizations through Neotask - charts, maps, and tables for editorial and business use through conversation.
- [Google Photos](https://neotask.ai/integrations/google-photos): Neotask searches, organizes, and manages your Google Photos library - find the right image, build albums, and share assets without scrolling.
- [Kaltura](https://neotask.ai/integrations/kaltura): Neotask creates and manages virtual events and sessions on Kaltura - your event calendar runs itself while you focus on content.
- [Mux](https://neotask.ai/integrations/mux): Manage video streaming infrastructure through natural conversation - Neotask runs on Neotask to give you API-level control over Mux without writing code.
- [Spotify](https://neotask.ai/integrations/spotify): Control your Spotify playback, build playlists, and discover music - all through conversation powered by Neotask.
- [Vimeo](https://neotask.ai/integrations/vimeo): Upload, organize, and share professional video through natural conversation - Neotask uses AI agents to give you full command of your Vimeo library.
- [YouTube](https://neotask.ai/integrations/youtube): Search, analyze, and manage YouTube content through natural conversation - Neotask runs on Neotask to give you real control over the platform.

## App Integrations: Other

- [Aqara](https://neotask.ai/integrations/aqara): Neotask turns your Aqara smart home into an AI-managed environment - controlling devices, running scenes, and optimizing automations through voice or text.
- [Baidu Map](https://neotask.ai/integrations/baidu-map): Neotask taps into Baidu Maps so your AI agent handles geocoding, routing, traffic monitoring, and location intelligence for operations in China and across Asia.
- [Candid](https://neotask.ai/integrations/candid): Neotask brings Candid's nonprofit and philanthropic intelligence into your grant research and fundraising workflows - search funders, grantees, and grants through natural conversation.
- [Google Maps](https://neotask.ai/integrations/google-maps): Neotask puts Google Maps Platform intelligence into your operations - geocode addresses, optimize routes, and analyze locations without writing API calls.
- [Google Places](https://neotask.ai/integrations/google-places): Neotask queries the Google Places API to surface business listings, ratings, and location data your operations depend on.
- [Google Directions](https://neotask.ai/integrations/google-routes): Neotask taps the Google Routes API to optimize delivery routes, calculate travel times, and plan logistics with real-time traffic data.
- [Kiwi.com](https://neotask.ai/integrations/kiwicom): Neotask connects to Kiwi.com so your agents can search flights across hundreds of airlines and find the best itineraries for business travel and team coordination.
- [lastminute.com](https://neotask.ai/integrations/lastminutecom): Neotask integrates lastminute.com so agents can search and book hotels, flights, and travel packages for business trips and team events through natural conversation.
- [Mapbox](https://neotask.ai/integrations/mapbox): Neotask calculates distances, generates directions, and analyzes geospatial data through Mapbox on Neotask - mapping intelligence for logistics, planning, and location features.
- [Philips Hue](https://neotask.ai/integrations/philips-hue): Control Philips Hue lights and scenes through conversation - Neotask uses AI agents to give you smart home lighting control through natural language.
- [Ticket Tailor](https://neotask.ai/integrations/tickettailor): Neotask automates your Ticket Tailor event operations - manage events, track attendance, process orders, and handle memberships without the dashboard.
- [Yardi Virtuoso](https://neotask.ai/integrations/yardivirtuoso): Neotask connects to Yardi Virtuoso so property management teams can automate leasing operations, tenant management, and financial workflows through natural conversation.

## App Integrations: Productivity

- [2Slides](https://neotask.ai/integrations/2slides): Neotask connects to 2Slides so your agents can generate professional presentations from prompts, outlines, or data without anyone opening PowerPoint.
- [Airtable](https://neotask.ai/integrations/airtable): Neotask keeps your Airtable bases working for your business around the clock - creating records, syncing data, and maintaining your operational backbone automatically.
- [Anytype](https://neotask.ai/integrations/anytype): Neotask turns Anytype into a self-organizing knowledge system - creating spaces, managing objects, and structuring your team's information without manual upkeep.
- [Archbee](https://neotask.ai/integrations/archbee): Neotask makes your Archbee documentation instantly accessible - letting agents search, retrieve, and surface the right docs at exactly the right moment.
- [Asana](https://neotask.ai/integrations/asana): Neotask runs your Asana projects autonomously - create tasks, update statuses, track goals, and get project reports through conversation instead of dashboards.
- [Atlassian](https://neotask.ai/integrations/atlassian): Neotask runs your Atlassian suite on autopilot - managing Jira issues, Confluence pages, and Compass components so your team stays coordinated without the busywork.
- [Basecamp](https://neotask.ai/integrations/basecamp): Manage Basecamp projects and to-dos through Neotask
- [Baserow](https://neotask.ai/integrations/baserow): Neotask turns Baserow into your autonomous no-code database - create tables, manage records, query data, and build collaborative workflows through natural conversation.
- [BoldSign](https://neotask.ai/integrations/boldsign): Neotask automates your e-signature workflow through BoldSign - sending documents, tracking signatures, and managing contacts so contracts close faster without manual follow-up.
- [Cal.com](https://neotask.ai/integrations/cal): Neotask automates your Cal.com scheduling - managing bookings, configuring event types, and handling rescheduling so your calendar runs itself.
- [Cal.com](https://neotask.ai/integrations/cal-com): Manage your Cal.com scheduling through conversation - Neotask uses AI agents to automate bookings, availability, and event types without touching the dashboard.
- [Cartesia](https://neotask.ai/integrations/cartesia): Neotask integrates Cartesia's ultra-low-latency voice AI into your automation stack - generate speech, clone voices, and manage voice infrastructure through natural conversation.
- [Circleback](https://neotask.ai/integrations/circleback): Neotask turns every Circleback meeting into actionable intelligence - searching transcripts, surfacing follow-ups, and connecting meeting insights to the rest of your business automatically.
- [ClickUp](https://neotask.ai/integrations/clickup): Neotask manages your entire ClickUp workspace - 47 actions covering tasks, time tracking, documents, and team chat, all through conversation.
- [Clockwise](https://neotask.ai/integrations/clockwise): Neotask coordinates your Clockwise scheduling - confirming meeting proposals and creating booking links so calendar management happens without back-and-forth.
- [Coda](https://neotask.ai/integrations/coda): Automate Coda docs and tables through Neotask
- [Confluence](https://neotask.ai/integrations/confluence): Search and manage Confluence pages through Neotask
- [Dart](https://neotask.ai/integrations/dart): Neotask manages your Dart project workspace - create tasks, write docs, and organize work through conversation so your team stays focused on execution.
- [Egnyte](https://neotask.ai/integrations/egnyte): Neotask transforms Egnyte into an intelligent content platform - searching files, analyzing documents with AI, managing projects, and collaborating through a single autonomous agent.
- [Evernote](https://neotask.ai/integrations/evernote): Search and manage your Evernote notes through Neotask
- [Fellow](https://neotask.ai/integrations/fellow): Neotask unlocks Fellow's meeting intelligence - surfacing action items, transcripts, summaries, and participant data so your meeting decisions drive real business outcomes.
- [Fibery](https://neotask.ai/integrations/fibery): Neotask automates your Fibery workspace - creating and updating entities, running queries, and keeping your interconnected business data in sync without manual effort.
- [Fireflies.ai](https://neotask.ai/integrations/fireflies): Neotask makes your Fireflies meeting data searchable and actionable -- Neotask queries transcripts, surfaces summaries, and tracks contacts so meeting knowledge flows into your business operations.
- [Gamma](https://neotask.ai/integrations/gamma): Neotask generates professional presentations, documents, and web pages through Gamma on demand - transforming raw ideas into polished visual content without design effort.
- [Glean](https://neotask.ai/integrations/glean): Search across all your company knowledge through Neotask - Glean enterprise search and AI assistant through conversation.
- [Google Calendar](https://neotask.ai/integrations/google-calendar): Neotask turns Google Calendar into a hands-free scheduling engine -- Neotask creates events, checks availability, and manages your calendar so scheduling never interrupts your real work.
- [Google Docs](https://neotask.ai/integrations/google-docs): Neotask automates document creation and management through Google Docs -- Neotask writes, formats, and shares documents so your team produces polished content without manual editing.
- [Google Forms](https://neotask.ai/integrations/google-forms): Neotask builds surveys, quizzes, and intake forms on Google Forms - your operations team collects data without designing a single question manually.
- [Google Keep](https://neotask.ai/integrations/google-keep): Neotask organizes your team's notes, checklists, and reminders on Google Keep - capture ideas and action items without opening the app.
- [Google Sheets](https://neotask.ai/integrations/google-sheets): Neotask reads, writes, and analyzes Google Sheets - your spreadsheets become live operational dashboards that update themselves.
- [Google Slides](https://neotask.ai/integrations/google-slides): Neotask builds and edits Google Slides presentations - decks go from brief to boardroom without opening the editor.
- [Google Tasks](https://neotask.ai/integrations/google-tasks): Neotask manages your Google Tasks - create, prioritize, and complete to-do items without switching apps or losing context.
- [Grain](https://neotask.ai/integrations/grain): Neotask extracts value from every Grain recording - retrieving transcripts, organizing views, and automating webhooks so meeting intelligence flows directly into your business processes.
- [Granola](https://neotask.ai/integrations/granola): Neotask unlocks your Granola meeting intelligence - querying transcripts, surfacing action items, and searching your meeting history to keep your business running on what was actually said.
- [Guru](https://neotask.ai/integrations/guru): Neotask taps into your Guru knowledge base - agents search, retrieve, and update company knowledge so your team always has the right answer.
- [Jira](https://neotask.ai/integrations/jira): Manage Jira issues and sprints through Neotask
- [JotForm](https://neotask.ai/integrations/jotform): Neotask creates, manages, and processes JotForm forms - build forms, capture submissions, and route data automatically without touching JotForm's interface.
- [Kagi Search](https://neotask.ai/integrations/kagisearch): Neotask uses Kagi's high-quality, ad-free search so your agents get clean, relevant web results without the noise of ad-driven search engines.
- [Linear](https://neotask.ai/integrations/linear): Neotask keeps your Linear workspace running on autopilot - agents create issues, manage cycles, track initiatives, and surface blockers before your standup.
- [Lumin](https://neotask.ai/integrations/lumin): Neotask integrates Lumin's PDF collaboration and document management platform so agents can handle documents, annotations, and signing workflows without opening a browser.
- [Make](https://neotask.ai/integrations/make): Neotask builds, triggers, and manages Make automation scenarios - workflow orchestration through conversation instead of the visual builder.
- [Microsoft 365](https://neotask.ai/integrations/microsoft-365): Neotask connects every corner of Microsoft 365 - email, calendar, SharePoint, OneDrive, Word, and the Microsoft Graph - into a single agent-driven workspace that runs itself.
- [Microsoft To Do](https://neotask.ai/integrations/microsoft-todo): Neotask turns Microsoft To Do into a hands-free task engine - add, organize, and complete tasks through natural conversation.
- [Microsoft 365](https://neotask.ai/integrations/microsoft365): Neotask turns Microsoft 365 into your autonomous productivity platform - manage email, calendar, files, users, and the complete M365 suite through conversation at enterprise scale.
- [Miro](https://neotask.ai/integrations/miro): Neotask brings Miro into your autonomous workflow - create diagrams, manage documents, sync tables, and explore boards without opening the canvas.
- [Monday.com](https://neotask.ai/integrations/monday): Neotask turns Monday.com into a self-managing project hub - create boards, update items, and track progress agents that keep your work organized.
- [Notion](https://neotask.ai/integrations/notion): Neotask turns your Notion workspace into a self-updating knowledge hub - AI agents create pages, manage databases, and surface information without you opening a tab.
- [Obsidian](https://neotask.ai/integrations/obsidian): Navigate and update your Obsidian vault through Neotask
- [Pipefy](https://neotask.ai/integrations/pipefy): Manage Pipefy processes, cards, and workflows through conversation - Neotask uses AI agents to operate your process automation without the Pipefy dashboard.
- [Plane](https://neotask.ai/integrations/plane): Neotask runs your Plane project management end-to-end - creating work items, managing cycles, organizing modules, and keeping your team's roadmap moving without manual overhead.
- [Play.ai](https://neotask.ai/integrations/playai): Neotask integrates Play.ai's conversational voice platform into your automation stack - create agents, generate speech, and manage voice experiences through natural conversation.
- [Roam Research](https://neotask.ai/integrations/roam-research): Neotask brings intelligence to your Roam graph - capture, link, and surface knowledge without breaking your flow.
- [Smartsheet](https://neotask.ai/integrations/smartsheet): Neotask automates your Smartsheet operations - managing sheets, updating rows, searching across workspaces, and keeping your collaborative work data in sync without manual intervention.
- [Todoist](https://neotask.ai/integrations/todoist): Manage Todoist tasks and projects through Neotask
- [Trello](https://neotask.ai/integrations/trello): Control your Trello boards and cards through Neotask
- [Wrike](https://neotask.ai/integrations/wrike): Neotask connects to Wrike so your project data becomes a conversation - assign tasks, track status, and run reports without leaving your workflow.
- [Zapier](https://neotask.ai/integrations/zapier): Trigger Zapier zaps and manage automation workflows through conversation - Neotask uses AI agents to operate your Zapier integrations without the dashboard.

## App Integrations: Research

- [Exa](https://neotask.ai/integrations/exa): Neotask uses Exa to research the web at depth - company intelligence, people profiles, LinkedIn data, and deep research reports through conversation.
- [Perplexity](https://neotask.ai/integrations/perplexity): Neotask adds Perplexity's real-time web intelligence to every agent workflow - searching, asking, researching, and reasoning across the live web without leaving your automation stack.
- [Scholar Gateway](https://neotask.ai/integrations/scholargateway): Neotask connects your research workflows to Wiley's Scholar Gateway - running semantic search across peer-reviewed scientific literature to ground your work in authoritative sources.

## App Integrations: Sales

- [Apollo.io](https://neotask.ai/integrations/apollo-io): Automate sales prospecting and outreach through Neotask
- [Outreach](https://neotask.ai/integrations/outreach): Neotask turns Outreach into your autonomous sales engagement engine - manage prospects, accounts, opportunities, and meeting intelligence.
- [Salesforce](https://neotask.ai/integrations/salesforce): Neotask connects to Salesforce so your CRM runs itself - query records, deploy metadata, manage orgs, and automate your entire sales operations.

## App Integrations: Sales & CRM

- [Apollo](https://neotask.ai/integrations/apollo): Prospect smarter with Neotask - AI agents that unlock Apollo's database through pure conversation.
- [Apollo.io](https://neotask.ai/integrations/apolloio): Neotask turns Apollo.io into an autonomous sales engine - searching for prospects, enriching companies, managing email sequences, and delivering analytics while your reps focus on closing.
- [Attio](https://neotask.ai/integrations/attio): Neotask turns Attio into a self-managing CRM - your agent searches records, manages deals, logs notes, and tracks tasks so your sales team sells instead of doing data entry.
- [Clay](https://neotask.ai/integrations/clay): Neotask automates your Clay outreach - enriching contacts, managing prospect lists, and orchestrating sales workflows so your pipeline fills itself.
- [Crossbeam](https://neotask.ai/integrations/crossbeam): Neotask unlocks your partner ecosystem data through Crossbeam -- Neotask surfaces account overlaps, partner signals, and co-sell opportunities so your partnerships actually drive revenue.
- [Explorium](https://neotask.ai/integrations/explorium): Neotask uses Explorium to enrich your business and prospect intelligence at unprecedented depth - combining firmographics, technographics, funding data, competitive intelligence, and workforce trends in a single automated workflow.
- [Freshsales](https://neotask.ai/integrations/freshsales): Manage your Freshsales pipeline with Neotask.
- [Gong](https://neotask.ai/integrations/gong): Extract revenue intelligence from Gong calls - Neotask surfaces insights while you focus on selling.
- [Harmonic](https://neotask.ai/integrations/harmonic): Neotask turns Harmonic into an autonomous startup intelligence operation - searching companies, tracking funding rounds, managing watchlists, and surfacing investor insights without manual research.
- [HubSpot](https://neotask.ai/integrations/hubspot): Neotask turns HubSpot into a conversational CRM - manage contacts, deals, workflows, and properties without navigating the dashboard.
- [Hunter](https://neotask.ai/integrations/hunter): Neotask automates your email prospecting with Hunter - finding verified contact emails, enriching company profiles, and building lead lists without leaving your workflow.
- [Pipedrive](https://neotask.ai/integrations/pipedrive): Neotask makes Pipedrive your most responsive deal manager - Neotask handles pipeline queries, updates, and reporting in plain English.
- [Vibe Prospecting](https://neotask.ai/integrations/vibeprospecting): Neotask makes Vibe Prospecting your autonomous B2B intelligence layer - matching businesses and prospects, enriching records, monitoring signals, and exporting datasets at scale.
- [Zoho CRM](https://neotask.ai/integrations/zoho-crm): Run your Zoho CRM from conversation - Neotask turns your words into CRM actions.
- [ZoomInfo](https://neotask.ai/integrations/zoominfo): Neotask uses ZoomInfo to build your ideal customer pipeline automatically - searching, enriching, and researching companies and contacts so your sales team always has the intelligence they need.

## App Integrations: Search

- [Algolia](https://neotask.ai/integrations/algolia): Neotask manages your Algolia search infrastructure - index data, tune relevance, run A/B tests, and monitor performance across 154 available actions.

## App Integrations: Security

- [1Password](https://neotask.ai/integrations/1password): Retrieve secrets, manage vaults, and audit access - Neotask secures your 1Password operations.
- [Asgardeo](https://neotask.ai/integrations/asgardeo): Neotask automates your Asgardeo identity platform - provisioning applications, managing API resources, and configuring auth flows so your team ships faster.
- [Auth0](https://neotask.ai/integrations/auth0): Neotask automates your Auth0 tenant - managing applications, deploying actions, publishing forms, and monitoring logs so your auth infrastructure runs itself.
- [BoostSecurity](https://neotask.ai/integrations/boostsecurity): Neotask makes application security continuous via BoostSecurity - validating packages, surfacing violations, and keeping your codebase free of exploitable dependencies.
- [Contrast Security](https://neotask.ai/integrations/contrast-security): Neotask automates application security triage with Contrast Security -- Neotask queries vulnerabilities, monitors attacks, and tracks library risks so your security team works at machine speed.
- [CrowdStrike](https://neotask.ai/integrations/crowdstrike): Investigate threats, query detections, and respond faster - Neotask brings AI agents to your CrowdStrike Falcon data.
- [CrowdStrike](https://neotask.ai/integrations/crowdstrike-falcon): Neotask brings CrowdStrike Falcon's threat intelligence to your fingertips -- Neotask investigates detections, queries hosts, and surfaces threat actor data so your security team responds in minutes, not hours.
- [Dependabot](https://neotask.ai/integrations/dependabot): Neotask keeps your dependencies secure - and Dependabot alerts.
- [Endor Labs](https://neotask.ai/integrations/endor-labs): Manage open-source dependency risk and software supply chain security through Neotask - Endor Labs insights through conversation.
- [Endor Labs](https://neotask.ai/integrations/endorlabs): Neotask makes supply chain security continuous - scanning dependencies, surfacing real risks, and keeping your software free of exploitable vulnerabilities automatically.
- [Snyk](https://neotask.ai/integrations/snyk): Scan dependencies, fix vulnerabilities, and monitor your code security - Neotask runs your Snyk workflows.
- [SonarQube](https://neotask.ai/integrations/sonarqube): Analyze code quality, track bugs, and enforce security standards - Neotask manages your SonarQube projects.
- [Stytch](https://neotask.ai/integrations/stytch): Neotask manages your Stytch authentication infrastructure - configure projects, manage tokens, and control SDK settings without the console.
- [Vault](https://neotask.ai/integrations/vault): Manage secrets, dynamic credentials, and encryption - Neotask handles your HashiCorp Vault operations.

## App Integrations: Storage

- [Backblaze B2](https://neotask.ai/integrations/backblaze-b2): Manage Backblaze B2 buckets and object storage through conversation - Neotask runs on Neotask to give you S3-compatible cloud storage control without CLI complexity.
- [Box](https://neotask.ai/integrations/box): Neotask automates your Box enterprise content - searching files, running AI analysis, and managing folders so your team finds what they need without digging through directories.
- [Dropbox](https://neotask.ai/integrations/dropbox): Search, share, and organize your Dropbox files through conversation - Neotask runs on Neotask to keep your cloud storage tidy without constant folder navigation.
- [Google Drive](https://neotask.ai/integrations/google-drive): Neotask automates your Google Drive operations -- Neotask searches, organizes, and shares files so your team finds what they need and keeps storage clean without manual housekeeping.
- [OneDrive](https://neotask.ai/integrations/onedrive): Search and manage your OneDrive files through conversation - Neotask runs on Neotask to organize Microsoft 365 content without navigating SharePoint or the OneDrive web app.
- [S3](https://neotask.ai/integrations/s3): Manage S3 buckets, objects, and access policies through conversation - Neotask uses AI agents to handle S3 operations without wrestling with the AWS console or CLI syntax.
- [Wasabi](https://neotask.ai/integrations/wasabi): Manage Wasabi hot cloud storage through conversation - Neotask uses AI agents to handle your S3-compatible object storage without egress fee surprises or console complexity.

## App Integrations: Web Services

- [GoDaddy](https://neotask.ai/integrations/godaddy): Neotask makes domain research instant - checking availability and generating name ideas through GoDaddy without leaving your business workflow.

## App Integrations: Web3

- [Blockscout](https://neotask.ai/integrations/blockscout): Neotask automates your blockchain analysis through Blockscout - inspecting contracts, tracking addresses, and monitoring chains so your Web3 operations run with full visibility.

## More App Integrations

- [1800Flowers](https://neotask.ai/integrations/1800flowers): 1800Flowers connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Browse flower, plant, and gift collections by occasion, view product details and pricing, and place an order for delivery directly through...
- [1Password Developer Docs MCP server](https://neotask.ai/integrations/1password-developer-docs-mcp-server): 1Password Developer Docs MCP server connects to Neotask so an agent can act on your behalf instead of you doing it by hand.
- [3Min API](https://neotask.ai/integrations/3min-api): 3Min API connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Define a JSON schema in chat and get a full CRUD REST API generated automatically, complete with sandbox and production endpoints, webhook...
- [ABEMA](https://neotask.ai/integrations/abema): ABEMA connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Browse and stream ABEMA's live TV channels, original dramas, anime, news, and sports programming from Japan without needing a separate...
- [Access Approval MCP server](https://neotask.ai/integrations/access-approval-mcp-server): Access Approval MCP server connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Review, approve, dismiss, and audit Google Cloud Access Approval requests, giving organizations control over when Google...
- [AccessOwl](https://neotask.ai/integrations/accessowl): AccessOwl connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Browse the SaaS applications your organization has enabled and submit access requests for the ones you need, streamlining internal app access...
- [AccurateScribe.ai – Transcribe](https://neotask.ai/integrations/accuratescribe-ai-transcribe): AccurateScribe.ai – Transcribe connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Upload or link audio and video from meetings, podcasts, lectures, or interviews and get accurate transcripts, meeting...
- [ACKO](https://neotask.ai/integrations/acko): ACKO connects to Neotask so an agent can act on your behalf instead of you doing it by hand. Enter a vehicle registration number to get instant car or bike insurance quotes with coverage options, and check whether the vehicle has any...
- [Activepieces MCP server](https://neotask.ai/integrations/activepieces-mcp-server): Activepieces is a workflow automation platform where people connect apps and services into automated flows, and this MCP server is the connector that puts a Neotask agent directly in touch with an existing Activepieces account rather than...
- [ADAC Mietwagen](https://neotask.ai/integrations/adac-mietwagen): ADAC Mietwagen compares rental car prices around the world across vehicle classes from economy up through premium, and it applies ADAC member benefits to that pricing automatically, so a Neotask agent connected to it can surface a rate...
- [Adalo](https://neotask.ai/integrations/adalo): Adalo turns plain descriptions of an app into working mobile and web screens, so instead of dragging elements around a visual builder, you describe a screen, a data structure, or a piece of logic and Adalo builds or changes it for you.
- [AdisInsight](https://neotask.ai/integrations/adisinsight): AdisInsight is Springer's drug development database, and it lets you search pipelines and clinical trials by developer, therapeutic area, phase, mechanism of action, sponsor, or indication.
- [Advanced Web Ranking](https://neotask.ai/integrations/advanced-web-ranking): Advanced Web Ranking tracks keyword rankings and search results pages for the sites you manage, and Neotask can pull that data straight into a conversation instead of you logging into a separate dashboard every time a question comes up.
- [Adyen MCP server](https://neotask.ai/integrations/adyen-mcp-server): Adyen MCP server connects a Neotask agent to Adyen, so requests you'd otherwise handle through Adyen's own tools can instead be made in conversation without a separate login step getting in the way of whatever else you're doing.
- [Agent card](https://neotask.ai/integrations/agent-card): The Houstonian Hotel, Club & Spa publishes an agent card, a machine-readable profile describing its reservations, club, and spa offerings for AI agents to read directly rather than requiring a person to browse the site page by page looking...
- [Agiflow](https://neotask.ai/integrations/agiflow): Agiflow is a project management tool built around projects, tasks, epics, and features moving through a kanban workflow, and Neotask lets you drive it entirely from chat without opening a browser tab for the board itself.
- [Agility CMS MCP Server](https://neotask.ai/integrations/agility-cms-mcp-server): Agility CMS is a headless content management system, and its MCP server lets Neotask fetch, manage, and query content, pages, and models without opening the CMS admin panel yourself for every small task.
- [Agora](https://neotask.ai/integrations/agora): Agora takes a claim or prediction and returns a verdict of true, false, or uncertain, backed by a confidence score rather than a flat answer with nothing behind it.
- [Aha!](https://neotask.ai/integrations/aha): Aha! integration connects a Neotask agent directly to your Aha! account, so requests that would otherwise mean opening Aha!'s own interface can instead be handled through conversation without switching context away from what you were...
- [AI Stats Docs MCP](https://neotask.ai/integrations/ai-stats-docs-mcp): AI Stats Docs MCP connects a Neotask agent directly to a documentation server built around AI statistics and metrics, giving the agent a place to check definitions, methodology notes, and source context instead of relying on a remembered...
- [AI Voice Generator](https://neotask.ai/integrations/ai-voice-generator): AI Voice Generator turns written text into a realistic spoken audio track, taking a script or transcript and returning a produced voiceover as a downloadable recording.
- [Airbyte Agent MCP](https://neotask.ai/integrations/airbyte-agent-mcp): Airbyte Agent MCP gives a Neotask agent a path into the business systems and data sources already wired up through Airbyte's data movement platform.
- [Airport Pickups London](https://neotask.ai/integrations/airport-pickups-london): Airport Pickups London books a private car transfer to or from any London airport or cruise port, quoting a fixed price instantly rather than leaving the cost to a running meter or a variable rate decided later.
- [AIS Insurance](https://neotask.ai/integrations/ais-insurance): AIS Insurance takes basic vehicle and driving history details and returns car insurance quotes compared across multiple top-rated carriers rather than a single provider's rate presented as the only option.
- [Aiwyn Tax (formerly Column Tax)](https://neotask.ai/integrations/aiwyn-tax-formerly-column-tax): Aiwyn Tax, formerly known as Column Tax, takes uploaded tax documents such as W-2s and calculates federal and state tax estimates before producing a completed 1040 PDF ready for review.
- [Akeneo MCP Server](https://neotask.ai/integrations/akeneo-mcp-server): Akeneo MCP Server connects a Neotask agent to Akeneo's product information management platform, covering the work of managing, enriching, and distributing product data across sales channels from one place.
- [ALL Accor](https://neotask.ai/integrations/all-accor): ALL Accor searches and compares more than 5,700 Accor hotels worldwide, spanning brands from ibis and Novotel through Sofitel, Fairmont, and Raffles, laid out on an interactive map with price, photos, amenities, and reviews attached to...
- [AllTrails](https://neotask.ai/integrations/alltrails): AllTrails helps find hiking, biking, and walking trails either near a specific location or within a chosen map area, with detailed information on distance, difficulty, reviews, and photos attached to each one that comes back.
- [Almanac by PassBy](https://neotask.ai/integrations/almanac-by-passby): Almanac by PassBy answers questions about physical retail store performance, drawing on aggregated, privacy-safe visit data rather than requiring someone to build a report from raw numbers first.
- [Almosafer.com](https://neotask.ai/integrations/almosafer-com): Almosafer.com is a Middle East travel booking platform that lets travelers search hotels by destination, dates, and personal preferences like star rating and cancellation policy.
- [Alpic](https://neotask.ai/integrations/alpic): Alpic is a hosting platform built specifically for MCP servers, and a Neotask agent connected to it can check on a deployed project without anyone opening a dashboard.
- [Alpy.com](https://neotask.ai/integrations/alpy-com): Alpy.com is a booking site for ski and snowboard rental shops across the Alps, and a Neotask agent connected to it can handle the comparison shopping that usually comes before a rental booking.
- [Alton Towers Tickets](https://neotask.ai/integrations/alton-towers-tickets): Alton Towers Tickets covers the official ticketing flow for the UK theme park, and a Neotask agent linked to it can walk someone through picking a visit date and ticket type before sending them to checkout.
- [Amapas Vacation Rental](https://neotask.ai/integrations/amapas-vacation-rental): Amapas Vacation Rental is the direct booking channel for a single property, the Amapas 353 beachfront condo in Puerto Vallarta's Zona Romantica, and a Neotask agent connected to it works more like a concierge than a search engine since...
- [Amentum MCP server](https://neotask.ai/integrations/amentum-mcp-server): The Amentum MCP server gives access to model-backed environmental and geospatial data spanning atmosphere, radiation, geomagnetic activity, gravity, and ocean conditions, and a Neotask agent connected to it can pull any of these directly...
- [아모레몰 AMORE MALL](https://neotask.ai/integrations/amore-mall): Amore Mall is the shopping platform for Amorepacific's cosmetics and skincare brands, including Sulwhasoo, Hera, Innisfree, and Espoir, and a Neotask agent connected to it can shop across all of them in one pass rather than brand by brand.
- [Amplifyr](https://neotask.ai/integrations/amplifyr): Amplifyr tracks how a brand shows up across AI models, and a Neotask agent connected to it turns that tracking into something a marketing or comms team can just ask about in plain English.
- [Android Management API MCP server](https://neotask.ai/integrations/android-management-api-mcp-server): The Android Management API MCP server gives access to Google's enterprise mobility management tools for Android, and a Neotask agent connected to it can handle the administrative side of managing a fleet of company devices.
- [Angi](https://neotask.ai/integrations/angi): Angi is a national directory of vetted home service professionals, and a Neotask agent connected to it can turn a vague home project into a shortlist of qualified pros without you opening a browser tab.
- [Angler Adventures: Lure Guide](https://neotask.ai/integrations/angler-adventures-lure-guide): Angler Adventures: Lure Guide turns a description of the water in front of you into a ranked set of bass lure picks, and a Neotask agent handles the translation.
- [AnyPDF - Your PDF Converter](https://neotask.ai/integrations/anypdf-your-pdf-converter): AnyPDF handles the routine PDF work that piles up on anyone dealing with documents: compressing large files, merging several into one, splitting one apart, pulling specific pages out, rotating pages that scanned sideways, and converting...
- [Apartment List](https://neotask.ai/integrations/apartment-list): Apartment List matches renters to units based on what actually matters to them, and a Neotask agent connected to it can run that matching from a plain conversation instead of a filtered search page.
- [API2Cart MCP OAuth server](https://neotask.ai/integrations/api2cart-mcp-oauth-server): API2Cart gives a Neotask agent one connection point into store, product, order, customer, category, and webhook data across many ecommerce platforms and marketplaces, so you're not building a separate integration for every cart software a...
- [Apideck MCP Server](https://neotask.ai/integrations/apideck-mcp-server): Apideck's unified APIs cover accounting, CRM, HRIS, applicant tracking, file storage, ecommerce, and issue tracking software, and a Neotask agent connected here can query any of them, plus manage the integrations and webhooks tied to your...
- [Apigee MCP server](https://neotask.ai/integrations/apigee-mcp-server): Apigee is Google Cloud's API management platform, and a Neotask agent connected to it can operate your organizations, environments, API proxies, and products directly instead of you working through the console for routine configuration...
- [Apixel](https://neotask.ai/integrations/apixel): Apixel generates images straight from a written description, and it does something a lot of image generators still stumble on: it renders the text you ask for accurately inside the image itself, not as garbled letters that need fixing...
- [AppDeploy](https://neotask.ai/integrations/appdeploy): AppDeploy takes a description of a website or web app and builds it, publishing a live public link you can share right away without any separate hosting setup.
- [Apple Music](https://neotask.ai/integrations/apple-music): Apple Music handles this without you switching tabs: Search the Apple Music catalog for songs, artists, albums, and playlists, preview short clips, and get music recommendations without needing a subscription.
- [ApplyBoard](https://neotask.ai/integrations/applyboard): ApplyBoard is a platform students and their families use to find and apply to international study programs, and Neotask's agent uses it to turn a rough idea of studying abroad into an actual shortlist.
- [Arcjet MCP Server](https://neotask.ai/integrations/arcjet-mcp-server): Arcjet is a security layer developers drop into an application to handle bot detection, rate limiting, prompt injection detection, and sensitive data protection, and Neotask's agent connects to it so a developer can check or adjust those...
- [Aris - Digital Consultant](https://neotask.ai/integrations/aris-digital-consultant): Aris is a digital consulting service built to answer business and technology questions about setting up a company website, internal systems, or an AI agent, and Neotask's agent uses it as a second opinion before those decisions get locked...
- [Article Galaxy MCP](https://neotask.ai/integrations/article-galaxy-mcp): Article Galaxy MCP connects Neotask's agent to Article Galaxy and Scite so researchers can search for scholarly articles and citation data without leaving the conversation.
- [Artue MCP Server](https://neotask.ai/integrations/artue-mcp-server): Artue is a marketplace for original contemporary art, including resale pieces from collectors, and Neotask's agent uses it to help someone browse and find artwork that fits a specific style or budget.
- [Assessment Generator](https://neotask.ai/integrations/assessment-generator): Assessment Generator turns a block of text, an uploaded document, or just a topic into a structured set of quiz questions, and Neotask's agent uses it so an instructional designer can go from source material to a working assessment in one...
- [AT&T](https://neotask.ai/integrations/at-and-t): AT&T's connection lets Neotask's agent check internet availability at a specific address and compare the fiber and internet air plans, speeds, and pricing available there.
- [atHome](https://neotask.ai/integrations/athome): atHome is Luxembourg's real estate listing platform, and Neotask's agent uses it to help someone browse apartments, houses, or investment properties in Luxembourg City and the Greater Region without opening the site directly.
- [Atlys](https://neotask.ai/integrations/atlys): Atlys answers visa questions by checking official government and embassy sources, and Neotask's agent uses it so a traveler can get a straight answer instead of piecing one together across multiple government portals.
- [Audible](https://neotask.ai/integrations/audible): Audible is Amazon's audiobook service, and Neotask's agent uses its catalog to recommend something to listen to based on genre, mood, or topic instead of making you scroll the app yourself.
- [Aura](https://neotask.ai/integrations/aura): Aura is a company-intelligence database that recruiters, investors, and market analysts use to pull structured data on employee headcount, hiring velocity, attrition, and organizational structure across industries.
- [AuthZed MCP Server](https://neotask.ai/integrations/authzed-mcp-server): AuthZed MCP Server gives an agent direct access to the authorization data stored in SpiceDB or AuthZed Cloud, the permission systems built on Google's Zanzibar model that many applications use to answer who can access what at scale.
- [Autodesk Product Help](https://neotask.ai/integrations/autodesk-product-help): Autodesk Product Help connects to Autodesk's own documentation across more than 110 products, including AutoCAD, Revit, and Fusion, so answers come from the official source instead of a forum post or an outdated tutorial.
- [AutoMotion](https://neotask.ai/integrations/automotion): AutoMotion connects Neotask to a network of more than 25,000 auto service locations covering oil changes, tires, brakes, and auto glass, so describing a car problem is enough to find a nearby shop and get it scheduled.
- [Autotrader](https://neotask.ai/integrations/autotrader): Autotrader gives Neotask access to a large marketplace of new and used vehicle listings along with tools that go beyond browsing: an instant cash offer for a car you own, an affordability calculator, and expert reviews and news.
- [Autoverzekering.nl](https://neotask.ai/integrations/autoverzekering-nl): Autoverzekering.nl is a Dutch car insurance comparison service, and connected to Neotask it lets someone check whether switching insurers would actually save money without visiting multiple insurer sites.
- [Autovit.ro](https://neotask.ai/integrations/autovit-ro): Autovit.ro is Romania's used and new car marketplace, the place most Romanian buyers already check before they visit a dealer.
- [Avalara MCP server family](https://neotask.ai/integrations/avalara-mcp-server-family): Avalara builds the tax infrastructure a lot of finance teams already run transaction tax through without thinking about it: sales tax, VAT, and the paperwork that follows a payment.
- [Avaza MCP Server](https://neotask.ai/integrations/avaza-mcp-server): Avaza bundles project management, time tracking, expenses, and client invoicing into one workspace, which is exactly the kind of spread that normally means jumping between several tabs to answer one question.
- [Avi A2A server](https://neotask.ai/integrations/avi-a2a-server): Avi is Avalara's agent-to-agent interface into its tax engine, built for the case where a tax lookup needs to happen inside an automated flow rather than through a person clicking through a portal.
- [Avito.ma](https://neotask.ai/integrations/avito-ma): Avito.ma is Morocco's general classifieds marketplace, covering everything from cars and apartments to phones, fashion, and home goods.
- [Avo MCP](https://neotask.ai/integrations/avo-mcp): Avo is where a lot of product and data teams keep their analytics tracking plan, the definitions of which events, properties, and metrics an app is supposed to send, and how those definitions branch across different releases.
- [AWeber](https://neotask.ai/integrations/aweber): AWeber is a long-standing email marketing platform built around subscriber lists and newsletter campaigns. Connected to Neotask, an agent can manage those subscribers and lists directly, draft or send a campaign, and pull up performance...
- [AWS Knowledge MCP Server](https://neotask.ai/integrations/aws-knowledge-mcp-server): The AWS Knowledge MCP Server gives an agent direct access to AWS Labs' own curated documentation, specifically the setup and configuration guidance for AWS-oriented developer tools and other agent servers in that ecosystem.
- [B12 Website Generator](https://neotask.ai/integrations/b12-website-generator): B12 turns a plain-language description of a business into a finished, editable website, matched to the goals and brand the owner already has in mind.
- [Backstage](https://neotask.ai/integrations/backstage): Backstage lists acting jobs across film, TV, theater, and commercials, and a Neotask agent can search, filter, and apply to those listings on an actor's behalf.
- [Ballpark](https://neotask.ai/integrations/ballpark): Ballpark helps product teams run customer research: building screeners, recruiting participants, and summarizing what respondents say.
- [Bandsintown](https://neotask.ai/integrations/bandsintown): Bandsintown tracks tour dates and live shows, and a Neotask agent can search that catalog by artist, city, venue, or date to answer a simple question: what's playing, and when.
- [Bankrate](https://neotask.ai/integrations/bankrate): Bankrate tracks current consumer banking rates, savings accounts, CDs, and similar products, and a Neotask agent can pull those figures directly instead of relying on outdated or remembered numbers.
- [Barceló Hotel Group](https://neotask.ai/integrations/barcelo-hotel-group): Barcelo Hotel Group covers over 160 properties across four brands, Barcelo, Occidental, Royal Hideaway, and Allegro, spanning Spain, the Caribbean, Mexico, and beyond.
- [BASBAS customer-account MCP server](https://neotask.ai/integrations/basbas-customer-account-mcp-server): BASBAS runs an ecommerce storefront for its herbal liqueur, and this connection lets a Neotask agent browse that store's product listings and manage a signed-in customer's account without the customer opening the site themselves.
- [Base44](https://neotask.ai/integrations/base44): Base44 takes a plain description of an app idea and generates a full stack project around it: data models, backend logic, and a working UI, in one pass.
- [BE-A](https://neotask.ai/integrations/be-a): BE-A connects a search directly into a hotel's own availability and pricing system, bypassing the markup that comes with third-party booking sites.
- [Bed-and-Breakfast.it](https://neotask.ai/integrations/bed-and-breakfast-it): Bed-and-Breakfast.it is a booking platform focused on Italy: thousands of bed and breakfasts, agriturismi, vacation rentals, and small guesthouses listed directly by their hosts.
- [Benevity](https://neotask.ai/integrations/benevity): Benevity maintains a verified database of more than 2.3 million nonprofits worldwide, built for corporate giving programs but useful to anyone trying to figure out where a donation or volunteer hour actually does the most good.
- [Best Lawyers](https://neotask.ai/integrations/best-lawyers): Best Lawyers ranks attorneys through peer review rather than paid placement or self-submitted reviews, which is why its rankings carry weight in the legal industry.
- [BestColleges.com](https://neotask.ai/integrations/bestcolleges-com): BestColleges.com catalogs accredited online degree programs from partner universities, aimed at people who need a flexible path to a degree rather than a traditional on-campus one.
- [Better Stack MCP server](https://neotask.ai/integrations/better-stack-mcp-server): Better Stack is a monitoring and observability platform covering uptime checks, on-call incident management, log aggregation, and error tracking, and this MCP server lets Neotask read straight from a connected account.
- [Bible](https://neotask.ai/integrations/bible): This integration connects Neotask to the YouVersion Bible platform, giving an agent access to any passage in a chosen translation, the ability to compare translations side by side, topic-based reading plans, and audio readings in different...
- [Bigdata.com](https://neotask.ai/integrations/bigdata-com): Bigdata.com indexes global news, earnings call transcripts, and regulatory filings with entity-aware search, meaning it understands that a query about a company covers its subsidiaries, executives, and related filings, not just literal...
- [BigGeo AI](https://neotask.ai/integrations/biggeo-ai): BigGeo AI answers location-based questions using geospatial datasets, administrative boundaries, and geocoding, effectively giving Neotask a way to reason about places and how they relate to each other without anyone touching a GIS tool.
- [BigQuery Data Transfer MCP](https://neotask.ai/integrations/bigquery-data-transfer-mcp): This integration connects Neotask to Google's BigQuery Data Transfer service, which handles scheduled imports from supported sources straight into a BigQuery data warehouse.
- [Binance](https://neotask.ai/integrations/binance): This integration gives Neotask read access to Binance market data across spot, futures, and options: current and historical prices, order books, recent trades, and candlestick charts.
- [bioRxiv](https://neotask.ai/integrations/biorxiv): With bioRxiv connected, Neotask's agent takes over the parts you'd normally do by hand. Search bioRxiv and medRxiv preprint servers for biology and medical research papers, filter by funder or category, retrieve full preprint details, and...
- [Bird MCP server (hosted)](https://neotask.ai/integrations/bird-mcp-server-hosted): Send and manage email, SMS, voice, and WhatsApp messages, and review messaging workflow data through a connected Bird account. Useful for teams automating customer communications across multiple channels.
- [Bites](https://neotask.ai/integrations/bites): Instead of opening Bites yourself, ask Neotask. Search local restaurants by cuisine or location, browse full menus with modifier options, check rewards balance, and place delivery or pickup orders.
- [Bitquery MCP Server](https://neotask.ai/integrations/bitquery-mcp-server): Bitquery MCP Server is one of the connected apps Neotask's agent can operate directly. Query indexed on-chain blockchain data and real time crypto market activity across many chains using structured queries.
- [Blackbaud](https://neotask.ai/integrations/blackbaud): Blackbaud handles this without you switching tabs: Look up nonprofit constituent records, event details, and gift or donation history, and generate content, all pulled live from a connected Blackbaud SKY API account.
- [BlazeSQL](https://neotask.ai/integrations/blazesql): With BlazeSQL connected, Neotask's agent takes over the parts you'd normally do by hand. Ask questions about a connected SQL database in plain language and get accurate answers, with built-in checks for reliability and accuracy over time.
- [Blossom](https://neotask.ai/integrations/blossom): Instead of opening Blossom yourself, ask Neotask. Blossom integration for connecting agents to Blossom. You describe the outcome in plain language and the agent carries it out using Blossom's real capabilities, then tells you what changed.
- [BlueConic Solutions Finder](https://neotask.ai/integrations/blueconic-solutions-finder): BlueConic Solutions Finder is one of the connected apps Neotask's agent can operate directly. Take a short quiz about a company's current marketing strategy and get matched marketing plays and growth strategies aligned to stated...
- [BlueSuite](https://neotask.ai/integrations/bluesuite): BlueSuite handles this without you switching tabs: Look up contacts, create and send quotes, log service requests, schedule jobs, and track invoices for a home service business through a connected BlueSuite account.
- [BluTranslate](https://neotask.ai/integrations/blutranslate): With BluTranslate connected, Neotask's agent takes over the parts you'd normally do by hand. Upload a document in conversation, choose a target language and translation options, then start, track, and download a translated version through...
- [BODi](https://neotask.ai/integrations/bodi): Describe fitness goals or preferences to find a matching guided workout program from a library of 140 plus programs and 8,000 plus workouts, including well-known series like P90X and Insanity.
- [Boleron AI](https://neotask.ai/integrations/boleron-ai): Instead of opening Boleron AI yourself, ask Neotask. Boleron AI integration for connecting agents to Boleron AI. You describe the outcome in plain language and the agent carries it out using Boleron AI's real capabilities, then tells you...
- [Book Direct](https://neotask.ai/integrations/book-direct): Book Direct is one of the connected apps Neotask's agent can operate directly. Search hotels and vacation rentals by location or property name to see photos, ratings, reviews, and direct contact details for the owner, then get a ready to...
- [Booking.com](https://neotask.ai/integrations/booking-com): Booking.com handles this without you switching tabs: Search and compare hotels, vacation homes, and other accommodations worldwide by price, rating, and amenities, and pull availability details for a stay.
- [Botlab](https://neotask.ai/integrations/botlab): With Botlab connected, Neotask's agent takes over the parts you'd normally do by hand. Have short automated phone calls placed on your behalf to confirm information like business hours or availability, returning a summary and transcript...
- [Botpress](https://neotask.ai/integrations/botpress): Create and iterate on an AI chatbot agent through conversation, describing its purpose, business logic, and knowledge sources, then open the result in a visual builder for further configuration.
- [Boxoffice](https://neotask.ai/integrations/boxoffice): Boxoffice is one of the connected apps Neotask's agent can operate directly. Ask for movie showtimes at nearby theaters, filter by genre or premium format like IMAX, and get direct ticketing links to book.
- [Braintrust MCP](https://neotask.ai/integrations/braintrust-mcp): Braintrust MCP handles this without you switching tabs: Review AI application traces, datasets, evaluation results, and prompt versions from a connected Braintrust account.
- [Brand24](https://neotask.ai/integrations/brand24): With Brand24 connected, Neotask's agent takes over the parts you'd normally do by hand. Summarize brand mentions, sentiment, and media coverage across social media, news, blogs, and forums, and track trends over time from a connected...
- [Breethe Meditations Made 4 You](https://neotask.ai/integrations/breethe-meditations-made-4-you): Describe a situation, activity, or intention to get a personalized 10-minute guided meditation audio generated on the spot for relaxation, sleep, or focus. Useful for a quick reset before a stressful moment or to wind down for sleep.
- [Bridebook](https://neotask.ai/integrations/bridebook): Instead of opening Bridebook yourself, ask Neotask. Browse thousands of UK wedding venues with real photos, pricing, and reviews from couples who got married there, then send enquiries directly or view full venue profiles.
- [broadbandnow.com](https://neotask.ai/integrations/broadbandnow-com): broadbandnow.com is one of the connected apps Neotask's agent can operate directly. Enter a zip code or city to see every major internet provider serving that area with pricing, speed tiers, service type, and customer ratings.
- [Browser Use Cloud MCP Server](https://neotask.ai/integrations/browser-use-cloud-mcp-server): Browser Use Cloud MCP Server handles this without you switching tabs: Run automated browser tasks in the cloud by describing a goal in natural language, letting an agent navigate websites, click, and extract information on your behalf.
- [Buf MCP Server](https://neotask.ai/integrations/buf-mcp-server): With Buf MCP Server connected, Neotask's agent takes over the parts you'd normally do by hand. Manage Protocol Buffer and gRPC schemas, generate code, and interact with resources stored in the Buf Schema Registry.
- [Builder MCP Publish Server](https://neotask.ai/integrations/builder-mcp-publish-server): Create, edit, and publish content and page sections for a connected Builder.io space, and manage headless CMS content used to power a website or app. Useful for content teams updating live site content or components without a developer.
- [Bump.sh APIs hub MCP server](https://neotask.ai/integrations/bump-sh-apis-hub-mcp-server): Instead of opening Bump.sh APIs hub MCP server yourself, ask Neotask. Search and read published API documentation from OpenAPI and AsyncAPI specifications hosted across an organization's API hub.
- [Bump.sh Portal docs MCP server](https://neotask.ai/integrations/bump-sh-portal-docs-mcp-server): Bump.sh Portal docs MCP server is one of the connected apps Neotask's agent can operate directly. Search and read a specific published API documentation portal, including endpoint details and guides.
- [Bump.sh Workspace docs MCP server](https://neotask.ai/integrations/bump-sh-workspace-docs-mcp-server): When an API team spreads its documentation across several published references inside one Bump.sh workspace, finding the right detail usually means opening reference after reference and scanning each one by hand.
- [Burger King](https://neotask.ai/integrations/burger-king): Deciding where to eat and what to order usually means opening an app, waiting for it to locate you, then scrolling a menu.
- [Busbud](https://neotask.ai/integrations/busbud): Comparing bus and train options for one trip across carriers like Greyhound, FlixBus, and Eurostar normally means checking each carrier's own site separately and keeping track of times and prices by hand.
- [Business Helper](https://neotask.ai/integrations/business-helper): Generic answers to strategy or marketing questions are easy to get anywhere; the harder thing is getting an answer grounded in real, trusted source material instead of a plausible-sounding guess.
- [Cafe24](https://neotask.ai/integrations/cafe24): Shopping across Cafe24-powered storefronts usually means visiting each store individually and hunting through its own catalog.
- [Calendly](https://neotask.ai/integrations/calendly): Calendly is where most teams keep event types, availability, and the links people use to book time on a calendar. Neotask connects to Calendly through Calendly's own MCP server at mcp.calendly.com, so an agent can create and update event...
- [Calorie Tracker](https://neotask.ai/integrations/calorie-tracker): Logging food is usually the part of tracking that people give up on first, because most apps want you to search a database and pick the exact match.
- [Calorie Tracker Guide MCP server](https://neotask.ai/integrations/calorie-tracker-guide-mcp-server): Calorie Tracker Guide MCP server is a narrower integration than a full tracking app — it's a connector purpose-built to link Neotask's agent to the Calorie Tracker Guide service so requests can be routed there directly.
- [Campfire](https://neotask.ai/integrations/campfire): Pulling a balance sheet together or answering "which vendor did we pay the most last quarter" usually means logging into the ERP and running the right report. Campfire connected to Neotask lets you ask the question instead.
- [CamScanner](https://neotask.ai/integrations/camscanner): A stack of paperwork — printed forms, handwritten notes, scanned contracts — is only useful once it's text you can search, summarize, or translate.
- [Canny](https://neotask.ai/integrations/canny): Feedback boards accumulate requests faster than anyone can read them, and the useful question is rarely "what did one customer say" — it's "what's trending, and for whom." Canny connected to Neotask lets you ask that directly instead of...
- [Capacities](https://neotask.ai/integrations/capacities): A knowledge base only earns its keep when notes are connected to each other, not filed away in isolation. Capacities connected to Neotask lets you build that structure by describing it, rather than manually creating each object and linking...
- [Cara](https://neotask.ai/integrations/cara): Car shopping across a nationwide inventory usually means filtering by feature after feature on a dealer site. Cara lets you describe the car you actually want — heated seats, a panoramic roof, "family friendly" — in plain language and get...
- [Carbon Arc](https://neotask.ai/integrations/carbon-arc): Private commercial and analytical data — the kind that supports real investment or market research — is usually locked behind a dashboard built for analysts, not a quick question.
- [Carbon Arc Research MCP](https://neotask.ai/integrations/carbon-arc-research-mcp): Carbon Arc Research MCP takes the same Carbon Arc data foundation and points it specifically at research workflows — discovering the right entities and ontologies for a question is the first step, not an afterthought.
- [CardChain AI](https://neotask.ai/integrations/cardchain-ai): Before submitting a trading card for grading, the real question is whether it's worth it — and that means knowing current pricing, price history, and how similar cards have graded.
- [CardChain AI MCP Server (Claude)](https://neotask.ai/integrations/cardchain-ai-mcp-server-claude): CardChain AI MCP Server (Claude) brings the same trading-card intelligence into Neotask through a Claude-specific server, covering the same core needs — pricing, grading data, and portfolio tracking — for collectors who want it in a...
- [CardChain MCP Server](https://neotask.ai/integrations/cardchain-mcp-server): CardChain MCP Server is the general CardChain connector for Neotask — the same market service behind the pricing and grading lookups, reached through its own dedicated endpoint rather than a client-specific one.
- [CARFAX](https://neotask.ai/integrations/carfax): Comparing cars for sale usually means weighing price against history, and those two things live in different places until you connect CARFAX.
- [CarGurus](https://neotask.ai/integrations/cargurus): CarGurus connects Neotask's agent directly to CarGurus, giving it a real channel into that listings and pricing service instead of relying on general knowledge about used-car shopping. CarGurus integration for connecting agents to CarGurus.
- [CarMax](https://neotask.ai/integrations/carmax): CarMax is an online and in-person used and new car retailer, and its integration lets an AI agent search live vehicle inventory using filters such as body style and fuel efficiency, then walk a shopper through the details of any match.
- [Carrefour](https://neotask.ai/integrations/carrefour): Carrefour is a European grocery retailer, and its chat integration turns a recipe idea or an upcoming event, a dinner party, a weekend brunch, a birthday cake, into a shopping cart that's ready to order rather than a list someone still has...
- [CARS24](https://neotask.ai/integrations/cars24): CARS24 operates a large used and new car marketplace in India, and its chat connection gives an agent access to that inventory along with the inspection reports attached to each pre-owned listing.
- [carsales](https://neotask.ai/integrations/carsales): carsales runs one of Australia's largest vehicle marketplaces, covering new, used, and demo listings, and its chat integration lets an agent search that inventory from a plain description rather than a form full of dropdowns.
- [Carta CRM](https://neotask.ai/integrations/carta-crm): Carta CRM is a deal and relationship tracking tool built for investment teams, and connecting it to an agent means pipeline work stops living in someone's head or a scattered set of documents.
- [CarWale](https://neotask.ai/integrations/carwale): CarWale is a research site for new cars sold in India, and its chat connection gives an agent direct access to pricing, specifications, and feature lists across hundreds of models, along with the expert and user reviews attached to each...
- [Cashback Hive: Deals & Coupons](https://neotask.ai/integrations/cashback-hive-deals-and-coupons): Cashback Hive tracks cashback rates and coupon codes across thousands of online retailers, and its chat integration means an agent can check what rate applies before someone clicks through to buy anything.
- [CB Insights](https://neotask.ai/integrations/cb-insights): CB Insights maintains data on private companies, their funding history, investors, and competitive relationships, and connecting it to an agent turns that data into something usable in the middle of actual analyst work rather than a...
- [Cdiscount](https://neotask.ai/integrations/cdiscount): Cdiscount is a French online retailer with a broad product catalog spanning electronics, home goods, and more, and its chat connection lets an agent search that catalog from a plain description instead of a keyword search that returns...
- [Channel3 MCP](https://neotask.ai/integrations/channel3-mcp): Channel3 MCP provides agentic commerce data that lets an agent search for products and compare pricing and availability across multiple retailers at once, rather than checking one store at a time.
- [Channel99](https://neotask.ai/integrations/channel99): Channel99 is a unified B2B marketing and go-to-market data platform that combines account identification with attribution reporting.
- [ChartMogul MCP server](https://neotask.ai/integrations/chartmogul-mcp-server): ChartMogul tracks subscription revenue for SaaS businesses, and its MCP server lets a Neotask agent pull those numbers straight into a conversation.
- [Cheat-Database](https://neotask.ai/integrations/cheat-database): Cheat-Database holds a large collection of video game cheats, codes, and walkthroughs spanning more than 30 platforms, and connecting it through Neotask lets an agent search that collection just by being told the name of the game.
- [CHECK24](https://neotask.ai/integrations/check24): CHECK24 is Germany's comparison portal for hotels, package vacations, flights, rental cars, tires, mobile plans, and insurance, and the Neotask connection lets an agent compare these categories and complete a booking or purchase without...
- [Checkatrade](https://neotask.ai/integrations/checkatrade): Checkatrade lists more than 50,000 vetted tradespeople across the UK for home improvement work, and connecting it through Neotask lets an agent search that directory on your behalf.
- [Checkly MCP Server](https://neotask.ai/integrations/checkly-mcp-server): Checkly runs synthetic monitoring checks against APIs, websites, and user workflows, and connecting it through Neotask lets an agent set up and review those checks in plain conversation.
- [Checkr Candidate MCP](https://neotask.ai/integrations/checkr-candidate-mcp): Checkr Candidate MCP is built for someone going through a background check as part of a hiring process, letting them ask about the status of their screening directly in conversation with a Neotask agent.
- [Checkr Customer MCP](https://neotask.ai/integrations/checkr-customer-mcp): Checkr Customer MCP is built for the hiring side of background screening, letting a hiring team manage and review checks they've ordered for candidates through conversation with a Neotask agent.
- [ChEMBL](https://neotask.ai/integrations/chembl): ChEMBL is a database of bioactive, drug-like compounds and the biological targets they interact with, and connecting it through Neotask lets an agent search that data directly in conversation.
- [Chessvia Openings](https://neotask.ai/integrations/chessvia-openings): Chessvia Openings covers chess openings and their variations, and connecting it through Neotask lets an agent walk you through exact move sequences for a named opening on request.
- [Chessy](https://neotask.ai/integrations/chessy): Chessy connects an AI agent to a live chess board where a full game can be played from the opening move through checkmate or resignation.
- [Chronosphere MCP server](https://neotask.ai/integrations/chronosphere-mcp-server): The Chronosphere MCP server integration is a direct connection between an agent and Chronosphere's own MCP endpoint, reached through a tenant-specific address rather than a shared public one that every customer would use.
- [Cinemô](https://neotask.ai/integrations/cinemo): Cinemô answers one question quickly: where can a specific movie or show actually be watched right now, without opening a single streaming app to check.
- [Clarify](https://neotask.ai/integrations/clarify): Clarify turns a CRM into something an agent can operate through plain conversation rather than a series of forms and dropdown menus scattered across several separate screens.
- [Clarity AI](https://neotask.ai/integrations/clarity-ai): Clarity AI connects an agent to fund-level sustainability classification under the EU's proposed SFDR 2.0 framework, a regulatory area many teams are only starting to prepare for.
- [Clerk](https://neotask.ai/integrations/clerk): Clerk's integration gives an agent direct access to current SDK code snippets and implementation guidance for adding authentication to an app, rather than relying on training data that may already be out of date by the time it actually...
- [Clidrive](https://neotask.ai/integrations/clidrive): Clidrive lets an agent search Spain's used car market from a plain description instead of a set of filter dropdowns spread across a listings site's own search page.
- [Clinical Trials](https://neotask.ai/integrations/clinical-trials): The Clinical Trials integration connects an agent directly to ClinicalTrials.gov, letting it search studies by condition, sponsor, or investigator instead of a person running that same search manually through the registry's own web...
- [Closai](https://neotask.ai/integrations/closai): Closai works from what someone has already bought rather than starting a conversation from scratch every single time a related question comes up.
- [Cloud Asset MCP server](https://neotask.ai/integrations/cloud-asset-mcp-server): The Cloud Asset MCP server gives an agent direct access to Google Cloud's Cloud Asset Inventory, letting it search, inventory, and analyze resources and IAM policy data across projects instead of a person pulling that information manually...
- [Cloud Bigtable Admin MCP server](https://neotask.ai/integrations/cloud-bigtable-admin-mcp-server): Cloud Bigtable Admin MCP server exposes Google Cloud Bigtable's administrative control plane to an AI agent through the Model Context Protocol.
- [Cloud Billing Budget MCP](https://neotask.ai/integrations/cloud-billing-budget-mcp): Cloud Billing Budget MCP connects an AI agent to Google Cloud's billing budget system, the part of Cloud Billing responsible for setting spend thresholds and triggering alerts on a billing account.
- [Cloud Logging MCP server](https://neotask.ai/integrations/cloud-logging-mcp-server): Cloud Logging MCP server gives an AI agent direct access to the log entries Google Cloud collects from resources and applications running in a project.
- [Cloud Monitoring MCP server](https://neotask.ai/integrations/cloud-monitoring-mcp-server): Cloud Monitoring MCP server connects an agent to the metrics, alerting policies, and dashboards Google Cloud Monitoring maintains for a project's infrastructure.
- [Cloud Resource Manager MCP server](https://neotask.ai/integrations/cloud-resource-manager-mcp-server): Cloud Resource Manager MCP server gives an agent access to the containers that hold Google Cloud resources: organizations, folders, projects, and the tags applied to them.
- [Cloud Run Admin API MCP server](https://neotask.ai/integrations/cloud-run-admin-api-mcp-server): Cloud Run Admin API MCP server puts an agent in direct touch with the control plane behind Google Cloud Run, the service that runs containerized workloads without requiring a team to manage the underlying servers.
- [Cloud Spanner MCP server](https://neotask.ai/integrations/cloud-spanner-mcp-server): Cloud Spanner MCP server gives an agent programmatic reach into Google Cloud Spanner, a distributed relational database built to serve globally consistent SQL workloads at scale.
- [Cloud SQL Admin MCP server](https://neotask.ai/integrations/cloud-sql-admin-mcp-server): Cloud SQL Admin MCP server opens up the administrative API behind Google Cloud SQL to an agent, the managed database service that runs MySQL, PostgreSQL, and SQL Server instances without a team having to operate the underlying servers...
- [Cloud Storage MCP server](https://neotask.ai/integrations/cloud-storage-mcp-server): Cloud Storage MCP server is a connection point between an AI agent and Google Cloud Storage, Google's object storage service used to hold files, backups, and static assets at scale.
- [Cloud Trace MCP server](https://neotask.ai/integrations/cloud-trace-mcp-server): Cloud Trace MCP server puts the distributed traces and spans Google Cloud Trace collects within an agent's reach, drawn from applications running across a service architecture.
- [Clutch](https://neotask.ai/integrations/clutch): Clutch sells used vehicles that have each passed a 210-point inspection before being listed, and it also publishes market valuations for cars found anywhere, not just on its own site.
- [Clutch.co](https://neotask.ai/integrations/clutch-co): Clutch.co lists thousands of vetted B2B service providers, agencies, consultants, and software partners, each backed by verified client reviews across categories like marketing, development, design, and IT services.
- [CMS Coverage](https://neotask.ai/integrations/cms-coverage): The CMS Coverage Database holds Medicare Part B coverage policy, including National Coverage Determinations and Local Coverage Determinations organized by topic or by the contractor that administers them.
- [Coalesce Quality MCP Server](https://neotask.ai/integrations/coalesce-quality-mcp-server): Synq's Coalesce Quality platform tracks data quality monitors, incidents, lineage, and alerts across connected data systems, and an agent working inside Neotask can pull all of that directly into conversation through this connection.
- [Cobalt MCP Server](https://neotask.ai/integrations/cobalt-mcp-server): Cobalt aggregates your bank, brokerage, transaction, budget, and subscription data in one place, and this integration lets an AI agent in Neotask read that aggregated picture and answer questions about it directly.
- [coches.net](https://neotask.ai/integrations/coches-net): coches.net is a Spanish vehicle marketplace covering new, used, and classic cars, searchable by make, model, price, year, and location.
- [CockroachDB Cloud MCP server](https://neotask.ai/integrations/cockroachdb-cloud-mcp-server): CockroachDB Cloud manages distributed SQL clusters, and this integration gives an AI agent in Neotask a way to handle provisioning, monitoring, and configuration for those clusters directly from conversation.
- [Code Tytor: Python](https://neotask.ai/integrations/code-tytor-python): Code Tytor: Python turns a plain description into working Python code, explains what an existing script is doing, and suggests ways to improve it.
- [Cogedim](https://neotask.ai/integrations/cogedim): Cogedim is a major French real estate developer that builds off-plan, or VEFA, housing developments, and this integration lets an AI agent in Neotask help someone explore that inventory and the process around it.
- [College Nation](https://neotask.ai/integrations/college-nation): College Nation connects an AI agent in Neotask to its education-focused catalog, giving you a conversational way to look up what's available rather than browsing a site yourself.
- [CometChat MCP server](https://neotask.ai/integrations/cometchat-mcp-server): Setting up in-app messaging, voice, video, or notifications usually means digging through docs across several SDKs and API references.
- [Common Room](https://neotask.ai/integrations/common-room): Building a target account list well normally means combining firmographic filters with actual buyer intent signals, which usually live in separate tools.
- [Compare the Market](https://neotask.ai/integrations/compare-the-market): Shopping UK car insurance usually means re-entering the same registration and postcode details across several comparison sites.
- [Computrabajo](https://neotask.ai/integrations/computrabajo): Job hunting across Latin America usually means checking country-specific job boards one at a time. Computrabajo connected to Neotask brings that into a single browsable flow: search by country or city, swipe through a carousel of listings,...
- [Conductor](https://neotask.ai/integrations/conductor): Brand visibility used to mean just search rankings; now it also means how often and how favorably a brand shows up in ChatGPT, Gemini, Copilot, and Perplexity answers. Conductor connected to Neotask tracks both sides at once.
- [Confident AI MCP Server](https://neotask.ai/integrations/confident-ai-mcp-server): LLM evaluation work tends to pile up in a separate browser tab from wherever the actual product conversation is happening — someone flags a bad model response, and checking whether it's a known regression means logging into Confident AI,...
- [ConfirmTkt](https://neotask.ai/integrations/confirmtkt): Checking train availability between two Indian cities usually means a form with a station picker, a date field, and a results page you scroll through manually.
- [Contact Center AI Insights MCP server](https://neotask.ai/integrations/contact-center-ai-insights-mcp-server): Contact center metrics and conversation-level analysis usually live in a Google Cloud console that only a couple of people on a team ever open.
- [Contentful MCP server (hosted, EU)](https://neotask.ai/integrations/contentful-mcp-server-hosted-eu): Managing structured content in Contentful — entries, assets, previews, who has access — is normally a handful of separate admin screens, and it matters here that the data stays on EU infrastructure.
- [Contentful MCP server (hosted, global)](https://neotask.ai/integrations/contentful-mcp-server-hosted-global): The global variant of the Contentful connection works the same way as the EU one, minus the region restriction — content entries and assets across your Contentful spaces route through Contentful's standard global infrastructure.
- [Contentsquare](https://neotask.ai/integrations/contentsquare): Contentsquare's whole pitch is understanding user behavior without wiring up a pile of custom tags first, and connecting it to Neotask keeps that promise going one step further — you can ask about sessions, errors, and conversion funnels...
- [Context.dev MCP server](https://neotask.ai/integrations/context-dev-mcp-server): Pulling structured information off the web — a competitor's pricing page, a company's public profile, a screenshot of a live site — usually means a separate scraping tool or a manual copy-paste job.
- [CorrentlyEnergy MCP](https://neotask.ai/integrations/correntlyenergy-mcp): Not every integration needs a long explanation — CorrentlyEnergy MCP connects Neotask's agent to the CorrentlyEnergy service so requests that would otherwise mean opening that platform directly can be handled from a normal conversation...
- [Cortellis Regulatory Intelligence](https://neotask.ai/integrations/cortellis-regulatory-intelligence): Regulatory research for drug or product approvals is exactly the kind of task that eats an afternoon — searching through licensed regulatory content, cross-checking requirements across jurisdictions, tracking down the right source...
- [Cosmic Hosted MCP (agent scope)](https://neotask.ai/integrations/cosmic-hosted-mcp-agent-scope): Agent-driven content work in Cosmic CMS needs its own access boundary — you don't want an agent creating projects with the same permissions as a human admin.
- [Cosmic Hosted MCP (bucket scope)](https://neotask.ai/integrations/cosmic-hosted-mcp-bucket-scope): Sometimes you just need to read what's already in a Cosmic bucket, not manage the whole project — the bucket-scope connection is built for exactly that narrower case.
- [Cottages](https://neotask.ai/integrations/cottages): Cottages.com's catalog of over 18,000 quality-checked UK cottage rentals is big enough that a location-and-dates search still leaves you scrolling — connected to Neotask, you can describe what you actually want (a location, a date range,...
- [Coveo](https://neotask.ai/integrations/coveo): Enterprise knowledge tends to be scattered across whatever content sources a company has plugged into its search platform, and Coveo exists to unify that search — connecting it to Neotask means you can ask a question in chat and have the...
- [Cowboy Model Finder](https://neotask.ai/integrations/cowboy-model-finder): Cowboy Model Finder connects Neotask's agent to the hosted Cowboy Model Finder service over its SSE endpoint, so whatever lookups that tool handles can be requested directly from chat rather than through a separate interface.
- [Craft](https://neotask.ai/integrations/craft): Craft is built as a workspace for notes, tasks, and structured documents, and connecting it to Neotask turns that workspace into somewhere the agent can actually work rather than just somewhere you write things down afterward.
- [Craime](https://neotask.ai/integrations/craime): Craime indexes Swedish court judgments, and searching that kind of archive by hand — matching a crime type against a specific court, or tracking down sentencing patterns for a case category — is slow, specialized work.
- [Create State](https://neotask.ai/integrations/create-state): Long-running software projects generate a lot of context that normally evaporates between sessions — the reasoning behind an architectural decision, a debugging insight that took an hour to reach, a snippet that solved a specific problem.
- [Creed MCP server](https://neotask.ai/integrations/creed-mcp-server): Repeating the same background and preferences to an AI assistant at the start of every conversation gets old fast. Creed solves that by keeping a structured Markdown profile of your context, preferences, and background that assistants read...
- [CrewAI MCP Server](https://neotask.ai/integrations/crewai-mcp-server): Keeping up with a fast-moving agent-framework ecosystem means checking blog posts, case studies, and event pages that don't always get bookmarked.
- [Cruise Critic](https://neotask.ai/integrations/cruise-critic): Cruise shopping usually means bouncing between a destination search, a price filter, and a date picker across multiple cruise-line sites before you can actually compare options.
- [Cruisebound](https://neotask.ai/integrations/cruisebound): Cruisebound connects Neotask's agent to Cruisebound's service through its OpenAI-compatible endpoint, so cruise-related requests can be handled from chat instead of the Cruisebound site directly.
- [Cube](https://neotask.ai/integrations/cube): Finance teams sitting on real numbers in Cube shouldn't have to open the platform every time someone asks for a variance breakdown or a quick forecast check.
- [CVpop](https://neotask.ai/integrations/cvpop): Writing a CV from scratch and then fighting with layout and formatting afterward is the part most people dread. CVpop connected to Neotask handles both halves in conversation: describe your background and career goals and the agent...
- [Daft.ie](https://neotask.ai/integrations/daft-ie): Property hunting on Daft.ie normally means setting a batch of filters and then paging through results one screen at a time.
- [DataCamp](https://neotask.ai/integrations/datacamp): Deciding what to learn next in SQL, Python, R, or AI is its own project — Neotask folds that decision into a normal conversation with DataCamp connected.
- [Dataflow MCP server](https://neotask.ai/integrations/dataflow-mcp-server): Checking on a Dataflow pipeline usually means opening the Google Cloud console, finding the right project, and clicking through to the job detail page.
- [DataForSEO MCP Server](https://neotask.ai/integrations/dataforseo-mcp-server): SEO research usually means bouncing between a rank tracker, a SERP viewer, and a backlink tool just to answer one question about a competitor.
- [Dataplex MCP server](https://neotask.ai/integrations/dataplex-mcp-server): In a large org, finding out which dataset or table actually has the data you need can eat an afternoon of asking around. Dataplex connected to Neotask turns that into a search: the agent browses Dataplex's Knowledge Catalog metadata to...
- [Dataproc MCP Server](https://neotask.ai/integrations/dataproc-mcp-server): Spark and Hadoop workloads on Google Cloud Dataproc still mean juggling cluster state, job submission, and status checks. With Dataproc connected, Neotask can manage clusters and jobs directly from a request — launching workloads and...
- [Datasite](https://neotask.ai/integrations/datasite): Running a Datasite virtual data room during an M&A deal involves a lot of fiddly setup work — folder structures, user roles, document uploads — plus ongoing tracking of buyer Q&A and access.
- [Datastream MCP server](https://neotask.ai/integrations/datastream-mcp-server): Change-data-capture streams quietly replicating your production database into BigQuery, Cloud Storage, Cloud SQL, or Spanner are easy to forget about until something breaks.
- [DatoCMS MCP Server](https://neotask.ai/integrations/datocms-mcp-server): Content teams running a headless CMS on DatoCMS spend real time in the model editor and content browser just to make routine updates.
- [Day AI](https://neotask.ai/integrations/day-ai): A CRM is only useful if updating it doesn't feel like a chore, and Day AI connected to Neotask removes most of that friction.
- [Dazzly](https://neotask.ai/integrations/dazzly): Building a real website — templates, pages, images, an ecommerce cart, hosting and SSL, basic SEO — is normally a string of separate decisions across a builder's UI.
- [DC PASS Contracts](https://neotask.ai/integrations/dc-pass-contracts): DC PASS Contracts connects Neotask's agent to the DC PASS Contracts service so requests that would otherwise mean logging into that system separately can be handled from the same conversation you're already having with the agent.
- [DealPilot](https://neotask.ai/integrations/dealpilot): Hunting for a working discount code before checkout usually means three browser tabs and a fair amount of guesswork about which codes still work.
- [Decolar](https://neotask.ai/integrations/decolar): Flight shopping means checking prices across dates and sometimes across destinations before you commit. With Decolar connected, Neotask searches flights by origin, destination, and travel dates, compares airline options, and can even...
- [Deel](https://neotask.ai/integrations/deel): Figuring out what it actually costs to hire someone in another country — and under which structure — is a research project on its own.
- [Deepgram hosted MCP server](https://neotask.ai/integrations/deepgram-hosted-mcp-server): Voice AI projects usually need both directions of audio handled well — turning speech into text and text back into speech — plus some project administration on the side.
- [Demandbase](https://neotask.ai/integrations/demandbase): B2B prospecting research means finding the right companies and the right people inside them, then understanding the account's broader context.
- [DepositPhotos](https://neotask.ai/integrations/depositphotos): Finding the right stock photo, vector, video, or music clip without leaving the flow of a project is the whole appeal here.
- [Descope MCP Server](https://neotask.ai/integrations/descope-mcp-server): Authentication and access-control settings — login methods, SSO, MFA, user management — are the kind of thing you configure rarely enough to forget where everything lives.
- [Despegar MCP Flights](https://neotask.ai/integrations/despegar-mcp-flights): For partner-facing travel bookings, flight data needs to come from the same B2B network the booking will actually run through.
- [Despegar MCP Hotels](https://neotask.ai/integrations/despegar-mcp-hotels): The hotel side of the same B2B network works the same way — availability and rates for partner bookings, retrievable directly for whatever booking workflow you're building on top of it, rather than a consumer-facing hotel search.
- [Developer Knowledge MCP server](https://neotask.ai/integrations/developer-knowledge-mcp-server): Digging through Google's developer documentation for the right API detail or SDK behavior usually means several search-and-scroll cycles.
- [DEWA](https://neotask.ai/integrations/dewa): Routine utility errands — checking a bill, paying it, finding an EV charger, locating a service center — are exactly the kind of task that's faster from a conversation than an app.
- [Dex](https://neotask.ai/integrations/dex): Personal relationship management lives or dies on whether you actually log the follow-up after a meeting, and Dex connected to Neotask makes that step nearly frictionless.
- [Dialogflow MCP server](https://neotask.ai/integrations/dialogflow-mcp-server): Building out a chatbot or voice interface in Dialogflow means managing agents, intents, and flows across either the ES or CX editions, which each have their own quirks.
- [Digits](https://neotask.ai/integrations/digits): Digits keeps a company's transactions, balances, and financial statements live and queryable, and Neotask's agent sits on top of that so you never have to open the app to get a number.
- [DirectBooker](https://neotask.ai/integrations/directbooker): Hotel search usually means five tabs open at once comparing neighborhoods, amenities, and price. DirectBooker collapses that into a conversation — describe the trip ("a boutique hotel near the Marais with a pool, under $300") and Neotask...
- [District](https://neotask.ai/integrations/district): Planning a night out in a city usually means jumping between a restaurant app, a ticketing site, and a maps app to see what's actually available.
- [Docket](https://neotask.ai/integrations/docket): In a live deal, the slow part is usually finding the right answer fast enough to keep momentum — a competitive comparison, a technical spec, a customer's past objection.
- [Docsie MCP](https://neotask.ai/integrations/docsie-mcp): Turning a training video or a pile of raw source files into an actual help center is normally a lot of manual restructuring.
- [DocuSeal](https://neotask.ai/integrations/docuseal): Getting a document signed usually starts with hunting for the right template and ends with checking who still hasn't signed.
- [Docusign](https://neotask.ai/integrations/docusign): Agreements pile up fast, and remembering which one renews next month or what a specific clause obligates you to is exactly the kind of thing that gets lost.
- [Domotz MCP Server (documented endpoint)](https://neotask.ai/integrations/domotz-mcp-server-documented-endpoint): Keeping tabs on network health across several client sites usually means logging into a monitoring dashboard site by site.
- [Domotz (Preview)](https://neotask.ai/integrations/domotz-preview): This is the preview build of the Domotz connection — the same network monitoring vendor as the documented Domotz endpoint, made available for agents to connect to ahead of full rollout.
- [DoneDeal](https://neotask.ai/integrations/donedeal): Car shopping in Ireland usually means scrolling through dozens of listings on DoneDeal, mentally filtering by fuel type and location as you go.
- [DoorDash](https://neotask.ai/integrations/doordash): Turning "we need groceries" or a recipe into an actual cart is the tedious part of grocery shopping, and DoorDash's grocery reach across stores like Kroger, Safeway, Sprouts, and local grocers gives Neotask real inventory to build from.
- [Dovetail](https://neotask.ai/integrations/dovetail): Customer research tends to live in a Dovetail workspace and stay there until someone remembers to dig it up for a decision.
- [Dow Jones Factiva](https://neotask.ai/integrations/dow-jones-factiva): Researching a company or an industry properly means going beyond a generic web search to sources you can actually cite. Dow Jones Factiva's archive covers premium Dow Jones sources and a large body of global news, and Neotask's agent can...
- [DraftKings](https://neotask.ai/integrations/draftkings): Checking on a game usually means opening an app and scrolling to find the score or the schedule you actually care about. DraftKings gives Neotask's agent a direct line into live scores, team performance, schedules, standings, and recent...
- [Draw Gen: Pyxl](https://neotask.ai/integrations/draw-gen-pyxl): Needing a quick visual — a mockup, a social post image, a slide illustration — usually means either opening a design tool or settling for a stock photo that doesn't quite fit.
- [Dremio-hosted MCP server](https://neotask.ai/integrations/dremio-hosted-mcp-server): Running analytics against a lakehouse usually means writing SQL by hand and separately tracking which jobs, catalogs, and engines are actually doing the work.
- [Driftless MCP server](https://neotask.ai/integrations/driftless-mcp-server): When a team and its AI agents are working out of a shared operational record, the value is in that record staying current for everyone reading it.
- [Dropbox (Legacy)](https://neotask.ai/integrations/dropbox-legacy): This is the legacy connection surface for Dropbox — the version of the integration kept available for accounts or workflows still wired to it, sitting alongside any newer Dropbox connection Neotask offers.
- [Dropbox Sign docs MCP server](https://neotask.ai/integrations/dropbox-sign-docs-mcp-server): Building an e-signature integration means constantly checking API reference material — which endpoint embeds a signing flow, how a webhook fires, what a status field actually means.
- [Dub Links MCP server](https://neotask.ai/integrations/dub-links-mcp-server): Short links are only useful if you actually look at what happens after someone clicks them, and that tracking usually lives in a separate dashboard nobody checks regularly.
- [Dub Partners MCP server](https://neotask.ai/integrations/dub-partners-mcp-server): Affiliate and referral programs get messy once there are more than a handful of partners to track — who has which link, what's converting, who to pay attention to.
- [Dunzo - Task Assistant](https://neotask.ai/integrations/dunzo-task-assistant): Small tasks and reminders scatter across sticky notes, phone reminders, and half-remembered intentions unless something catches them consistently.
- [EAN-Search MCP server](https://neotask.ai/integrations/ean-search-mcp-server): A barcode by itself is just a string of numbers until something looks it up — EAN-Search does that lookup against a large product database, and Neotask's agent can query it directly from an EAN, GTIN, UPC, or ISBN.
- [EaseMyTrip](https://neotask.ai/integrations/easemytrip): Comparing flights and hotels for a trip usually means opening several booking sites and mentally reconciling what you see across them.
- [Edmunds](https://neotask.ai/integrations/edmunds): Figuring out what a car is actually worth usually means guessing between a few online estimates that don't agree with each other.
- [eDreams](https://neotask.ai/integrations/edreams): eDreams runs one of the larger travel booking engines in Europe, and Neotask can sit in front of it as a conversational travel assistant.
- [edX](https://neotask.ai/integrations/edx): edX hosts course catalogs from Harvard, MIT, IBM, and dozens of other institutions, and normally finding the right one means filtering through a crowded search page.
- [elmah.io MCP server](https://neotask.ai/integrations/elmah-io-mcp-server): elmah.io is where a lot of teams' application errors and uptime data quietly pile up, and this connection turns that pile into something Neotask can dig through on request.
- [EMPLOYERS](https://neotask.ai/integrations/employers): Getting a workers' comp quote usually means filling out an insurer's intake form field by field. With EMPLOYERS connected, you can just describe the business — its type, where it's located, roughly how many people it employs, and payroll —...
- [Enhancv](https://neotask.ai/integrations/enhancv): Enhancv's whole job is turning resume content into something that looks like a professional actually designed it, and Neotask can trigger that directly.
- [Enterpret](https://neotask.ai/integrations/enterpret): Enterpret builds a knowledge graph out of everything customers say — support tickets, sales calls, app reviews, NPS responses, Slack threads — and normally getting an answer out of it means someone who knows the query language.
- [Envoy MCP server](https://neotask.ai/integrations/envoy-mcp-server): Envoy is the system a lot of offices run visitor check-ins, desk bookings, and deliveries through, and this connection lets Neotask act as the front desk when you'd rather not open the app.
- [Enzo Reader for RSS](https://neotask.ai/integrations/enzo-reader-for-rss): If you subscribe to more feeds than you can realistically read, Enzo Reader for RSS connected to Neotask turns "catching up" into a conversation instead of a scroll session.
- [EODHD MCP Server v1](https://neotask.ai/integrations/eodhd-mcp-server-v1): EODHD's v1 API is a deep well of market data — historical and live prices, company fundamentals, news, technical indicators, options, and macro figures — and connecting it to Neotask means you can pull any of that with a question instead...
- [Error Reporting MCP](https://neotask.ai/integrations/error-reporting-mcp): Google Cloud's Error Reporting service groups application errors and tracks how often each one fires, and this connection puts that directly in reach of a Neotask conversation.
- [Ethos](https://neotask.ai/integrations/ethos): Getting a real sense of term life insurance pricing usually means an application form up front, before you know if the number is even in range.
- [Evaneos](https://neotask.ai/integrations/evaneos): Evaneos is built around local travel experts crafting itineraries rather than a pure self-service booking engine, and connected to Neotask it works the same way in conversation.
- [Every AI](https://neotask.ai/integrations/every-ai): Every AI handles the back-office side of running a small business — invoices, tax calculation, expense tracking, proposals — and connected to Neotask it becomes something you can operate by describing what you need rather than clicking...
- [Excalidraw](https://neotask.ai/integrations/excalidraw): Excalidraw's hand-drawn diagram style is popular precisely because it doesn't feel like corporate clip art, and connected to Neotask you can build that directly from a description instead of dragging shapes.
- [Executor Dev](https://neotask.ai/integrations/executor-dev): Executor is a layer that sits between chat and a whole catalog of external tools and APIs, so instead of wiring up credentials for each one separately, you configure them once in Executor and Neotask calls them through a single policy...
- [Expedia](https://neotask.ai/integrations/expedia): Expedia's flight and hotel inventory is huge, and connected to Neotask you get real-time pricing and availability from it without opening the site.
- [Experian Insurance](https://neotask.ai/integrations/experian-insurance): Auto insurance pricing varies a lot by location, and Experian Insurance connected to Neotask gives you a localized estimate from just a ZIP code, drawing on Experian's own carrier network data.
- [Faces MCP server](https://neotask.ai/integrations/faces-mcp-server): Faces builds interactive, web-based slide decks rather than static PowerPoint files, and connected to Neotask you can manage that whole lifecycle from chat — creating a new deck, editing slides, and publishing the finished result.
- [FactSet AI-Ready Data](https://neotask.ai/integrations/factset-ai-ready-data): FactSet is a staple in institutional equity research, and this connection puts its content directly in reach of a Neotask conversation — global prices, fundamentals, earnings estimates, ownership data, M&A activity, and supply chain...
- [Faire Wholesale](https://neotask.ai/integrations/faire-wholesale): Faire connects independent brands with retail stores buying wholesale, and connected to Neotask you can search that catalog by describing what you're looking for — eco-friendly, handmade, seasonal, whatever fits the store — instead of...
- [fal MCP server](https://neotask.ai/integrations/fal-mcp-server): fal.ai hosts generative models for image, video, and audio generation on serverless infrastructure, and connected to Neotask you can trigger those runs directly instead of managing the fal dashboard yourself.
- [FareHarbor](https://neotask.ai/integrations/fareharbor): FareHarbor is the booking backbone behind a lot of tour and activity operators, and connected to Neotask it turns "find something to do" into a real, bookable search.
- [Fathom](https://neotask.ai/integrations/fathom): Fathom records and summarizes meetings, and connected to Neotask that archive becomes searchable in plain language instead of something you scroll through call by call.
- [FeedClaw](https://neotask.ai/integrations/feedclaw): FeedClaw is built around a human checkpoint before anything goes out publicly — drafts get reviewed in an inbox before they're published to X or sent as email.
- [Fern Docs MCP server](https://neotask.ai/integrations/fern-docs-mcp-server): A lot of API onboarding time gets burned just hunting through documentation for the right endpoint or the right SDK call. The Fern Docs connection points Neotask's agent at documentation generated with Fern, so when you ask a question like...
- [Ferryhopper](https://neotask.ai/integrations/ferryhopper): Planning a ferry leg of a trip usually means checking a handful of route/date combinations before anything makes sense. Connected to Ferryhopper, Neotask's agent can answer questions in plain language — "is there a ferry from Piraeus to...
- [Fever Event Discovery](https://neotask.ai/integrations/fever-event-discovery): "What's a good show happening this weekend" is a question that usually takes ten browser tabs to answer properly. With Fever connected, Neotask's agent searches live entertainment events worldwide — concerts, festivals, immersive...
- [Figaro Immobilier](https://neotask.ai/integrations/figaro-immobilier): House hunting in France usually means re-entering the same filters — location, budget, size, buy vs. rent — across multiple searches.
- [Fin Agent API MCP Server](https://neotask.ai/integrations/fin-agent-api-mcp-server): Support teams running on Intercom accumulate a lot of institutional knowledge inside conversations and help center articles that's hard to act on outside Intercom itself.
- [FINN](https://neotask.ai/integrations/finn): Car subscriptions are supposed to be simpler than financing or leasing, but comparing plans still takes some digging. Through FINN, Neotask's agent can browse car subscription plans with a fixed monthly rate — insurance, maintenance,...
- [Fintables](https://neotask.ai/integrations/fintables): Following Borsa Istanbul names closely means juggling real-time quotes, balance sheets, dividend data, and brokerage targets across multiple sources.
- [Fintual](https://neotask.ai/integrations/fintual): Retirement math is the kind of thing people put off because running the numbers feels like a project. Fintual connected to Neotask turns it into a conversation: describe your situation and it simulates retirement savings scenarios and...
- [Firehouse Subs](https://neotask.ai/integrations/firehouse-subs): Deciding on lunch shouldn't require opening an app just to see what's on the menu near you. With Firehouse Subs connected, Neotask can find the nearest location, walk through the menu including anything new, and check what deals are...
- [Firestore MCP Server](https://neotask.ai/integrations/firestore-mcp-server): Firestore is often where the actual state of an application lives, and poking at it usually means opening the console or writing a one-off script.
- [Fiscal.ai](https://neotask.ai/integrations/fiscal-ai): Earnings season moves fast, and the value of financial data drops the longer it takes to get from filing to analysis. Fiscal.ai connected to Neotask pulls institutional-grade metrics, ratios, and company-specific KPIs shortly after they're...
- [Flagsmith MCP Server](https://neotask.ai/integrations/flagsmith-mcp-server): Feature flag hygiene tends to slip when toggling a flag or checking a segment requires a trip to a separate dashboard. With Flagsmith connected, Neotask's agent can manage feature flags and remote configuration directly — toggling flags,...
- [Flarehawk MCP Server](https://neotask.ai/integrations/flarehawk-mcp-server): Security incidents rarely show up as one clean alert — they're usually scattered telemetry from edge, gateway, and application layers that someone has to piece together.
- [Flight Network](https://neotask.ai/integrations/flight-network): Flight Network connects agents to Flight Network's flight search and booking surface, giving Neotask a way to work flight-related requests into the same conversation as the rest of your travel planning instead of routing you to a separate...
- [FlixBus](https://neotask.ai/integrations/flixbus): Cross-country and cross-border trips by bus or train often come down to comparing a handful of departure times before booking.
- [Flixor](https://neotask.ai/integrations/flixor): "What should I watch" is a hard question to answer from a static genre list, because the real answer usually depends on mood more than category.
- [FMP MCP Server](https://neotask.ai/integrations/fmp-mcp-server): Financial Modeling Prep covers a lot of ground — stocks, fundamentals, historical prices, economic indicators, ETFs, forex, and crypto — and this connection puts all of it in reach of Neotask's agent for market research, financial...
- [Formbyte](https://neotask.ai/integrations/formbyte): Building a form usually means opening a builder, dragging fields around, and configuring logic by hand. Formbyte connected to Neotask lets you describe the form, survey, or poll you need in plain language and it builds it, then keeps...
- [Formula Genius](https://neotask.ai/integrations/formula-genius): Nobody enjoys staring at a broken VLOOKUP or a regex that almost works. Formula Genius connected to Neotask writes, explains, and debugs formulas for Excel and Google Sheets, and extends past spreadsheets into SQL queries, regex patterns,...
- [Fotocasa](https://neotask.ai/integrations/fotocasa): Searching Spanish property listings usually means re-filtering by city, price, and room count every time you refine what you want.
- [Foxit PDF Editor](https://neotask.ai/integrations/foxit-pdf-editor): PDF busywork — rotating a page, merging two contracts, splitting out one section — rarely needs a full desktop app open just for that.
- [Framagit MCP](https://neotask.ai/integrations/framagit-mcp): Framasoft's self-hosted GitLab instance runs a bit outside the usual GitHub-centric workflow, which can make it easy to overlook in tooling.
- [Framagit Orbit MCP](https://neotask.ai/integrations/framagit-orbit-mcp): This is the Orbit-side integration for Framagit — connecting Neotask's agent to Framagit Orbit MCP directly, so requests that touch that surface can be handled inside the same conversation rather than through a separate tool.
- [Frankfurter MCP server](https://neotask.ai/integrations/frankfurter-mcp-server): Frankfurter is a currency data feed built on rates published by central banks and official government sources, and Neotask can query it directly inside a conversation.
- [Freediver](https://neotask.ai/integrations/freediver): Freediver is training-log software built specifically for the sport, and connecting it to Neotask means you can plan a session — say, a CO2 tolerance table or a set of dynamic apnea repeats — by describing it in chat rather than building...
- [freee確定申告](https://neotask.ai/integrations/freee): freee確定申告 is a Q&A layer over Japan's kakutei shinkoku (final tax return) process, drawing its answers from real advice given by licensed tax accountants across more than ten thousand past consultations.
- [Freshworks Developer MCP server](https://neotask.ai/integrations/freshworks-developer-mcp-server): This connection puts the Freshworks developer platform's own documentation in reach of Neotask, aimed at people building or submitting custom apps that plug into Freshdesk and the rest of the Freshworks suite.
- [Frontegg Agen for SaaS MCP Gateway](https://neotask.ai/integrations/frontegg-agen-for-saas-mcp-gateway): Frontegg's SaaS MCP Gateway sits over customer identity and access management for a B2B product, so connecting it lets Neotask handle the operational side of that — looking up a user, reviewing SSO configuration, checking SCIM provisioning...
- [Fugle-Stock](https://neotask.ai/integrations/fugle-stock): Fugle-Stock is built around tracking Taiwanese equities, and connected to Neotask it becomes a conversational watchlist manager: add a ticker, jot an investment note against it, or set an alert for a price or valuation move, all without...
- [Full Stack GTM MCP server](https://neotask.ai/integrations/full-stack-gtm-mcp-server): Full Stack GTM's MCP server audits CRM data — HubSpot or Salesforce — in a strictly read-only mode and then proposes a staged plan for cleaning it up, with any actual write-back held behind human approval.
- [Fullstory MCP](https://neotask.ai/integrations/fullstory-mcp): Fullstory captures how people actually behave inside your web and mobile product — session replays, click paths, rage clicks, the works — and connecting it to Neotask means you can ask questions about that behavioral data directly.
- [Fun Generators MCP Server](https://neotask.ai/integrations/fun-generators-mcp-server): Fun Generators is a grab-bag of random-content utilities — names, trivia, riddles, lottery numbers, barcodes, QR codes, fake identities, UUIDs, and more — reachable straight from a Neotask conversation.
- [Fun Translations MCP server](https://neotask.ai/integrations/fun-translations-mcp-server): Fun Translations turns ordinary text into one of over 150 novelty, fictional, or encoding styles — Pirate speak, Yoda speak, Morse code, and plenty of stranger ones — and Neotask can run the conversion for you directly.
- [Function Health](https://neotask.ai/integrations/function-health): Function Health gives you a secure window into your own lab results across more than 160 tests, and once connected Neotask can help you actually make sense of them — ask about a specific thyroid, hormone, or inflammation marker and get an...
- [Fyxer](https://neotask.ai/integrations/fyxer): Fyxer writes email drafts that sound like you, because it learns from the messages you've already sent and pulls in context from your calendar and meeting notes when it puts a draft together.
- [G2](https://neotask.ai/integrations/g2): G2 tracks buyer intent and customer reviews across the software market, and connected to Neotask it becomes a research tool you can query mid-conversation: look up how a competitor is reviewed, compare two products in the same category, or...
- [Gainsight (Staircase AI)](https://neotask.ai/integrations/gainsight-staircase-ai): Staircase AI, part of Gainsight, aggregates live customer-intelligence signals — sentiment, churn risk, engagement trends — so that whatever an agent tells you about an account is grounded in real, current data rather than a stale CRM...
- [Galileo MCP Server](https://neotask.ai/integrations/galileo-mcp-server): Galileo is built for evaluating and monitoring LLM applications and autonomous agents in production, and connected to Neotask it lets you interrogate that evaluation layer conversationally — pull the traces from a specific run to see what...
- [Gametime - Last Minute Tickets](https://neotask.ai/integrations/gametime-last-minute-tickets): Gametime specializes in last-minute tickets to sports, concerts, and theater — including seats you can still grab up to 90 minutes after an event has already started.
- [Gathrd](https://neotask.ai/integrations/gathrd): Gathrd is a way of saving and organizing products you come across while chatting, instead of losing track of a link the moment the conversation moves on.
- [Genspark AI Slides](https://neotask.ai/integrations/genspark-ai-slides): Genspark AI Slides turns a topic, outline, or existing document into a complete slide deck, and connected to Neotask you can describe the subject in one message and get a formatted set of slides back, ready to present or keep editing.
- [GetMeDesign](https://neotask.ai/integrations/getmedesign): GetMeDesign is a marketplace for finding freelance and full-time creative and technical talent, and connected to Neotask you can ask it to surface professionals for a specific project by browsing profiles, portfolios, and experience...
- [GetYourGuide](https://neotask.ai/integrations/getyourguide): GetYourGuide catalogs more than 75,000 tours, activities, and attractions worldwide, and connected to Neotask it turns trip planning into a conversation — search by destination, by interest like food or history, or by a specific landmark,...
- [GitMCP remote MCP server](https://neotask.ai/integrations/gitmcp-remote-mcp-server): GitMCP maps any public GitHub repository's URL to a hosted MCP endpoint, so Neotask can read a project's README, documentation, and code structure directly, without you cloning the repo locally first.
- [GKE MCP server](https://neotask.ai/integrations/gke-mcp-server): This connects Neotask to Google Kubernetes Engine so you can check on your clusters conversationally — view a cluster's status, look at node pool configuration, or see what workloads are running, across your Google Cloud project.
- [Glama MCP Gateway](https://neotask.ai/integrations/glama-mcp-gateway): Glama MCP Gateway is a connection profile-based gateway that routes Neotask's requests to whatever tools you've set up behind your Glama endpoint, giving you a single managed access point instead of wiring up each tool separately.
- [Glama MCP Inspector](https://neotask.ai/integrations/glama-mcp-inspector): If you're building or wiring up MCP servers, Glama gives you a place to actually see what you're connecting to before you commit to it.
- [Glow](https://neotask.ai/integrations/glow): Glow is built around a fairly specific and human problem: finding the right person, not just any person. Whether that's a hiking partner, a mentor, a co-founder-shaped collaborator, or someone in a very particular niche, Glow works from a...
- [GoCardless](https://neotask.ai/integrations/gocardless): GoCardless connects less like an app you operate and more like a reference desk for building payments — it knows its own API inside and out, including endpoint details, integration patterns, webhook handling, and code samples across...
- [Gogs Docs MCP server](https://neotask.ai/integrations/gogs-docs-mcp-server): Gogs is a lightweight, self-hosted Git service, and this server is really a documentation lookup layer for it — covering repository hosting, users, organizations, and Git operations over web, SSH, and HTTP.
- [Gogs Docs MCP server (Mintlify host)](https://neotask.ai/integrations/gogs-docs-mcp-server-mintlify-host): Same underlying Gogs documentation as the standard docs server above, just served from a different host — a Mintlify-hosted documentation site rather than the default.
- [Goodnotes](https://neotask.ai/integrations/goodnotes): Goodnotes closes a gap that shows up constantly in AI conversations: you talk through a diagram, a sketch, or a well-formatted document with an assistant, and then it just... lives in the chat window.
- [Google Cloud Pub/Sub MCP server](https://neotask.ai/integrations/google-cloud-pub-sub-mcp-server): Pub/Sub is the plumbing behind a lot of asynchronous systems, and managing it — topics, subscriptions, snapshots, schemas — is exactly the kind of infrastructure bookkeeping that's easy to let get stale.
- [Google Cloud Support MCP server](https://neotask.ai/integrations/google-cloud-support-mcp-server): When something breaks on Google Cloud and you need to escalate to Google's own support team, this is the connection that handles it — creating, viewing, and updating technical support cases, along with their attachments, comments, and...
- [Google Compute Engine](https://neotask.ai/integrations/google-compute-engine): Compute Engine is where a lot of infrastructure actually lives, and this connection gives Neotask real control over it: creating, starting, stopping, resetting, and deleting virtual machine instances, plus inspecting disks, machine types,...
- [Google Play Developer Reporting MCP server](https://neotask.ai/integrations/google-play-developer-reporting-mcp-server): If you ship an Android app, this connection puts your Play Console quality metrics within reach of a conversation — crash rates, ANR rates, and other Android vitals data for apps in your Google Play developer account.
- [Google Shopping Content MCP](https://neotask.ai/integrations/google-shopping-content-mcp): This connection reaches into Google Merchant Center through the Content API for Shopping — managing product listings and shopping data, adding or updating products, and checking account status.
- [Gorgias MCP server](https://neotask.ai/integrations/gorgias-mcp-server): Gorgias is ecommerce customer support infrastructure, and this connection lets Neotask work inside it directly — reading and responding to customer tickets and conversations, and checking on automation rules.
- [GovTribe](https://neotask.ai/integrations/govtribe): Government contracting data is notoriously scattered and hard to search meaningfully — GovTribe pulls together federal, state, and local contract opportunities, awards, grants, and vendor data and lets you search it in plain language.
- [Granted](https://neotask.ai/integrations/granted): Granted specializes in the grants a bigger database would miss — niche state programs, small family foundations, that kind of long-tail funding that's hard to find by searching alone.
- [GraphOS MCP Tools](https://neotask.ai/integrations/graphos-mcp-tools): For anyone building on GraphQL, this connection puts Apollo's official documentation and the Apollo Connectors specification directly in reach — schema design guidance, deployment practices, and specifically how to build REST-to-GraphQL...
- [Grep MCP Server](https://neotask.ai/integrations/grep-mcp-server): Grep answers a question every developer has asked at some point: how does someone else actually use this function or library in real code? It searches code, files, and paths across public GitHub repositories at scale, so instead of...
- [Greptile MCP Server](https://neotask.ai/integrations/greptile-mcp-server): Greptile reviews code with whole-repository context rather than just looking at the diff in isolation, which matters for catching the bugs, style drift, and logic problems that only show up once you understand how a change fits into the...
- [Greystar Apartment Scout](https://neotask.ai/integrations/greystar-apartment-scout): Apartment hunting across a big curated inventory is exactly the kind of search-and-filter task that's tedious by hand and quick with an agent doing the legwork. Greystar Apartment Scout covers rental homes across most major U.S.
- [groundcover MCP server](https://neotask.ai/integrations/groundcover-mcp-server): Most observability tools make you learn a query language before you can answer a simple question like "why is checkout slow." groundcover's MCP server skips that step: it hands your agent direct access to the metrics, logs, and traces...
- [GS Plant Foods UCP MCP server](https://neotask.ai/integrations/gs-plant-foods-ucp-mcp-server): GS Plant Foods sells organic and natural fertilizers, soil products, and garden supplies through Shopify, and this connector lets an agent shop that storefront directly on your behalf.
- [Gumloop MCP Server](https://neotask.ai/integrations/gumloop-mcp-server): If your team already builds automations and agent workflows in Gumloop that reach into internal and external tools, this connection means those flows don't have to live behind a separate dashboard anymore.
- [Gusto](https://neotask.ai/integrations/gusto): Gusto holds the payroll and people data most companies otherwise dig for manually — employee records, contractor details, compensation, time sheets, pay schedules.
- [Habitify](https://neotask.ai/integrations/habitify): Habitify turns habit tracking into something you can just talk through instead of tapping through an app. Mark a workout done, log a reading session, check how a streak is holding up — all through conversation.
- [HAPI MCP Registry](https://neotask.ai/integrations/hapi-mcp-registry): Before connecting to yet another MCP server, it helps to know what's actually out there. The HAPI MCP Registry is a centralized, community-supported directory of publicly accessible MCP servers, and this connection lets you search it by...
- [Happenstance](https://neotask.ai/integrations/happenstance): Happenstance treats your professional network as something you can query, not just scroll through. Describe who you're looking for in plain language — the right person for an introduction, someone with specific expertise — and it searches...
- [Harvey](https://neotask.ai/integrations/harvey): Harvey is built for legal work specifically, and this connection reflects that: you can ask general legal questions, run analysis over a specific Vault project, or research a question against curated legal knowledge sources.
- [hauptsache.net MCP server](https://neotask.ai/integrations/hauptsache-net-mcp-server): hauptsache.net runs both a web agency and a separate document service called Komitee Akten, and this server is the bridge to that infrastructure.
- [Headout](https://neotask.ai/integrations/headout): Planning what to actually do in a city is usually the slow part of travel planning, and Headout is built for exactly that step.
- [HealthEx](https://neotask.ai/integrations/healthex): HealthEx gives you a conversational way into your own health records — medications, lab results, vitals, allergies, immunizations, conditions, and care plans — all accessed securely.
- [Heepsy Influencer Search](https://neotask.ai/integrations/heepsy-influencer-search): Heepsy Influencer Search connects your agent to Heepsy's influencer search platform, giving you a conversational path into the same influencer discovery capability the tool is built around.
- [Heffl MCP server](https://neotask.ai/integrations/heffl-mcp-server): Heffl covers a lot of the day-to-day CRM and business operations ground — contacts, companies, deals, tasks, quotations, projects, and forms — and this connection puts all of it inside your AI client.
- [Helium 10](https://neotask.ai/integrations/helium-10): Helium 10's ASIN Insights capability turns an Amazon ASIN or product URL into an instant performance snapshot — the keywords actually driving traffic, sales and price trends, review analysis, and a listing health score with concrete...
- [Help Scout](https://neotask.ai/integrations/help-scout): Help Scout syncs your support mailboxes and conversations so past customer support history becomes something you can search and reason over rather than something buried in a mailbox archive.
- [hepsiemlak](https://neotask.ai/integrations/hepsiemlak): Hepsiemlak covers hundreds of thousands of for-sale and rental real estate listings across Turkey — apartments, houses, commercial properties, land — and this connection lets you search that catalog in everyday language rather than...
- [Hercules MCP](https://neotask.ai/integrations/hercules-mcp): Hercules is built to take a described idea straight to a running app — web apps, internal tools, ecommerce sites, mobile apps — with hosting, database, auth, and payments handled as part of the build rather than separate setup steps.
- [HeyGen](https://neotask.ai/integrations/heygen): HeyGen turns a prompt, a script, or even a conversation into a finished talking-head video, complete with AI avatars, voices, and visual scenes.
- [Hibbett](https://neotask.ai/integrations/hibbett): Hibbett lets you shop sneakers, athletic apparel, and sports gear from brands like Nike, Adidas, New Balance, and Jordan just by describing what you're after.
- [HighLevel](https://neotask.ai/integrations/highlevel): HighLevel pulls your CRM data — contacts, opportunities, appointments, conversations — into the conversation itself, turning raw records into summaries, next-step recommendations, and ready-to-use content.
- [Hindsight Cloud MCP server (multi-bank)](https://neotask.ai/integrations/hindsight-cloud-mcp-server-multi-bank): Hindsight Cloud gives your agent persistent memory that survives across sessions, and the multi-bank version of this server splits that memory into multiple isolated banks.
- [Hindsight Cloud MCP server (single-bank)](https://neotask.ai/integrations/hindsight-cloud-mcp-server-single-bank): This is the single-bank version of Hindsight Cloud's memory server: one persistent memory bank the agent can store facts into and retrieve from across interactions, without the added complexity of managing multiple banks.
- [Hjarni](https://neotask.ai/integrations/hjarni): Hjarni is a place to write and organize markdown notes in folders and tags that your AI can read directly, so context doesn't have to be copy-pasted into every conversation.
- [エネチェンジ Home 電気・ガス比較](https://neotask.ai/integrations/home): エネチェンジ Home compares electricity and gas plans available across Japan based on your region and household size, estimating annual costs across providers so you're not left guessing.
- [HomeGraph MCP](https://neotask.ai/integrations/homegraph-mcp): Google's Home Graph is the state layer behind Google Home's cloud-to-cloud platform — it knows what every synced smart home device reports as its current status, and it can be told to resync.
- [Horizon-hosted MCP server endpoint](https://neotask.ai/integrations/horizon-hosted-mcp-server-endpoint): This is a self-hosted MCP endpoint on the Horizon platform (fastmcp.app) — a server you or your team stood up to expose your own tools and data to any MCP-aware agent.
- [Hosted X MCP](https://neotask.ai/integrations/hosted-x-mcp): Hosted X MCP is X's (formerly Twitter's) own integration for letting an agent act on the platform — reading posts, checking profiles, and posting on your behalf, plus touching account and advertising data through X's API.
- [Hostinger](https://neotask.ai/integrations/hostinger): Hostinger Horizons turns a plain description into a real, running website or app and puts it live with one click — built for people who don't want to touch code.
- [Hot100.ai](https://neotask.ai/integrations/hot100-ai): Hot100.ai tracks hundreds of AI projects, tools, and agents, ranking them weekly and tagging each by tech stack, use case, and the builder tools behind them (Cursor, Replit, and similar).
- [HumanCopy](https://neotask.ai/integrations/humancopy): HumanCopy takes AI-generated text and rewrites it so it reads like a person wrote it, with adjustable tone and style. It's built for polishing the emails, articles, and support replies that come out of an AI draft sounding a little stiff.
- [Hyatt](https://neotask.ai/integrations/hyatt): Hyatt's connection lets you search hotels and resorts worldwide, compare room types, amenities, and live rates, then jump to Hyatt.com to finish booking.
- [Hygraph MCP Server](https://neotask.ai/integrations/hygraph-mcp-server): Hygraph is a headless content management platform, and this MCP server is the bridge that lets an agent work with your Hygraph content graph directly.
- [Hyperping MCP server](https://neotask.ai/integrations/hyperping-mcp-server): Hyperping is a status-page and uptime monitoring tool, and this MCP server pulls that monitoring data into a conversation.
- [IBM watsonx BI remote MCP server](https://neotask.ai/integrations/ibm-watsonx-bi-remote-mcp-server): This MCP server reaches into IBM Cloud and watsonx BI, pulling account, catalog, and analytics data into a conversation for reporting and diagnostics.
- [ICD-10 Codes](https://neotask.ai/integrations/icd-10-codes): ICD-10 Codes gives you a fast, accurate way to look up, validate, and browse ICD-10-CM diagnosis codes and ICD-10-PCS procedure codes — by code, description, category, or body system.
- [ICE Data Services](https://neotask.ai/integrations/ice-data-services): ICE Data Services gives fixed income analysts real-time and historical trade data for corporate, agency, securitized, and municipal bonds reported to FINRA TRACE and MSRB, plus reference data like coupons and maturities.
- [Icons8 MCP Server](https://neotask.ai/integrations/icons8-mcp-server): Icons8's library holds icons, illustrations, photos, and other design assets, searchable by keyword or style. Rather than browsing the Icons8 website to find the right visual, Neotask can search and pull a matching asset straight into your...
- [idealista](https://neotask.ai/integrations/idealista): Idealista covers property listings for sale or rent across Spain, Italy, and Portugal, searchable with plain-language descriptions of location, price, and features — photos and maps included per listing.
- [idealo](https://neotask.ai/integrations/idealo): Idealo compares prices for products across thousands of online shops — including vouchers, delivery costs, shipping times, and shop ratings — and surfaces product reviews and expert opinions before you buy.
- [IFTTT](https://neotask.ai/integrations/ifttt): IFTTT connects over a thousand apps and smart home devices through triggers, actions, and prebuilt Applets. Neotask can search that catalog of services and Applets, and create or enable new Applets that link two apps together, so...
- [ImmoScout24](https://neotask.ai/integrations/immoscout24): ImmoScout24 is Germany's largest real estate marketplace, and Neotask can search apartments, houses, and investment properties there using natural-language criteria, then browse photos, pricing, and neighborhood details per listing.
- [Imovirtual](https://neotask.ai/integrations/imovirtual): Imovirtual covers sale and rental listings across Portugal. Describe an ideal property in plain language and Neotask gets matched listings back, then lets you refine with follow-up questions and compare details — useful for first-time...
- [incident.io](https://neotask.ai/integrations/incident-io): incident.io runs incident response — declaring and triaging incidents, checking on-call schedules and escalation status, managing follow-ups, and querying telemetry and catalog data across the workspace.
- [Infermedica MCP server](https://neotask.ai/integrations/infermedica-mcp-server): Infermedica's MCP server connects Neotask to Infermedica's medical intelligence platform, giving an agent a route into whatever capabilities that server exposes.
- [Inline MCP server](https://neotask.ai/integrations/inline-mcp-server): Inline is a team messaging workspace, and this MCP server lets an agent read and interact with the approved conversations and content inside it, plus trigger bot automations in team chats.
- [InPost](https://neotask.ai/integrations/inpost): InPost handles parcel lockers and delivery, and this connection checks the current delivery status of one or more InPost parcels at once, right from the conversation.
- [Insight Timer](https://neotask.ai/integrations/insight-timer): Insight Timer's free library holds guided meditations, sleep tracks, breathwork sessions, ambient soundscapes, and talks. Rather than scrolling the app, describe what you want — a duration, a specific teacher, or a topic — and Neotask...
- [Instacart](https://neotask.ai/integrations/instacart): Instacart lets you search groceries and household products, build a cart, and check what's available nearby for delivery in as fast as 30 minutes.
- [Insurance GPT](https://neotask.ai/integrations/insurance-gpt): Insurance back offices run on forms: proposal templates, policy schedules, claims intake sheets. Insurance GPT replaces the form with a conversation.
- [Insurify](https://neotask.ai/integrations/insurify): Shopping for car insurance usually means visiting three or four carrier sites, re-entering your driver profile each time, and squinting at rate tables that don't line up.
- [integrations.sh MCP server](https://neotask.ai/integrations/integrations-sh-mcp-server): Before wiring up a new integration, most developers spend the first hour just figuring out what a service actually exposes — is there a REST API, a GraphQL schema, an MCP server, a CLI, and how does auth work? integrations.sh is an...
- [Internshala](https://neotask.ai/integrations/internshala): Internshala handles this without you switching tabs: search internships and entry-level jobs across India by role, location, stipend, or PPO availability, and view listings from trusted employers.
- [Interzoid MCP Server](https://neotask.ai/integrations/interzoid-mcp-server): Duplicate customer records, addresses spelled three different ways, names with inconsistent casing — every data team eventually inherits a mess like this.
- [Intuit Credit Karma](https://neotask.ai/integrations/intuit-credit-karma): A credit score by itself doesn't tell you what to fix. Intuit Credit Karma does — payment history, credit utilization, account age, credit mix, recent inquiries — and Neotask can pull that breakdown into a conversation so you're not...
- [Intuit QuickBooks](https://neotask.ai/integrations/intuit-quickbooks): Instead of opening QuickBooks yourself, ask Neotask. Generate profit and loss and cash flow statements, benchmark your business performance against industry peers, and import transactions from CSVs, PDFs, images, or pasted text into...
- [invideo](https://neotask.ai/integrations/invideo): Turn a text prompt into a finished video, complete with script, visuals, voiceover, and music, without manual editing. Useful for creators, marketers, and businesses who want to produce shareable video content quickly from an idea.
- [Inviton](https://neotask.ai/integrations/inviton): "What's happening tonight" is a harder question than it should be once you're outside your home city. Inviton answers it for Slovakia, Czechia, Poland, and nearby regions — concerts, festivals, exhibitions, nightlife — and Neotask can...
- [IPinfo MCP Server](https://neotask.ai/integrations/ipinfo-mcp-server): IPinfo MCP Server handles this without you switching tabs: look up IP address details such as geolocation, network or ASN ownership, and proxy or VPN detection for any given IP.
- [isango!](https://neotask.ai/integrations/isango): With isango! connected, Neotask's agent takes over the parts you'd normally do by hand. Search and compare tours, attraction tickets, transfers, cruises, and local experiences across major cities worldwide by budget and interest.
- [italki language learning](https://neotask.ai/integrations/italki-language-learning): Browse language teachers and book one-on-one or group lessons across more than 150 languages based on your schedule, level, and budget. Useful for language learners who want flexible, per-lesson study with native speakers.
- [ixigo](https://neotask.ai/integrations/ixigo): ixigo is one of the connected apps Neotask's agent can operate directly. Search and compare domestic and international flights and hotels with live fares, schedules, and amenity filters, track flights in real time, and see fare trends...
- [JARS LT](https://neotask.ai/integrations/jars-lt): Instead of opening JARS LT yourself, ask Neotask. Search and retrieve official Lithuanian company registry records, including company details, and run macro-level analysis of the Lithuanian economy.
- [Jaz MCP Server](https://neotask.ai/integrations/jaz-mcp-server): Jaz MCP Server handles this without you switching tabs: create and manage invoices, bills, and reconciliations, and pull financial reports from your Jaz accounting workspace.
- [Jentic](https://neotask.ai/integrations/jentic): Search across more than 1,500 APIs, load the execution details for the one you need, and run the call, all without exposing credentials in the conversation since Jentic's vault manages authentication centrally.
- [Jerry.ai Car Insurance & Care](https://neotask.ai/integrations/jerry-ai-car-insurance-and-care): Get personalized comparisons of car insurance and repair options using data from repair shops and insurance providers. Useful for drivers deciding between insurance plans or looking for trustworthy repair pricing.
- [Jimdo](https://neotask.ai/integrations/jimdo): Instead of opening Jimdo yourself, ask Neotask. Turn a conversation about your business into a complete website, with content and design generated from what was already discussed, then publish it or move to Jimdo's growth tools for...
- [Jina Remote MCP Server](https://neotask.ai/integrations/jina-remote-mcp-server): Jina Remote MCP Server is one of the connected apps Neotask's agent can operate directly. Read and search the web, generate embeddings, rerank results, classify content, and run deep multi-step research over web pages and multimodal...
- [Jinko](https://neotask.ai/integrations/jinko): Ask for flight recommendations by budget, dates, or trip type and get tailored deals drawn from billions of daily airfares.
- [Jinko MCP server (documented endpoint)](https://neotask.ai/integrations/jinko-mcp-server-documented-endpoint): Discover, price, book, and manage flights and hotels through Jinko's shared trip model, available consistently across MCP, HTTP API, and CLI interfaces.
- [Jitty Property Search](https://neotask.ai/integrations/jitty-property-search): Property listings describe square footage and bedroom counts. They don't tell you how quiet the street is or whether there's a decent pub nearby.
- [Job Search by Jobtome](https://neotask.ai/integrations/job-search-by-jobtome): Job Search by Jobtome handles this without you switching tabs: search job listings by title, keyword, or city across multiple industries and job boards worldwide.
- [Jobkorea](https://neotask.ai/integrations/jobkorea): Search Korean job openings by district, company type, industry, required skills or licenses, education, and experience level.
- [Jobs by Wizehire](https://neotask.ai/integrations/jobs-by-wizehire): Job hunting usually means the same tab-juggling ritual: search a board, open a listing, retype your details into an application form, repeat. Jobs by Wizehire connected to Neotask cuts that loop down to a conversation.
- [Johns Hopkins Careers](https://neotask.ai/integrations/johns-hopkins-careers): Johns Hopkins runs one of the largest healthcare and research employers in the country, and its careers site reflects that scale — hundreds of open roles across clinical, research, and administrative tracks.
- [Jooble](https://neotask.ai/integrations/jooble): Jooble exists to solve a specific annoyance: job postings are scattered across thousands of individual boards and company sites, and checking each one is a part-time job of its own.
- [Jora](https://neotask.ai/integrations/jora): Jora indexes hundreds of thousands of listings from boards and company career pages, and it also does something most aggregators skip: match roles to an actual resume.
- [Jumpseller MCP Server](https://neotask.ai/integrations/jumpseller-mcp-server): Running an online store means constantly context-switching between the products page, the orders queue, customer records, and shipping settings — all inside the Jumpseller admin panel.
- [JustFix](https://neotask.ai/integrations/justfix): Getting a home repair quote normally means calling around, waiting for callbacks, and comparing bids yourself. JustFix connected to Neotask compresses that: describe the issue, and the agent gets an upfront cost estimate back, then books a...
- [Kactus](https://neotask.ai/integrations/kactus): Picking a venue for an offsite, seminar, or corporate gathering usually means comparing a pile of PDFs and half-answered emails.
- [Kaggle MCP Server](https://neotask.ai/integrations/kaggle-mcp-server): Kaggle holds a huge amount of what data scientists actually need day to day — datasets, notebooks, pretrained models, competition details — but finding the right one still means leaving your workflow to search the site.
- [Kelley Blue Book](https://neotask.ai/integrations/kelley-blue-book): Trading in a car or selling it yourself pays very differently, and the gap only shows up once you know both numbers. Kelley Blue Book connected to Neotask looks up both at once — Trade-In Value and Private Party Value — for a used car you...
- [Kernel MCP Server](https://neotask.ai/integrations/kernel-mcp-server): Building browser-automation agents at any real scale runs into the same wall: you need remote browsers and sandboxes you can spin up and tear down on demand, not a laptop tab.
- [Kestra MCP server](https://neotask.ai/integrations/kestra-mcp-server): Kestra workflows are defined as YAML flows with their own tasks, triggers, inputs, and outputs, and keeping tabs on them across self-hosted or cloud deployments usually means jumping into the Kestra UI.
- [Ketryx](https://neotask.ai/integrations/ketryx): Regulated software teams keep compliance data spread across Jira, Jama, GitHub, TestRail, Polarion, and Azure DevOps, and answering a single audit question can mean checking all of them.
- [KeyBid Puls](https://neotask.ai/integrations/keybid-puls): Evaluating a short-term rental property usually means building your own revenue model from scratch. KeyBid Puls connected to Neotask skips that: paste a listing URL or screenshot from Airbnb, Booking.com, or similar sites, and the agent...
- [Kickresume](https://neotask.ai/integrations/kickresume): Starting a resume from a blank page is its own kind of writer's block. Kickresume connected to Neotask removes the blank page: describe your work experience, skills, and career goals in plain conversation, and the agent builds a polished,...
- [Kindora Funder Discovery](https://neotask.ai/integrations/kindora-funder-discovery): Prospect research for nonprofit fundraising means digging through giving patterns and IRS 990 filings across a huge field of foundations — Kindora Funder Discovery covers more than 168,000 of them.
- [Kindora Pro MCP Server](https://neotask.ai/integrations/kindora-pro-mcp-server): Beyond just searching for funders, nonprofit development teams need to actually manage the pipeline of prospects they're pursuing.
- [Kit Developer Docs MCP](https://neotask.ai/integrations/kit-developer-docs-mcp): Building against Kit's email marketing and creator platform means digging through API references for subscribers, broadcasts, and tagging while you code.
- [Kit MCP](https://neotask.ai/integrations/kit-mcp): Running email campaigns in Kit means juggling subscribers, broadcasts, sequences, forms, and tags across the creator dashboard.
- [Klarna Shopping Search](https://neotask.ai/integrations/klarna-shopping-search): Klarna Shopping Search brings Klarna's product search capability directly into Neotask, so a shopping question can be answered by searching Klarna's catalog on the spot rather than switching to a separate shopping app or browser tab.
- [Kleinanzeigen](https://neotask.ai/integrations/kleinanzeigen): Kleinanzeigen is Germany's largest classifieds marketplace, spanning furniture, electronics, used cars, real estate, and job listings — a lot of ground to cover by hand.
- [Klook](https://neotask.ai/integrations/klook): Planning what to actually do on a trip is often the hardest part of travel planning. Klook connected to Neotask handles that step: describe a destination, your dates, and your interests, and the agent returns matching tours, attractions,...
- [Kokiko](https://neotask.ai/integrations/kokiko): Comparing beauty and personal care products in Moldova usually means checking prices site by site. Kokiko connected to Neotask searches its catalog by product name directly, returning prices in Moldovan lei alongside discounts, brand...
- [komoot](https://neotask.ai/integrations/komoot): Planning an outdoor route usually means cross-referencing a map with reviews from people who've actually done it. Komoot connected to Neotask surfaces cycling, hiking, and running routes along with the highlights that community members...
- [Kong Konnect MCP Server](https://neotask.ai/integrations/kong-konnect-mcp-server): Kong Konnect MCP Server connects Neotask directly to Kong Konnect, giving the agent a way to work with Kong Konnect's platform through the conversation instead of a separate console.
- [Košík.cz](https://neotask.ai/integrations/kosik-cz): A full grocery run in the Czech Republic normally means building a cart item by item on Kosik.cz, then handling delivery details separately.
- [Kraken](https://neotask.ai/integrations/kraken): Crypto markets move fast, and most of the useful context — why a coin jumped, whether a move is part of a longer trend, how it stacks up against the rest of the top assets — is scattered across charts, news feeds, and social sentiment...
- [Kredit](https://neotask.ai/integrations/kredit): Comparing UAE credit cards usually means opening a handful of bank sites, squinting at fee tables, and trying to remember which card had the better cashback on dining versus travel.
- [Krikey AI Animation](https://neotask.ai/integrations/krikey-ai-animation): Making a short animated video used to mean either learning animation software or hiring someone who already had. Krikey flips that: pick a 3D character, a music genre, and a topic, and its text-to-animation engine generates the clip for...
- [Krisp](https://neotask.ai/integrations/krisp): Krisp already sits in your meetings recording transcripts and notes — the problem is finding anything in that archive later.
- [Krónan](https://neotask.ai/integrations/kronan): Before a grocery run, it's useful to know whether the store actually has what you need without driving over to check. With Krónan connected, Neotask can search Krónan's product catalog directly, so you can confirm an item exists, check its...
- [Kubera](https://neotask.ai/integrations/kubera): Kubera already tracks your full financial picture, but pulling a quick answer out of it usually means opening the app and digging through accounts.
- [KYN AI Fitness Planner](https://neotask.ai/integrations/kyn-ai-fitness-planner): Booking studio classes usually means opening a separate app, scrolling the schedule, and hoping your preferred time isn't full.
- [La Pyme MCP](https://neotask.ai/integrations/la-pyme-mcp): Small business owners running their operations through La Pyme don't always have time to open the dashboard for a quick number.
- [Lambus](https://neotask.ai/integrations/lambus): Planning a multi-stop trip usually means juggling a spreadsheet for expenses, a separate note for points of interest, and a rough mental itinerary that lives in your head.
- [Laminar MCP Server](https://neotask.ai/integrations/laminar-mcp-server): When an AI agent misbehaves in production, the hard part is usually finding out why. Laminar traces agent and LLM application runs so you can inspect exactly where something went wrong.
- [Langfuse MCP Server](https://neotask.ai/integrations/langfuse-mcp-server): Once your app is running in production, the useful questions shift from "how does Langfuse work" to "what's actually happening in my traces." This connection queries your own Langfuse project — traces, prompts, datasets, and evaluation...
- [LangSmith Remote MCP](https://neotask.ai/integrations/langsmith-remote-mcp): If your agent or chain is built with LangChain, LangSmith is usually where the debugging trail lives. Connected to Neotask, you can search traces, evaluations, and prompt versions directly, so when a chain produces an unexpected result you...
- [Le Parisien](https://neotask.ai/integrations/le-parisien): Staying current on French and Paris-area news usually means checking the site directly or scrolling a news app. With Le Parisien connected, Neotask can read the latest headlines, search articles by topic or location, and follow live...
- [LeapScholar](https://neotask.ai/integrations/leapscholar): Researching study-abroad programs means comparing tuition, rankings, and intake dates across a long list of universities and countries — tedious to do by hand.
- [Learning Commons Knowledge Graph](https://neotask.ai/integrations/learning-commons-knowledge-graph): Building standards-aligned curriculum means knowing exactly what a state expects at each grade level, and how a given standard evolves as students progress.
- [leboncoin](https://neotask.ai/integrations/leboncoin): France's largest classifieds marketplace has an enormous catalog — electronics, furniture, fashion, real estate, vehicles, and everything in between.
- [LegalZoom](https://neotask.ai/integrations/legalzoom): Reviewing a contract or a small business legal document is easier when you have a quick read on what's actually risky before deciding whether to bring in a lawyer.
- [lemlist MCP server](https://neotask.ai/integrations/lemlist-mcp-server): Running cold outreach usually means checking campaign performance, lead status, and inbox activity across a separate tool from wherever else your team works.
- [lesfurets](https://neotask.ai/integrations/lesfurets): Wondering if your car insurance is still competitively priced usually means going through a comparison site's full form. With lesfurets connected, you can get a free instant estimate by giving Neotask your age, vehicle, and coverage level,...
- [LFmall](https://neotask.ai/integrations/lfmall): Scrolling through thousands of premium fashion items to find something for a specific occasion gets old fast. With LFmall connected, Neotask can get product recommendations matched to your style and the occasion you describe, drawing on...
- [liblab hosted MCP server](https://neotask.ai/integrations/liblab-hosted-mcp-server): Turning an API specification into an SDK, documentation, or an MCP server is exactly the kind of build-pipeline work teams don't want to maintain themselves.
- [LiftTrack](https://neotask.ai/integrations/lifttrack): Your Garmin-synced strength training data is already there, but reviewing progress usually means opening the app and scrolling through session after session.
- [Linkly URL Shortener](https://neotask.ai/integrations/linkly-url-shortener): Marketing links get messy fast — a dozen campaigns, a dozen UTM variants, and no easy way to see what's actually getting clicked.
- [Linktree](https://neotask.ai/integrations/linktree): Finding the right creators for an influencer campaign usually means scrolling through profiles by hand. Neotask's connection to Linktree does that filtering for you — describe the campaign and the kind of creator you're after, and it...
- [ListingForgeAI](https://neotask.ai/integrations/listingforgeai): Building a product listing from scratch — title, description, specs, images — is repetitive work that ListingForgeAI automates from just a barcode or product name.
- [Little Caesars](https://neotask.ai/integrations/little-caesars): Deciding what to order for a group is more about the back-and-forth than the ordering itself. Neotask's connection to Little Caesars Concierge handles that conversation — finding a nearby store, pulling up the menu, and working out...
- [LlamaIndex documentation MCP server](https://neotask.ai/integrations/llamaindex-documentation-mcp-server): LlamaIndex's docs cover a lot of ground — parsing, extraction, classification, indexing pipelines, LlamaParse, LlamaCloud.
- [LobeHub WebMCP](https://neotask.ai/integrations/lobehub-webmcp): LobeHub's marketplace of agents, MCP servers, and skills is worth browsing when you're assembling a workflow, but scrolling a catalog is slower than just asking for what you need.
- [Local Falcon](https://neotask.ai/integrations/local-falcon): Local SEO performance lives across geo-grid scans, competitor comparisons, and review sentiment — usually scattered across separate reports.
- [Localiza](https://neotask.ai/integrations/localiza): Choosing between renting, buying, or subscribing to a car involves comparing options that aren't always laid out side by side.
- [LogRocket MCP Server](https://neotask.ai/integrations/logrocket-mcp-server): Debugging a user-reported bug usually means digging through session replays and error logs by hand. Neotask connects to LogRocket so you can ask for the session replay, captured error, or performance signal directly and get it pulled from...
- [Logto MCP Server](https://neotask.ai/integrations/logto-mcp-server): Managing identity and access — users, roles, tenants, auth configuration — usually means working through an admin console one screen at a time.
- [LONA Trading Assistant](https://neotask.ai/integrations/lona-trading-assistant): Testing a trading idea properly means writing the strategy, feeding it real market data, and backtesting it before risking anything live.
- [Lorikeet](https://neotask.ai/integrations/lorikeet): Building and tuning an AI support agent means juggling workflows, tools, integrations, guardrails, and then checking how it actually performs against real conversations.
- [Lovable](https://neotask.ai/integrations/lovable): Turning an app idea into a working project usually means either learning to code it or hiring someone who can. Neotask connects to Lovable so you can describe the idea in plain language and have it build and iterate on the actual project...
- [loveholidays](https://neotask.ai/integrations/loveholidays): Planning a package holiday means weighing dates, budget, destination, and what the hotel is actually like once real guests have stayed there.
- [Lowe’s](https://neotask.ai/integrations/lowe-s): Sourcing or comparing home improvement products across tools, appliances, and building materials means browsing a large catalog.
- [Lucid](https://neotask.ai/integrations/lucid): Making a diagram from scratch, or finding one you made months ago, both eat time you'd rather spend on the actual work. Neotask connects to Lucid so you can generate a diagram from a description or pull up an existing Lucidchart document...
- [Lucid Developer Documentation MCP server](https://neotask.ai/integrations/lucid-developer-documentation-mcp-server): Building against Lucid's REST API or writing a custom editor extension means knowing the right endpoint and authentication pattern.
- [Lugg](https://neotask.ai/integrations/lugg): Getting something moved — furniture, a marketplace pickup, a storage run, a junk haul — usually starts with wondering what it'll cost and whether it's even worth booking a mover.
- [Lusha](https://neotask.ai/integrations/lusha): Building a lead list usually means running searches across a B2B database and manually enriching each contact. Neotask connects to Lusha so you can search for a specific person's contact details or filter companies by industry or size...
- [Luxauto](https://neotask.ai/integrations/luxauto): Shopping for a car or motorcycle in Luxembourg and the Greater Region across multiple marketplace listings takes time to narrow down.
- [Luxury Escapes](https://neotask.ai/integrations/luxury-escapes): Finding a hotel that matches a destination, dates, and a particular vibe usually means opening several booking sites and comparing manually.
- [Lytical](https://neotask.ai/integrations/lytical): Raw analytics dashboards tell you numbers changed but rarely tell you why. Neotask connects to Lytical so you can ask what happened on your site in plain language and get an explanation along with suggested next steps, instead of digging...
- [M6plus](https://neotask.ai/integrations/m6plus): Scrolling a streaming catalog to find something worth watching is its own chore. Neotask connects to M6+ so you can describe the mood you're in and get tailored recommendations back instead of browsing categories yourself.
- [Maersk](https://neotask.ai/integrations/maersk): Tracking an ocean shipment usually means logging into a separate carrier tracking portal to check on a booking or container.
- [Magneto Empleos](https://neotask.ai/integrations/magneto-empleos): Job hunting usually means juggling filters across two or three job boards before you find anything worth a second look. Magneto Empleos connects Neotask directly to Magneto365's job listings, so you can describe the role you actually want...
- [Mailchimp Transactional Messaging MCP](https://neotask.ai/integrations/mailchimp-transactional-messaging-mcp): Mandrill handles the transactional email most apps depend on — password resets, receipts, shipping notices — while Mailchimp itself holds the marketing side: lists, campaigns, subscriber data.
- [MailerLite](https://neotask.ai/integrations/mailerlite): MailerLite runs a lot of the mechanics behind an email marketing program — subscriber lists, segments, scheduled campaigns, automations, forms — and normally that means logging into MailerLite itself every time something needs adjusting.
- [MailerSend MCP Server](https://neotask.ai/integrations/mailersend-mcp-server): MailerSend covers both transactional email and SMS, plus the account-level plumbing around them — domains, templates, inbound routing, suppression lists.
- [MainFunc MCP server](https://neotask.ai/integrations/mainfunc-mcp-server): MainFunc MCP server is a general connector: it wires Neotask's agent into MainFunc's own MCP server so whatever tools MainFunc exposes there become available inside a normal conversation instead of a separate integration.
- [Make A Viz](https://neotask.ai/integrations/make-a-viz): Explaining a process out loud is easy; drawing it in a diagramming tool is the slow part. Make A Viz closes that gap — describe a sales approval flow, an auth sequence, or a decision tree in plain language and it builds the flowchart,...
- [Malwarebytes](https://neotask.ai/integrations/malwarebytes): Before you click a link, call a number back, or reply to an email that feels slightly off, Malwarebytes can check it against real threat intelligence.
- [MangaBoom](https://neotask.ai/integrations/mangaboom): A chat log, a story idea, or a rough concept can become an actual comic through MangaBoom — full color manga-style panels, including quick 4-panel skits pulled straight from a conversation.
- [Manus](https://neotask.ai/integrations/manus): Some requests are bigger than a single reply — a full presentation deck, a working website, a marketing video, a research report with real sources behind it.
- [Map Maker](https://neotask.ai/integrations/map-maker): Trip planning tends to live scattered across a travel PDF, a saved TikTok, and a mental list of places to hit. Map Maker pulls all of that into one place — attach a travel PDF or a TikTok link and it extracts the locations automatically,...
- [Mapify](https://neotask.ai/integrations/mapify): Dense topics get easier to hold in your head once they're laid out spatially. Mapify takes a topic — or a block of content you've already organized — and turns it into an interactive, zoomable mind map, useful for organizing ideas,...
- [MarcoPolo](https://neotask.ai/integrations/marcopolo): Real data work usually means bouncing between a database console, an API, a CRM, and a shell — MarcoPolo puts all of that behind one secure, scoped-credential workspace instead.
- [MarketCheck MCP Server](https://neotask.ai/integrations/marketcheck-mcp-server): Whether you're pricing a trade-in or building something that needs real automotive data, MarketCheck gives Neotask direct access to vehicle listings, VIN decoding, and market analytics across North America and the UK — so a valuation or...
- [Marrow MCP server](https://neotask.ai/integrations/marrow-mcp-server): Marrow MCP server is a general connector into Marrow — it links Neotask's agent to Marrow's own MCP server so whatever tools Marrow exposes there are reachable from a normal conversation instead of a separate app.
- [Massive Remote MCP Server](https://neotask.ai/integrations/massive-remote-mcp-server): Financial market data spans a lot of asset classes and a lot of ways to consume it — Massive covers both. Stocks, options, futures, indices, forex, and crypto are all reachable through on-demand queries, streaming feeds, or bulk historical...
- [McAfee](https://neotask.ai/integrations/mcafee): McAfee's connection turns a suspicious email, text, or link into a quick answer instead of a guess. Share what you received and it gives you a real-time read on whether it's a scam, along with a plain explanation of the risk and what to do...
- [MCP server](https://neotask.ai/integrations/mcp-server): This connection links Neotask's agent to BigDataCloud's MCP server, bringing whatever tools BigDataCloud exposes there into a normal conversation instead of a separate integration to manage.
- [MCPTotal](https://neotask.ai/integrations/mcptotal-mcp-server): MCPTotal's own MCP server becomes reachable from Neotask once connected, giving the agent a direct line to whatever MCPTotal exposes without you needing to interact with its interface separately.
- [MDN](https://neotask.ai/integrations/mdn-mcp-server): MDN (the Mozilla Developer Network) is the reference most web developers already trust for HTML, CSS, and JavaScript documentation.
- [Medeloop Grants](https://neotask.ai/integrations/medeloop-grants): Medeloop Grants is aimed at researchers hunting for funding. Connected to Neotask, it matches grant opportunities to your research background and focus, checks eligibility and deadlines against that opportunity, and can draft a first-pass...
- [Medidata](https://neotask.ai/integrations/medidata): Medidata covers clinical trial platforms like Rave EDC, Data Connect, and Clinical Data Studio. Connected to Neotask, you can ask how to use those products and get answers pulled from Medidata's own platform documentation — and,...
- [MeetGeek](https://neotask.ai/integrations/meetgeek-cloud-mcp): MeetGeek records and transcribes meetings, and this connection lets Neotask reach into that captured history directly. Ask it to retrieve a recording, a transcript, or a summary of a past meeting, or to pull insights for sharing — and it...
- [Meetup](https://neotask.ai/integrations/meetup): Meetup connects Neotask to local event discovery. Describe what you're looking for in plain language — a topic, a type of gathering, a date range — and it finds matching events, lets you filter by location and dates, and surfaces who's...
- [Melon](https://neotask.ai/integrations/melon): Melon is a music platform, and connected to Neotask you can browse charts, look up detailed artist/album/track information, and get personalized song recommendations pulled from your own listening history — plus search for songs and...
- [Mem](https://neotask.ai/integrations/mem): Mem is a notes and knowledge-base tool, and this connection lets Neotask search your existing notes for context, capture new ideas as they come up, save whole chat conversations into notes, and edit documents you already have — so your...
- [Mem0](https://neotask.ai/integrations/mem0-mcp-server): Mem0 is a memory layer built for AI agents and LLM applications. Connected to Neotask, you can store, search, update, and manage persistent memories across different users and agents — useful if you're building or operating something that...
- [MeranO](https://neotask.ai/integrations/merano): MeranO is a local guide for Merano, South Tyrol, Italy — connected to Neotask, it surfaces up-to-date information on events, hotels, bike tours, hiking trails, and dining in the area.
- [Merge](https://neotask.ai/integrations/merge-agent-handler-mcp-server): Merge normalizes access to a huge range of third-party business systems — HR, ATS, accounting tools — behind unified APIs, using linked accounts per end customer.
- [Mermaid Chart](https://neotask.ai/integrations/mermaid-chart): This connection is specifically for validating and rendering Mermaid diagrams. Ask Neotask to check your Mermaid syntax, and it validates it, renders the diagram as a high-quality SVG, and displays it in an interactive viewer where you can...
- [Mermaid Chart (diagram generation)](https://neotask.ai/integrations/mermaid-mcp-server): Where the Mermaid Chart connection above focuses on validating and rendering diagrams you already wrote, this Mermaid MCP Server connection is about creating them — AI-assisted diagram generation from a description, plus browsing the open...
- [Metaview](https://neotask.ai/integrations/metaview): Metaview specializes in recruiting conversation intelligence. Connected to Neotask, it lets you search and analyze your recruiting conversations, group and query candidate data, and extract insights from hiring interactions — plus source...
- [Meter Energy](https://neotask.ai/integrations/meter-energy-mcp-server): Meter Energy is focused on Texas electricity and solar buyback plan comparison for homeowners. Connected to Neotask, you can ask it to compare plans and pull the energy data behind them, so finding the best electricity rate — or the best...
- [MGS Agile Toolkit](https://neotask.ai/integrations/mgs-agile-toolkit): Scrum masters and product owners spend a surprising amount of their week on things a good coach could do in minutes: splitting a vague epic into shippable stories, sanity-checking a backlog item against INVEST, or digging up the right...
- [Middesk MCP server](https://neotask.ai/integrations/middesk-mcp-server): Before you extend credit, sign a vendor, or onboard a business customer, you want a straight answer to one question: is this a real, legitimately registered company, or is something off? Middesk MCP server gives Neotask's agent direct...
- [Midify](https://neotask.ai/integrations/midify): Most people with a melody stuck in their head have no idea what to do with it next — they don't read music, don't know a ii-V-I from a power chord, and definitely don't want to open a DAW to find out.
- [Midpage Legal Research](https://neotask.ai/integrations/midpage-legal-research): Midpage lets you search a database of case law, review court opinions, and find specific passages within an opinion, with everything linked to real sources for verification.
- [Migma MCP Server](https://neotask.ai/integrations/migma-mcp-server): Running email marketing well means juggling a dozen small operational tasks — building a template, checking it renders correctly, keeping audiences segmented, watching which campaigns are actually landing, and making sure your sending...
- [Migros](https://neotask.ai/integrations/migros): Grocery runs are the kind of task that's easy to describe and tedious to execute — you know you need milk, bread, and whatever's on sale this week, but browsing a storefront and building a cart takes real time.
- [Milanuncios](https://neotask.ai/integrations/milanuncios): Milanuncios lets you search Spain's classifieds marketplace for second hand goods, vehicles, real estate, and professional services, filtering by category, location, and price.
- [MineSkin MCP server](https://neotask.ai/integrations/mineskin-mcp-server): MineSkin MCP server connects Neotask's agent directly to the MineSkin service, so requests that would normally mean visiting the site and working through its interface can instead be handled from a conversation.
- [Mintlify Admin MCP](https://neotask.ai/integrations/mintlify-admin-mcp): Search Mintlify-hosted documentation content, and for trusted/admin users, edit pages, manage docs structure, and review site analytics. Requests go through natural language rather than the dashboard UI.
- [MintMCP Gateway MCP endpoint](https://neotask.ai/integrations/mintmcp-gateway-mcp-endpoint): MintMCP Gateway MCP endpoint is a gateway endpoint that lets Neotask's agent route requests through the MintMCP Gateway to whatever MCP-connected services sit behind it.
- [Minty](https://neotask.ai/integrations/minty): Search for coupon codes, cashback rates, and promotional deals at partner stores by name, and get the tracking link needed to claim the cashback. Does not process payments or sell anything directly.
- [Miracle Of Mind](https://neotask.ai/integrations/miracle-of-mind): Start a short guided meditation, breathing exercise, or grounding practice suited to how you're feeling, such as stress, anxiety, overthinking, or trouble sleeping, drawing on Sadhguru's yogic teachings.
- [Mixedbread MCP Server](https://neotask.ai/integrations/mixedbread-mcp-server): Ingest files into searchable stores, then run semantic search and question answering over that content to turn unstructured documents into a queryable knowledge base for an AI agent.
- [Mixmax](https://neotask.ai/integrations/mixmax): Review sales outreach sequences, prospect engagement, and AI-generated meeting notes from a connected Mixmax account, and pull insights on which campaigns and messages are performing.
- [Model Context Protocol docs MCP server](https://neotask.ai/integrations/model-context-protocol-docs-mcp-server): Search and read the Model Context Protocol specification, implementation guides, and examples for building tools, servers, and clients that connect AI applications to external resources.
- [Model Context Protocol docs MCP server (Mintlify host)](https://neotask.ai/integrations/model-context-protocol-docs-mcp-server-mintlify-host): Search and read the Model Context Protocol specification, implementation guides, and examples through the Mintlify-hosted mirror of the MCP documentation site.
- [MODO](https://neotask.ai/integrations/modo): Discover active discounts and cashback benefits at supermarkets, gas stations, and restaurants tied to your Argentine bank cards, filtered by category or day of the week.
- [Mollie MCP Server](https://neotask.ai/integrations/mollie-mcp-server): Create and manage payments, payment links, and recurring billing, and pull transaction and settlement reports through Mollie's European payment processing platform for merchants and marketplaces.
- [Mondoir.art](https://neotask.ai/integrations/mondoir-art): Explore contemporary artworks, artists, and gallery collections, and get context on valuation, provenance, and editions for pieces you're researching or considering collecting.
- [MoneySuperMarket](https://neotask.ai/integrations/moneysupermarket): Compare car, home, pet, and van insurance, credit cards, loans, current accounts, savings products, and broadband deals, and sign in to a MoneySuperMarket account for personalized comparisons and a broadband speed check.
- [Monito](https://neotask.ai/integrations/monito): Compare international money transfer providers such as Wise, Revolut, Western Union, and Remitly on exchange rate, fees, and delivery speed to find the cheapest way to send money abroad.
- [Monoscope MCP Server](https://neotask.ai/integrations/monoscope-mcp-server): Query API request logs, traces, performance metrics, and anomaly alerts from a Monoscope (formerly APItoolkit) observability project to debug and monitor API traffic.
- [MoodTrip.Ai](https://neotask.ai/integrations/moodtrip-ai): Describe your ideal hotel stay in plain language, such as destination, dates, budget, or style, and get real-time pricing, availability, photos, and direct booking links across thousands of properties.
- [MoSPI](https://neotask.ai/integrations/mospi): Query official Indian government statistics on GDP, inflation, employment, industrial production, education, health, energy, and trade sourced directly from the Ministry of Statistics and Programme Implementation.
- [Motion Creative Analytics](https://neotask.ai/integrations/motion-creative-analytics): Analyze Meta ad creative performance across campaigns, surface top and declining performers, research competitor ad libraries, and view demographic breakdowns, creative transcripts, and performance reports.
- [MotionHub MCP](https://neotask.ai/integrations/motionhub-mcp): Media production runs on scattered pieces — a shot list in one place, a shooting calendar in another, footage transcripts nobody wants to scrub through, and a contact sheet for the crew.
- [motos.net](https://neotask.ai/integrations/motos-net): Shopping for a motorbike usually means bouncing between listing sites, opening ten tabs, and manually lining up specs in a spreadsheet.
- [Motra](https://neotask.ai/integrations/motra): Motra tracks the sets, reps, and lifts you've logged — but pulling insight out of a workout log usually means manually flipping through history.
- [Movie Pick](https://neotask.ai/integrations/movie-pick): Movie Pick already tracks what you've watched, rated, and want to watch — Neotask just makes it conversational. Ask for something to watch tonight and get recommendations built from your actual ratings, log a film you just finished, add...
- [MSCI](https://neotask.ai/integrations/msci): MSCI holds the index data institutional investors rely on for exposure and factor analysis — but digging through it usually means navigating dense research portals.
- [MutuiOnline.it](https://neotask.ai/integrations/mutuionline-it): Comparing mortgage offers across Italian banks usually means visiting each lender's site separately. MutuiOnline.it connected to Neotask flips that around — describe your loan needs and the agent pulls offers from dozens of Italian banks...
- [MyChoice Insurance](https://neotask.ai/integrations/mychoice-insurance): Getting an insurance quote usually means filling out the same form on multiple carrier sites. MyChoice Insurance connected to Neotask lets you answer a handful of questions once, and the agent returns auto and life insurance quotes from...
- [MYCOlive](https://neotask.ai/integrations/mycolive): MYCOlive connects to Neotask so agents can reach MYCOlive directly from a conversation rather than through a separate login.
- [mydeetz](https://neotask.ai/integrations/mydeetz): Filling out contact forms for every business you deal with gets old fast. mydeetz connected to Neotask flips the process: ask what information a business actually requires, then share your name, email, and other details straight from the...
- [MyFitnessPal](https://neotask.ai/integrations/myfitnesspal): MyFitnessPal connected to Neotask turns meal planning into a conversation. Describe your dietary goals — cutting, bulking, plant-based, whatever it is — and the agent generates personalized meal plans and recipes, plus nutrition guidance...
- [MyFonts](https://neotask.ai/integrations/myfonts): Picking the right font for a project is half taste, half catalog-browsing fatigue. MyFonts connected to Neotask cuts the browsing out — describe your project's mood, style, or industry, and the agent returns curated font recommendations...
- [MyRegistry.com](https://neotask.ai/integrations/myregistry-com): Wedding and baby-shower registries usually mean picking items from five different stores and hoping guests find the right list.
- [Naked Insurance](https://neotask.ai/integrations/naked-insurance): Naked Insurance connected to Neotask brings car insurance quoting into the conversation. Describe your vehicle and coverage needs, get a quote and compare cover options directly in chat, then continue to Naked's site when you're ready to...
- [Namecheap](https://neotask.ai/integrations/namecheap): Naming a new project usually means checking the same domain across a dozen TLDs by hand. Namecheap connected to Neotask does that lookup for you — check domain availability and pricing across multiple TLDs, get alternative name suggestions...
- [Natoma managed MCP profile/connection endpoints](https://neotask.ai/integrations/natoma-managed-mcp-profile-connection-endpoints): Organizations running Natoma set up governed connection profiles that determine exactly which internal business applications an AI agent is allowed to reach and how.
- [Natural MCP Server](https://neotask.ai/integrations/natural-mcp-server): Natural is a payments platform built specifically for AI agents and apps to move money, and connecting it to Neotask lets the agent send, receive, and manage funds through wallets and transfers directly from a conversation.
- [NatWest Mortgages](https://neotask.ai/integrations/natwest-mortgages): Checking what you could borrow on a mortgage usually means logging into online banking or filling out a form that pulls your credit.
- [Neartail](https://neotask.ai/integrations/neartail): Neartail turns a plain menu or product list into a working order system — describe what you sell and it builds an order form with automatic pricing and payment already wired in, with every order syncing to Google Sheets for kitchen prep...
- [Neighbor](https://neotask.ai/integrations/neighbor): Finding storage or parking space nearby usually means checking several listing sites and reading through reviews one by one.
- [Neptune Flood](https://neotask.ai/integrations/neptune-flood): Flood insurance quotes usually require navigating a carrier's own quoting tool property by property. Neptune Flood connected to Neotask gets you a real-time flood insurance quote for a specific US property, including pricing details,...
- [Nesha](https://neotask.ai/integrations/nesha): Finding the right strategic partner — for co-sell, channel alliances, tech partnerships, or licensing — usually means a lot of manual research and cold outreach.
- [Netatmo MCP server](https://neotask.ai/integrations/netatmo-mcp-server): This is the Netatmo MCP server connection — the link that lets Neotask's agent talk to Netatmo directly. Once connected, requests involving Netatmo's smart home and weather devices can be handled through chat instead of Netatmo's own app.
- [Network Management MCP server](https://neotask.ai/integrations/network-management-mcp-server): Diagnosing a Google Cloud network issue usually means jumping into Network Intelligence Center and clicking through connectivity test results manually.
- [Network Solutions](https://neotask.ai/integrations/network-solutions): Network Solutions connected to Neotask makes domain shopping start from an idea instead of a name. Describe a business idea in plain language, and the agent instantly checks domain name availability across relevant extensions, surfacing...
- [Neurnav](https://neotask.ai/integrations/neurnav): Long research reports fall apart the moment more than one person or one draft is touching them at once. Neurnav treats a report like a codebase: it lives on branches, every edit is a revision you can diff against the last one, and nothing...
- [Newegg](https://neotask.ai/integrations/newegg): Instead of digging through Newegg's filters yourself to find the right GPU or motherboard, tell Neotask what you're building and it searches Newegg's catalog directly.
- [Newsify](https://neotask.ai/integrations/newsify): If you want to know what's happening in a country or on a topic without opening a browser and scrolling past a dozen articles, Newsify gives Neotask a shortcut.
- [NEWT（ニュート）](https://neotask.ai/integrations/newt): Planning an overseas trip through NEWT usually means browsing tour packages one by one. Tell Neotask where you want to go, roughly when, and what you're willing to spend, and it searches NEWT's available overseas tour packages in Japan for...
- [Nexafin](https://neotask.ai/integrations/nexafin): Nexafin connects to your bank accounts in a read-only way, so Neotask can answer questions about your finances without ever being able to move money.
- [Nexafin MCP server (nexafin.com)](https://neotask.ai/integrations/nexafin-mcp-server-nexafin-com): This is Nexafin's personal finance API surface, letting Neotask query balances, transactions, and recurring bills either from connected bank accounts or from statements you've imported.
- [Noodle Seed](https://neotask.ai/integrations/noodle-seed): When your team's documents and notes live in a Noodle Seed workspace, finding the right one usually means opening the app and searching manually.
- [Norton](https://neotask.ai/integrations/norton): Suspicious email, a text that feels off, a link you're not sure about — instead of guessing, hand it to Neotask and it runs the analysis through Norton Genie AI.
- [NoteItHub](https://neotask.ai/integrations/noteithub): A lot of good intentions from conversations never turn into action because nobody wrote them down as a task. NoteItHub lets Neotask turn a conversation directly into a todo list organized by goal, and save the important parts of a chat as...
- [Notra MCP Server](https://neotask.ai/integrations/notra-mcp-server): Shipping code is only half the job — someone still has to tell people what changed. Notra connects Neotask to your GitHub activity so it can turn engineering and product work into changelogs, blog posts, and social posts written in your...
- [Novu MCP Server](https://neotask.ai/integrations/novu-mcp-server): Notifications that need to go out across email, SMS, push, and in-app all at once usually mean juggling several services. Novu gives Neotask one place to send and manage multi-channel notifications through workflows and subscriber lists,...
- [NPI Registry](https://neotask.ai/integrations/npi-registry): Verifying a US healthcare provider's identity usually means checking their National Provider Identifier against the official registry by hand.
- [NutriMindz](https://neotask.ai/integrations/nutrimindz): Instead of sifting through conflicting nutrition advice online, ask Neotask directly and it draws on NutriMindz for coaching-style answers.
- [Nuwa](https://neotask.ai/integrations/nuwa): If you take notes by hand with a Nuwa Pen, your handwriting is already digitized somewhere — Nuwa just makes it searchable.
- [Nuxt MCP Server](https://neotask.ai/integrations/nuxt-mcp-server): When you're building a full-stack Vue app with Nuxt and need to check how a feature is supposed to work, this connection lets Neotask search Nuxt.js documentation, deployment guides, and blog content directly.
- [Nuxt UI MCP Server](https://neotask.ai/integrations/nuxt-ui-mcp-server): Building a UI with Nuxt UI means looking up which component to use and how to wire it together with Tailwind. This connection lets Neotask search Nuxt UI's component library documentation directly, so when you ask how to use a specific...
- [NyquistAI](https://neotask.ai/integrations/nyquistai): Regulatory and compliance research around medical devices means digging through FDA databases that aren't built for quick lookups.
- [OK幹事](https://neotask.ai/integrations/ok): Organizing a group event usually starts with someone manually building a poll to find a date everyone can make. OK幹事 lets Neotask create that attendance poll for you — describe the event's purpose, date range, and time, and it generates a...
- [OKX](https://neotask.ai/integrations/okx): Checking crypto prices across multiple tokens usually means bouncing between exchange tabs and chart apps. Ask Neotask instead, and it pulls real-time prices, market stats, and trading volume straight from OKX, including interactive...
- [OLS4 MCP Server](https://neotask.ai/integrations/ols4-mcp-server): Life-science research often needs a specific ontology term, sequence record, or structural data point that lives in EMBL-EBI's public bioinformatics services.
- [Omio](https://neotask.ai/integrations/omio): Comparing trains, buses, flights, and ferries across dozens of operators and countries is exactly the kind of tedious research Neotask can take off your hands.
- [Omni Analytics](https://neotask.ai/integrations/omni-analytics): Getting an answer from your company's data usually means either writing SQL or hunting through a dashboard someone else built.
- [Omnisend](https://neotask.ai/integrations/omnisend): Instead of logging into Omnisend to check how a campaign performed, ask Neotask directly. It connects to your Omnisend account to answer questions about email and SMS campaign performance, where automations are losing subscribers along the...
- [OneUptime MCP Server](https://neotask.ai/integrations/oneuptime-mcp-server): Uptime dashboards are only useful if someone is actually watching them. Connect OneUptime and Neotask's agent becomes that someone: it can check whether your services are up, pull the current alert queue, look at who is on-call right now,...
- [OpenBudget MCP server](https://neotask.ai/integrations/openbudget-mcp-server): OpenBudget links your bank accounts through Plaid and gives Neotask's agent read-only access to balances and transaction history, so budgeting questions get answered directly from your real accounts instead of a spreadsheet you have to...
- [OpenCapital MCP server](https://neotask.ai/integrations/opencapital-mcp-server): When you need a number from a public company's actual filings — revenue by segment, a quarterly margin, a specific line item — OpenCapital gives Neotask's agent a direct path to SEC-derived fundamentals and metrics, with a link back to the...
- [OpenLens MCP server](https://neotask.ai/integrations/openlens-mcp-server): Being mentioned by ChatGPT or another AI assistant is becoming its own kind of visibility — and OpenLens tracks it. Connected to Neotask, it lets the agent check how your brand, your competitors, and specific topics get mentioned across...
- [OpenObserve MCP server](https://neotask.ai/integrations/openobserve-mcp-server): If your observability stack runs on OpenObserve, Neotask's agent can query it directly — logs, metrics, traces, and real user monitoring data — using SQL or PromQL under the hood while you just describe what you're looking for.
- [OpenRouter MCP server](https://neotask.ai/integrations/openrouter-mcp-server): OpenRouter sits in front of dozens of language and multimodal models with one API and automatic fallback if a provider is down.
- [OpenStatus MCP Server](https://neotask.ai/integrations/openstatus-mcp-server): OpenStatus runs synthetic uptime checks from multiple regions and manages your branded status pages. With it connected, Neotask's agent can run those checks, review results by region, and publish incident or maintenance updates directly to...
- [OptionsCalc](https://neotask.ai/integrations/optionscalc): Structuring a multi-leg options trade by hand — checking the Greeks, the implied volatility, the profit-and-loss curve across strike prices — is tedious even for a simple iron condor.
- [Organization Policy API MCP server](https://neotask.ai/integrations/organization-policy-api-mcp-server): Google Cloud's Organization Policy API is how you constrain which resource configurations are allowed across organizations, folders, and projects — and it's easy to lose track of what's actually enforced where.
- [OriginalVoices](https://neotask.ai/integrations/originalvoices): Before you ship messaging, a landing page headline, or a product concept, OriginalVoices lets you test it against a network of AI-represented real people rather than guessing how it'll land.
- [Orion by Gravity](https://neotask.ai/integrations/orion-by-gravity): Orion by Gravity connects your business metrics and workflows so questions about performance don't require rebuilding an analysis from scratch each time.
- [Ory Docs MCP Server](https://neotask.ai/integrations/ory-docs-mcp-server): Ory's documentation covers authentication, OAuth2/OIDC, and permissions across both cloud and self-hosted deployments — a lot of ground when you just need one answer.
- [Otomoto](https://neotask.ai/integrations/otomoto): Otomoto is Poland's big car marketplace, and searching it usually means filling out filter forms and clicking through listings one by one.
- [OTseek](https://neotask.ai/integrations/otseek): OTseek connects Neotask's agent to the OTseek platform so its capabilities are available directly from chat instead of a separate app.
- [Otter.ai](https://neotask.ai/integrations/otter-ai): Meeting notes are useless if you can't find the one line you need six weeks later. Connect Otter.ai to Neotask and the agent can search your transcripts by date, participant, or topic, then pull back summaries, action items, or the full...
- [Overstappen.nl](https://neotask.ai/integrations/overstappen-nl): Switching energy providers in the Netherlands means comparing rates, green-energy options, and contract terms across suppliers — exactly what Overstappen.nl is built for.
- [Ovios Home](https://neotask.ai/integrations/ovios-home): Furnishing an outdoor space usually starts with a lot of browsing before anything looks right together. Ovios Home connected to Neotask flips that: describe your space, your style, and your budget, and the agent pulls personalized...
- [OVIOS store MCP server](https://neotask.ai/integrations/ovios-store-mcp-server): Ovios also runs a Shopify-based storefront for indoor and outdoor furniture, and this connection lets Neotask's agent browse that catalog and complete purchases through agent-facing commerce flows — meaning you can go from "I need a...
- [Owkin](https://neotask.ai/integrations/owkin): Owkin works at the intersection of pathology and drug discovery — analyzing H&E histopathology slides from the TCGA database to quantify cell types, study how tumors interact with surrounding tissue spatially, and run survival analysis...
- [Paardplaats Horse Finder](https://neotask.ai/integrations/paardplaats-horse-finder): Paardplaats.nl is the Dutch equestrian marketplace for horses, and searching it usually means paging through listings that don't quite match what you actually need.
- [Paddle docs MCP server](https://neotask.ai/integrations/paddle-docs-mcp-server): When you're integrating Paddle for billing and hit a question about subscriptions, checkout, or invoicing, this connection lets Neotask's agent search Paddle's actual documentation and API reference directly, so you get the documented...
- [Paddle MCP server (live)](https://neotask.ai/integrations/paddle-mcp-server-live): This connection puts Neotask's agent directly into your live Paddle account — managing subscriptions, pricing, checkout configuration, invoicing, and customer billing data in production.
- [Paddle MCP server (sandbox)](https://neotask.ai/integrations/paddle-mcp-server-sandbox): Before a subscription or checkout change touches real customers, this connection lets Neotask's agent test it against Paddle's sandbox environment — subscriptions, checkout flows, invoicing, and billing workflows can all be run safely...
- [Paloma](https://neotask.ai/integrations/paloma): Finding a venue for a party or a large group dinner usually means calling around to check availability that turns out to be wrong.
- [PandaDoc MCP Server](https://neotask.ai/integrations/pandadoc-mcp-server): Chasing down whether a contract was signed, or finding the right template for a new proposal, shouldn't require opening PandaDoc every time.
- [Pane](https://neotask.ai/integrations/pane): Pane links to your bank accounts, credit cards, and investments and gives Neotask's agent a read-only window into your finances.
- [Parqet](https://neotask.ai/integrations/parqet): Parqet tracks your investment portfolios, and connecting it to Neotask means you can ask about your holdings the way you'd ask a person, not a dashboard.
- [Particl Market Research](https://neotask.ai/integrations/particl-market-research): Particl Market Research turns Neotask into an ecommerce research assistant. Instead of manually crawling competitor storefronts and stitching together spreadsheets, you can ask the agent to discover companies and product catalogs, compare...
- [Payabli Documentation MCP server](https://neotask.ai/integrations/payabli-documentation-mcp-server): When you're integrating Payabli's embedded payments platform and need an answer fast, this connection lets Neotask search Payabli's own documentation instead of you tabbing over to their docs site.
- [PAYONE MCP server](https://neotask.ai/integrations/payone-mcp-server): PAYONE is a payment service provider, and this connection gives Neotask's agent a direct line into PAYONE's own MCP server rather than routing requests through a generic integration layer.
- [PC Express](https://neotask.ai/integrations/pc-express): PC Express covers grocery ordering from Canadian retailers like Loblaws, No Frills, and Real Canadian Superstore, and connecting it means you can build your grocery order right from chat.
- [PDF Generator API MCP server](https://neotask.ai/integrations/pdf-generator-api-mcp-server): Instead of exporting a document manually or hand-building a PDF, describe what you need and let Neotask generate it through PDF Generator API.
- [PDF Viewer](https://neotask.ai/integrations/pdf-viewer): PDF Viewer lets Neotask display and read PDF files inline in the chat itself, using a visual viewer widget rather than dropping a raw link.
- [Peec AI](https://neotask.ai/integrations/peec-ai): Peec AI tracks how your brand shows up across AI search engines and chatbots, and connecting it gives Neotask a direct way to check that visibility on demand.
- [Peloton](https://neotask.ai/integrations/peloton): Connect your Peloton account and Neotask can browse classes and instructors on your behalf, then build a personalized training plan spanning cycling, strength, yoga, running, and meditation.
- [Penny](https://neotask.ai/integrations/penny): Penny turns shopping into something you can describe rather than search for. Tell the agent what you're looking for — or upload a photo of it — and Penny finds matching products, lets you compare options, and starts price tracking on the...
- [People API MCP server](https://neotask.ai/integrations/people-api-mcp-server): The People API connection reads and manages contacts, profile information, and Google Workspace directory data for a connected Google account.
- [Perseuss](https://neotask.ai/integrations/perseuss): When you need a quick read on shipping cost before committing to a carrier, Perseuss compares rate and delivery-time estimates across USPS, UPS, and FedEx.
- [Pet Project Co-Founder AI](https://neotask.ai/integrations/pet-project-co-founder-ai): Pet Project Co-Founder AI is built for the early, uncertain stage of a product idea — before you know if there's real demand.
- [PetPals](https://neotask.ai/integrations/petpals): PetPals covers pet shopping — toys, beds, and general supplies — starting from a plain description of what you need. Describe it to the agent, browse curated product results complete with prices and photos, and add the ones you want...
- [pg-aiguide](https://neotask.ai/integrations/pg-aiguide): pg-aiguide is a documentation and guidance connection for anyone writing PostgreSQL or TimescaleDB code. Rather than digging through official docs or half-remembered blog posts, ask the agent a question about Postgres or Tiger Data and it...
- [PGA – Play Golf](https://neotask.ai/integrations/pga-play-golf): PGA – Play Golf helps you find a PGA of America golf coach that actually fits — matched to your goals, skill level, location, and schedule.
- [PhantomBuster MCP server](https://neotask.ai/integrations/phantombuster-mcp-server): PhantomBuster automations — the "Phantoms" and scripts you've already built — become something you can operate from chat with this connection.
- [Phish.in](https://neotask.ai/integrations/phish-in): Phish.in is the archive of live Phish audience recordings, and this connection lets you search and browse it conversationally — by song, venue, or tour — with audio playing directly in chat.
- [Pine Labs](https://neotask.ai/integrations/pine-labs): Pine Labs' shopping assistant lets you discover products from trusted merchants through chat. Describe what you're after, compare the options the agent surfaces, build a cart from what you like, and head to checkout — a conversational...
- [Pipedream Remote MCP server](https://neotask.ai/integrations/pipedream-remote-mcp-server): Pipedream connects to thousands of third-party apps, and this remote MCP server exposes that same reach to Neotask's agent — running reusable actions and triggers and building integration workflows on your behalf.
- [Plain MCP server](https://neotask.ai/integrations/plain-mcp-server): Plain is where your customer support threads live, and this connection lets Neotask work inside them directly — managing threads, looking up customers, and working with help center content, with AI-assisted workflows layered on top of your...
- [Play Sheet Music](https://neotask.ai/integrations/play-sheet-music): Turning a musical idea into notation used to mean opening a score editor and placing notes by hand. Connect Play Sheet Music to Neotask and you can describe a melody, chord progression, or short piece in plain language, and the agent...
- [PlayMCP](https://neotask.ai/integrations/playmcp): PlayMCP works as a connector toolbox rather than a single service — it's the layer that lets you add and use a variety of MCP servers, including Kakao's own services, without wiring up each one separately.
- [PocketMind: Texas Hold'em](https://neotask.ai/integrations/pocketmind-texas-hold-em): PocketMind: Texas Hold'em turns a chat with Neotask into a poker table. It plays real hands of Texas Hold'em against AI opponents set at different skill levels, so you can pick a table that matches how much of a challenge you want.
- [PocketSmith MCP Server (full access)](https://neotask.ai/integrations/pocketsmith-mcp-server-full-access): This is the full-access connection to PocketSmith, meaning Neotask isn't limited to just looking at your finances — it can act on them.
- [PocketSmith MCP Server (read-only)](https://neotask.ai/integrations/pocketsmith-mcp-server-read-only): If you want Neotask looking at your finances without any risk of it changing anything, this is the connection for that. It gives the agent visibility into your PocketSmith accounts, transactions, budgets, and forecasts — enough to answer...
- [Polar MCP Server (production)](https://neotask.ai/integrations/polar-mcp-server-production): This connects Neotask to your live Polar storefront — the real one customers are buying from. Once connected, the agent can manage products, checkouts, subscriptions, and customers, and handle benefit delivery like license keys and...
- [Polar MCP Server (sandbox)](https://neotask.ai/integrations/polar-mcp-server-sandbox): Before pointing Neotask at your real Polar storefront, the sandbox connection lets you try things out safely. It runs the same kinds of checkout, subscription, and benefit-delivery workflows, but against a test environment, so you can...
- [Policy Troubleshooter MCP Server](https://neotask.ai/integrations/policy-troubleshooter-mcp-server): Google Cloud IAM permission problems are notoriously hard to diagnose by hand — a denial can come from any layer of policy.
- [PolicyNote](https://neotask.ai/integrations/policynote): Tracking policy across multiple jurisdictions usually means monitoring a scattered set of government sites and legislative trackers.
- [Popeyes](https://neotask.ai/integrations/popeyes): Connect Popeyes to Neotask and ordering food becomes a conversation instead of an app hunt. The agent can browse the current menu, find the Popeyes location closest to you, check what offers are running right now, and place an order ahead...
- [Portkey MCP Gateway](https://neotask.ai/integrations/portkey-mcp-gateway): Portkey sits between your AI agents and the upstream MCP providers they use, and connecting its gateway to Neotask means every tool call the agent makes can be routed, observed, and logged through that layer.
- [Preply Language Tutor Finder](https://neotask.ai/integrations/preply-language-tutor-finder): Finding the right language tutor on Preply usually means scrolling through a lot of profiles. Connect it to Neotask and you can describe exactly what you want — the language, a price range, tutors who speak a particular language...
- [Priceline](https://neotask.ai/integrations/priceline): Booking a trip on Priceline usually means opening the site, setting your dates and destination, and scrolling a map of deals.
- [PrintHive MCP server](https://neotask.ai/integrations/printhive-mcp-server): Running more than one 3D printer at once means juggling print queues, model files, and status across every machine. PrintHive connects Neotask to your fleet of Bambu Lab printers, letting the agent route print jobs to the right printer,...
- [Priority Guardian](https://neotask.ai/integrations/priority-guardian): Priority Guardian connects to your Google Calendar to keep track of where your time is actually going against what you said mattered.
- [Prisma MCP server](https://neotask.ai/integrations/prisma-mcp-server): Setting up a Postgres database usually means jumping into a console to provision it and configure the related resources. Connect the Prisma MCP server to Neotask and you can create and manage Prisma Postgres databases through...
- [Prismic MCP Server](https://neotask.ai/integrations/prismic-mcp-server): Prismic is a headless CMS, and connecting its MCP server to Neotask gives the agent access to a content repository the way an editor would use it — reading and managing content, content models, assets, and migrations — so content updates...
- [Process Street](https://neotask.ai/integrations/process-street): Process Street runs the checklists and workflows a lot of operational teams live in day to day. Connected to Neotask, the agent can look up workflows, workflow runs, and individual tasks, check their status and history, update form field...
- [Product Hunt](https://neotask.ai/integrations/product-hunt): Product Hunt's daily leaderboard is where a lot of new tools first surface, and connected to Neotask you can browse that leaderboard by category or topic, read what reviewers are saying, and get tools, apps, and startups compared side by...
- [Productboard MCP Server](https://neotask.ai/integrations/productboard-mcp-server): Productboard is where product teams track features, insights, and the roadmap that ties them together. Connecting its MCP server to Neotask lets the agent read and manage that data through Productboard's REST API — so checking on a...
- [Profit.co](https://neotask.ai/integrations/profit-co): Profit.co is where OKRs, balanced scorecards, and strategy execution plans live for a lot of teams, and connected to Neotask the agent can manage those directly — tracking project and task progress and reviewing performance and engagement...
- [Promocodes.com](https://neotask.ai/integrations/promocodes-com): Before checking out at an online store, Promocodes.com connected to Neotask means the agent can go search for working promo codes and discount coupons for that store on your behalf, so you're not manually hunting through a coupon site or...
- [Prompt Perfect](https://neotask.ai/integrations/prompt-perfect): Most people write a prompt the way they'd leave a voicemail — half the context implied, half assumed. Prompt Perfect connected to Neotask closes that gap.
- [PromptLayer MCP](https://neotask.ai/integrations/promptlayer-mcp): If your team is running LLM prompts in production, PromptLayer is probably where the version history, request traces, and evaluation scores live — but checking it usually means opening a separate dashboard mid-conversation.
- [ProofX](https://neotask.ai/integrations/proofx): Authorship disputes are hard to win with nothing but a file's "date modified" timestamp. ProofX exists to fix that: it registers a piece of writing, code, or other text and issues a tamper-proof cryptographic proof that you had it at a...
- [Property24.com](https://neotask.ai/integrations/property24-com): Property24 is one of the main places South Africans look for property to buy or rent, but scrolling filters on a listings site is still scrolling filters.
- [PubNub MCP Server (hosted)](https://neotask.ai/integrations/pubnub-mcp-server-hosted): Real-time messaging infrastructure tends to be invisible until something's wrong with it — a channel isn't publishing, an app config got changed, usage spiked.
- [pumaDB MCP server](https://neotask.ai/integrations/pumadb-mcp-server): Agents and small apps often need somewhere durable to keep structured facts between runs — a user's preferences, a running tally, a small dataset — without standing up a database.
- [Pylon](https://neotask.ai/integrations/pylon): Support work is mostly context-switching: read the ticket, check the account, look at message history, decide what to do, update the record. Pylon connected to Neotask keeps all of that in one conversation.
- [Qovery MCP Server](https://neotask.ai/integrations/qovery-mcp-server): Qovery manages deployments and environments on top of your own cloud infrastructure, which usually means jumping into its console to check on something mid-conversation about a release.
- [Quartr](https://neotask.ai/integrations/quartr): Following a company through earnings season means tracking calls, transcripts, slides, and filings scattered across sources.
- [Quicknode](https://neotask.ai/integrations/quicknode): Running blockchain infrastructure means keeping an eye on endpoints, usage, security settings, and billing — usually across a few different screens in Quicknode's dashboard.
- [QuiverAI MCP server](https://neotask.ai/integrations/quiverai-mcp-server): Turning a text description or a raster image into a clean, scalable SVG is exactly the kind of task that's tedious to do by hand but simple to describe.
- [Quizlet](https://neotask.ai/integrations/quizlet): Turning notes or a topic into a usable flashcard set is a chore most people put off. Quizlet connected to Neotask removes the manual formatting step — give it a topic or a set of notes, and it creates the flashcard set in Quizlet directly,...
- [Quo](https://neotask.ai/integrations/quo): Customer conversations pile up across calls, texts, and contacts, and the useful patterns — common complaints, why deals are being lost — tend to stay buried unless someone goes looking.
- [Rabbu](https://neotask.ai/integrations/rabbu): Buying a short-term rental property involves more than browsing listings — you need to know if it's actually Airbnb-eligible, how the market compares to others, and what kind of revenue it could realistically generate.
- [Rakhys](https://neotask.ai/integrations/rakhys): Shopping for electronics across Saudi Arabia and the Gulf usually means checking a handful of retailer sites separately to see who has the best price or who actually has stock.
- [Ranked AI](https://neotask.ai/integrations/ranked-ai): SEO and PPC work spans a lot of separate reports — keyword rankings, site audits, backlinks, content performance, and increasingly, how a site shows up in AI search results.
- [Raptive Food](https://neotask.ai/integrations/raptive-food): Deciding what to cook is a search problem most recipe sites make more annoying than it needs to be. Raptive Food connected to Neotask searches over a million recipes from real food creators directly — tell it a cuisine, a course, how much...
- [Reach MCP Server](https://neotask.ai/integrations/reach-mcp-server): Bad contact data — a mistyped address, a dead phone number, an unverifiable bank account — causes problems downstream that are much cheaper to catch upfront.
- [Read AI](https://neotask.ai/integrations/read-ai): Meeting notes tend to scatter across whichever tool recorded them, and finding the one thing someone said three meetings ago is painful.
- [Readwise](https://neotask.ai/integrations/readwise): Highlights you save are only useful if you can find them again, and most people's Readwise library grows faster than they revisit it.
- [Reagent Travel](https://neotask.ai/integrations/reagent-travel): Flight search sites give you a wall of filters when what you actually want is a good option quickly. Reagent Travel connected to Neotask flips that: describe your trip in plain language, get flights compared for you, receive suggested...
- [RealEstateAPI Developer MCP Server](https://neotask.ai/integrations/realestateapi-developer-mcp-server): Building on top of real estate data usually means wiring up API calls by hand and mapping responses yourself. The RealEstateAPI Developer MCP Server connects Neotask directly to the RealEstateAPI Developer MCP Server, so agent workflows...
- [RealEstateAPI MCP Server](https://neotask.ai/integrations/realestateapi-mcp-server): Same idea as the developer endpoint, aimed at the standard service: the RealEstateAPI MCP Server connects Neotask agents directly to RealEstateAPI's MCP Server, so property and market data queries can happen inside a conversation instead...
- [Realtor.com](https://neotask.ai/integrations/realtor-com): House hunting involves three separate questions that usually live in three separate tools: what's actually for sale or rent, what can I afford, and which neighborhood is right for my commute and my kids' school.
- [Rebilly MCP server](https://neotask.ai/integrations/rebilly-mcp-server): Rebilly runs subscription billing, payment orchestration, and dunning for a lot of teams, and normally checking on any of it means logging into the Rebilly dashboard and clicking through account and transaction views.
- [Rebound Scam Assistant](https://neotask.ai/integrations/rebound-scam-assistant): Scam texts and phishing emails are getting harder to eyeball with confidence, and the ones that look almost legitimate are exactly the ones that catch people out.
- [Reclaim.ai](https://neotask.ai/integrations/reclaim-ai): Reclaim.ai sits on top of your calendar and task tools to actually protect the time you keep meaning to protect. Connect it to Google Calendar or Outlook, plus Jira, Linear, ClickUp, Asana, or Todoist, and Neotask can use it to find real...
- [Recommender MCP server](https://neotask.ai/integrations/recommender-mcp-server): Google Cloud's Recommender API is normally something you check inside the console — cost recommendations here, an idle-resource flag there, a security suggestion somewhere else.
- [ReCommerce](https://neotask.ai/integrations/recommerce): Shopping for a refurbished phone or laptop usually means bouncing between a few marketplace tabs trying to compare condition grades and prices.
- [Recommerce Trade-in](https://neotask.ai/integrations/recommerce-trade-in): Getting a real number for a used phone or laptop usually starts with typing the model into three different trade-in sites just to see who's paying what.
- [Recorrido](https://neotask.ai/integrations/recorrido): Booking an intercity bus in Chile means checking a handful of operators to see who actually runs the route you want and at what price.
- [redBus](https://neotask.ai/integrations/redbus): redBus is one of the go-to booking platforms for intercity and regional bus travel, and connecting it to Neotask means you can work with it through a normal conversation instead of a booking flow built around forms and dropdowns.
- [Redocly MCP Server](https://neotask.ai/integrations/redocly-mcp-server): Keeping OpenAPI specs, a developer portal, and hosted docs all in sync is one of those jobs that quietly rots if nobody owns it.
- [Redox MCP server](https://neotask.ai/integrations/redox-mcp-server): Healthcare data rarely lives in one clean format — EHRs, payers, and providers all speak slightly different dialects, and Redox is the layer that translates between them.
- [Reducto MCP Server (hosted)](https://neotask.ai/integrations/reducto-mcp-server-hosted): Every business ends up with a pile of PDFs, scanned forms, spreadsheets, and office documents that need to become usable data at some point.
- [Refiner](https://neotask.ai/integrations/refiner): NPS and CSAT surveys pile up fast, and reading through every response to spot what's actually driving a score change is slow work.
- [Regal](https://neotask.ai/integrations/regal): Deciding on a movie night usually means checking showtimes across a couple of apps before you even know if the format you want is playing near you.
- [Releases MCP server](https://neotask.ai/integrations/releases-mcp-server): Keeping up with what changed in the tools your team relies on usually means checking a handful of changelog pages one at a time, if you remember to at all.
- [Remedy Legal](https://neotask.ai/integrations/remedy-legal): Remedy Legal connected to Neotask gives the agent a direct line into Remedy Legal's platform, so legal-workflow requests can be handled through a normal conversation instead of a separate login and interface.
- [Remote.com MCP](https://neotask.ai/integrations/remote-com-mcp): Managing a globally distributed team through an employer-of-record platform means juggling employment records, payroll, benefits, and time off across countries with different rules.
- [Render MCP Server](https://neotask.ai/integrations/render-mcp-server): Checking on a deploy or digging through logs for a service running on Render usually means opening the Render dashboard and clicking into the right project.
- [RentCafe](https://neotask.ai/integrations/rentcafe): RentCafe connected to Neotask gives the agent a direct line into Yardi's RentCafe platform, so property and leasing-related requests can be handled through a conversation instead of navigating RentCafe's own interface for every lookup.
- [RentCast MCP server](https://neotask.ai/integrations/rentcast-mcp-server): Pricing a rental property or checking on a market usually means pulling data from a few different real-estate sources and reconciling them yourself.
- [Replit](https://neotask.ai/integrations/replit): Turning an app idea into something running usually means setting up a project, writing the scaffolding, and configuring an environment before you've even seen it work.
- [Resend MCP](https://neotask.ai/integrations/resend-mcp): Sending transactional or marketing email, and keeping contacts, domains, broadcasts, and templates organized, is usually split across a dashboard and whatever code sends the actual messages.
- [Responsive](https://neotask.ai/integrations/responsive): Answering RFP and proposal questions well usually means finding the right previously-approved answer buried somewhere in a shared drive.
- [Resume Builder](https://neotask.ai/integrations/resume-builder): Most resumes have the same problem: bullet points that describe a task instead of an outcome, and gaps nobody has flagged.
- [Resume.io](https://neotask.ai/integrations/resume-io): Resume.io turns a blank page into a finished, ATS-friendly resume, and Neotask can drive the whole build for you. Tell the agent what job you're targeting and what your background looks like, and it works through Resume.io's templates,...
- [Resy](https://neotask.ai/integrations/resy): Booking a table shouldn't mean checking five different restaurant sites. Connect Resy and Neotask can search nearby restaurants, check what's actually open at the time you want, and lock in a reservation for your party size — filtered by...
- [Retell AI](https://neotask.ai/integrations/retell-ai): Retell AI is how a business gets an AI voice agent answering its phones — a receptionist, a sales caller, or a support line.
- [RIFO Home](https://neotask.ai/integrations/rifo-home): House hunting rarely fits into dropdown filters. RIFO Home lets you describe what you actually want — budget, commute, school quality, being near a park — and Neotask's agent turns that description into a real search, surfacing listings...
- [Rillet](https://neotask.ai/integrations/rillet): Rillet holds the live general ledger, and Neotask can query it the way you'd ask a finance analyst. Ask for revenue trends or SaaS metrics in plain English, pull a balance sheet or cash flow statement, or have the agent draft a journal...
- [RoadOps](https://neotask.ai/integrations/roadops): Touring runs on the day sheet — venue details, hotel info, lobby call, who's where. RoadOps keeps that live, and Neotask's agent can update it in real time or just tell you what's next: which venue, what time the bus leaves, where the team...
- [Rome2Rio](https://neotask.ai/integrations/rome2rio): Getting from one place to another isn't always one mode of transport. Rome2Rio compares train, bus, ferry, car, and flight in one search, and Neotask's agent can run that comparison for you door-to-door, laying out travel times,...
- [Rootly MCP Server](https://neotask.ai/integrations/rootly-mcp-server): When something breaks, Rootly is where the incident lives. Connect the Rootly MCP Server and Neotask's agent can open or update an incident directly, check who's on call right now, and pull up the retrospective from a past incident without...
- [RunReady](https://neotask.ai/integrations/runready): Race day goes better with course knowledge you didn't have to dig up yourself. RunReady covers supported races with an interactive course map, elevation and aid-station breakdowns, and a pacing plan built around your goal time — and...
- [Safe2Sign: Contract Risk Check](https://neotask.ai/integrations/safe2sign-contract-risk-check): Before signing a lease, employment agreement, or freelance contract, it helps to know what's actually risky in it. Safe2Sign runs the document against a checklist of common red flags and Neotask can walk you through it clause by clause — a...
- [Saleor MCP server](https://neotask.ai/integrations/saleor-mcp-server): Saleor runs the store's product catalog, orders, and configuration as a headless GraphQL backend. Connect the Saleor MCP server and Neotask can look up catalog data, check an order's details, and run commerce operations directly against...
- [SalesEQ](https://neotask.ai/integrations/saleseq): SalesEQ holds the recordings and transcripts from your sales calls and meetings. Neotask's agent can search across them, pull an AI-generated summary or insight from a specific call, and share a meeting with a teammate — turning a pile of...
- [SAMRUM](https://neotask.ai/integrations/samrum): SAMRUM runs a 10-axis, Big Five-inspired personality assessment for family members and stores the results so they can be compared.
- [Sanity](https://neotask.ai/integrations/sanity): Sanity holds your structured content — documents, schemas, datasets — and Neotask's agent can operate directly inside a project: run a content audit, publish or unpublish a document, coordinate a content release across items, search across...
- [Savecraft](https://neotask.ai/integrations/savecraft): Savecraft reads your actual game save files and linked profile data, so Neotask's agent can give advice grounded in your real character, gear, and progress — not generic strategy.
- [Scalar Docs MCP](https://neotask.ai/integrations/scalar-docs-mcp): Scalar Docs MCP puts Scalar's own developer documentation and its OpenAPI spec registry in reach of the agent. Ask how to publish docs, manage a spec, or expose an API to AI agents through Scalar, and it searches the actual docs and...
- [Scalar Installation MCP](https://neotask.ai/integrations/scalar-installation-mcp): This connection wires an agent directly to a specific Scalar installation — the one tied to your own install ID rather than Scalar's general public docs.
- [ScholarXIV MCP server](https://neotask.ai/integrations/scholarxiv-mcp-server): ScholarXIV is a research library, and Neotask's agent can work it like a research assistant — searching academic papers, saving the useful ones into a collection, adding bookmarks and comments as you go, and using ScholarXIV's own...
- [SciSpace](https://neotask.ai/integrations/scispace): SciSpace turns a pile of academic papers into a working table — columns for abstract, summary, methods, results, conclusions — so you can actually compare them side by side.
- [Scite](https://neotask.ai/integrations/scite): Scite grounds an answer in whether the underlying research actually supports it — showing per-claim whether a study has been supported, disputed, or given more context by later work.
- [Scrapfly MCP Server](https://neotask.ai/integrations/scrapfly-mcp-server): Scrapfly handles the messy parts of web scraping — rendering JavaScript-heavy pages, rotating through proxies, and getting past anti-bot defenses — behind a single API.
- [ScrapingBee Remote MCP Server](https://neotask.ai/integrations/scrapingbee-remote-mcp-server): ScrapingBee fetches rendered pages, search results, and e-commerce data without you running your own scraping proxies. Neotask's agent can pull a page through ScrapingBee, grab a screenshot, or use its AI-assisted extraction to pull out...
- [Scrimba](https://neotask.ai/integrations/scrimba): Scrimba's interactive screencasts let you edit the code shown in the lesson while you watch, and Neotask's agent can point you to the right one — JavaScript, React, Node, Next, SQL — for whatever you're trying to learn, then help you dig...
- [Seaclub Alcudia](https://neotask.ai/integrations/seaclub-alcudia): Seaclub Alcudia is a beach-club venue in Alcudia, Mallorca, and this connection lets Neotask's agent reach its booking and information surface directly from a conversation instead of you hunting down a reservations page or WhatsApp thread.
- [Search Console MCP server](https://neotask.ai/integrations/search-console-mcp-server): This connection puts Google Search Console's own data -- impressions, clicks, position, indexing status -- into a conversation with Neotask's agent.
- [Seat Unique](https://neotask.ai/integrations/seat-unique): Seat Unique specializes in premium tickets and official hospitality packages for sports, music, and cultural events, and this connection lets you ask Neotask's agent to find and compare those options instead of browsing the site's category...
- [SeatGeek](https://neotask.ai/integrations/seatgeek): SeatGeek's strength is its Deal Score and venue map view, and this connection brings that into a conversation with Neotask's agent -- describe what you're looking for nearby and it surfaces real listings instead of you scrolling an app.
- [Seatpin](https://neotask.ai/integrations/seatpin): Seatpin aggregates listings from independent ticket sellers for football matches, concerts, and live events worldwide, and this connection lets Neotask's agent do the price-and-seat comparison work across sellers instead of you opening...
- [SecretBox](https://neotask.ai/integrations/secretbox): SecretBox curates dining, hotel, and spa experiences across France and Europe, and this connection lets Neotask's agent ask it directly for recommendations rather than you scrolling through curated guides yourself.
- [SecureLend](https://neotask.ai/integrations/securelend): SecureLend compares financing across business loans, personal loans, mortgages, auto financing, and student loans, and this connection lets Neotask's agent run those rate and payment comparisons for you instead of filling out a comparison...
- [SecureLend AI Agents MCP](https://neotask.ai/integrations/securelend-ai-agents-mcp): This is SecureLend's lender-facing surface rather than its consumer comparison tool -- it runs AI-driven loan comparison and underwriting workflows for lenders and financial institutions, and Neotask's agent can drive those workflows...
- [Selah - Bible stories](https://neotask.ai/integrations/selah-bible-stories): Selah - Bible stories gives Neotask's agent access to its library of Bible stories, so instead of searching an app or a website you can ask directly for a story, a passage retelling, or content built around a specific theme.
- [Semaphore MCP Server](https://neotask.ai/integrations/semaphore-mcp-server): Semaphore runs CI/CD pipelines, and this connection lets Neotask's agent operate those pipelines directly -- trigger a build, check whether tests passed, or review a deployment configuration without you opening the Semaphore dashboard.
- [Semrush](https://neotask.ai/integrations/semrush): Semrush is a heavyweight SEO and traffic data platform, and this connection lets Neotask's agent pull that data directly -- keyword metrics, backlink profiles, competitor comparisons -- instead of you exporting reports from the Semrush...
- [Send](https://neotask.ai/integrations/send): Send turns documents, one-pagers, decks, and presentations into shareable interactive web pages, and this connection lets Neotask's agent build and publish those pages directly, plus track what happens after you share them.
- [SEPHORA](https://neotask.ai/integrations/sephora): Sephora's connection lets Neotask's agent give personalized beauty recommendations across skincare, makeup, hair care, and fragrance, grounded in the goals and budget you describe rather than a generic product list.
- [Sequel.io](https://neotask.ai/integrations/sequel-io): Sequel.io turns webinar recordings into written content, and this connection lets Neotask's agent pull a transcript from your connected Sequel account and generate blog posts, social content, or a newsletter from what was actually said on...
- [Serasa](https://neotask.ai/integrations/serasa): Serasa is a Brazilian credit bureau and financial-services company, and this connection lets Neotask's agent answer questions about credit scores, debt negotiation, payment plans, and data protection using Serasa's own information instead...
- [SerpApi MCP Server](https://neotask.ai/integrations/serpapi-mcp-server): SerpApi scrapes real-time results from Google, Bing, Baidu, and other search engines, and this connection lets Neotask's agent request a specific type of result -- organic, local, images -- and get back parsed data instead of a raw...
- [Serpstat](https://neotask.ai/integrations/serpstat): Serpstat is built for keyword clustering and technical SEO auditing, and this connection lets Neotask's agent build semantic keyword clusters, research competitor strategy, and flag technical site issues without you running each report...
- [Setu Bharat Connect BillPay](https://neotask.ai/integrations/setu-bharat-connect-billpay): Setu Bharat Connect BillPay lets you pay utility bills -- electricity, broadband, and similar recurring bills -- through conversation instead of opening a separate app or web form for each provider.
- [Shippo hosted MCP server](https://neotask.ai/integrations/shippo-hosted-mcp-server): Shippo handles multi-carrier shipping logistics, and this connection lets Neotask's agent compare rates, validate addresses, create labels, and track packages directly instead of you logging into the Shippo dashboard for each shipment.
- [Shipt](https://neotask.ai/integrations/shipt): Shipt delivers groceries and household essentials, and this connection lets Neotask's agent go from a meal or shopping idea straight to a scheduled delivery -- planning the list, adding items to a cart, and picking a delivery window in one...
- [Shop](https://neotask.ai/integrations/shop): Shop.app searches across local stores and major brands and lets you check out directly, and this connection lets Neotask's agent turn a shopping idea described in plain language into a real product search with a purchase option, instead of...
- [ShopBack](https://neotask.ai/integrations/shopback): ShopBack surfaces cashback offers tied to specific merchants and categories, and this connection lets Neotask's agent check current cashback rates for a brand before you buy, instead of you searching the ShopBack app merchant by merchant.
- [Shopee](https://neotask.ai/integrations/shopee): Shopee is a major Southeast Asian e-commerce marketplace, and this connection lets Neotask's agent turn a plain-language shopping idea into a focused product search on Shopee, compare listings, and complete the purchase there.
- [Shopify storefront UCP MCP server](https://neotask.ai/integrations/shopify-storefront-ucp-mcp-server): This connection queries a Shopify-hosted storefront's product catalog, pricing, and inventory directly through GraphQL, so Neotask's agent can pull real storefront data for a specific shop into a commerce workflow instead of you checking...
- [Shortcut MCP Server](https://neotask.ai/integrations/shortcut-mcp-server): Shortcut is project-management software for software teams, and this connection lets Neotask's agent manage stories, epics, iterations, and docs directly -- creating or updating a story, checking iteration progress, or reviewing roadmap...
- [Shotstack MCP Server (hosted)](https://neotask.ai/integrations/shotstack-mcp-server-hosted): Video and image edits usually mean a timeline editor and a lot of manual dragging. Shotstack skips that: it takes a JSON edit timeline describing clips, transitions, overlays, and audio, then renders the finished asset on its own render...
- [Shufti MCP Server](https://neotask.ai/integrations/shufti-mcp-server): Shufti connects agent workflows to Shufti's identity verification platform. Once linked, Neotask's agent can trigger and manage Shufti-based checks as part of a broader task instead of you jumping into a separate verification dashboard.
- [Shutterstock](https://neotask.ai/integrations/shutterstock): Finding the right stock photo, clip, or sound effect usually means opening Shutterstock, typing search terms, scrolling results, and downloading manually.
- [Sigma documentation MCP server](https://neotask.ai/integrations/sigma-documentation-mcp-server): Standing up dashboards or data apps on Sigma often means digging through docs to figure out how to connect a warehouse and set up governed access correctly.
- [SignNow](https://neotask.ai/integrations/signnow): E-signature workflows involve several small steps — sending a request, pre-filling a template, checking who has and hasn't signed, retrieving the final file.
- [SigNoz MCP Server](https://neotask.ai/integrations/signoz-mcp-server): SigNoz is an observability platform for traces, metrics, and logs. This connection gives Neotask's agent a direct line into your SigNoz instance, so observability questions and tasks can be handled as part of a normal conversation rather...
- [Simcardo Connectivity Planner](https://neotask.ai/integrations/simcardo-connectivity-planner): Trip connectivity planning through Simcardo helps figure out how much mobile data you'll actually need and which eSIM setup fits a specific trip.
- [Simcardo MCP](https://neotask.ai/integrations/simcardo-mcp): Before a trip, figuring out an eSIM plan usually means comparing multiple providers' data allowances and prices by hand. With Simcardo connected, you tell the agent where you're going and it estimates the data you'll need for that...
- [Simply Business](https://neotask.ai/integrations/simply-business): Getting a business insurance quote usually starts with a form asking about your business, coverage needs, and risk factors.
- [Simply Family Budget](https://neotask.ai/integrations/simply-family-budget): Your bank balance rarely tells the real story — rent hasn't cleared yet, a shared bill is about to split three ways, payday is four days out.
- [Simsurf](https://neotask.ai/integrations/simsurf): Comparing eSIM plans across dozens of providers for a single trip is tedious by hand. Simsurf does that comparison at scale — thousands of plans from many providers — so once you tell the agent your data needs and how long you'll be...
- [SIXT](https://neotask.ai/integrations/sixt): Renting a car usually means opening a rental site, entering pickup location and dates, and scrolling through vehicle classes and prices.
- [Skillshare](https://neotask.ai/integrations/skillshare): Finding the right class on Skillshare out of thousands of options is easier described than searched. Tell the agent your learning goal or project idea — "I want to get better at watercolor landscapes" — and it searches Skillshare for...
- [SkipCalls MCP server](https://neotask.ai/integrations/skipcalls-mcp-server): SkipCalls runs an AI phone receptionist for a business — answering calls, capturing leads, booking appointments, and handling outbound calls and texts on a dedicated business number.
- [SkipCalls MCP server (legacy SSE)](https://neotask.ai/integrations/skipcalls-mcp-server-legacy-sse): This is the legacy SSE transport variant of the SkipCalls connection, kept available for setups still wired to that endpoint.
- [Skyscanner](https://neotask.ai/integrations/skyscanner): Cross-airline fare comparisons through Skyscanner are exactly the kind of research task the agent can take off your plate.
- [SkyWatch](https://neotask.ai/integrations/skywatch): Sourcing satellite imagery normally means checking multiple providers separately — Vantor, Planet, Airbus — each with its own catalog and pricing.
- [SkyWatch.ai](https://neotask.ai/integrations/skywatch-ai): Non-owned aircraft renters insurance for pilots usually involves a long form covering flight experience and coverage preferences.
- [Sleep Cycle](https://neotask.ai/integrations/sleep-cycle): When you're having trouble winding down, opening an app and scrolling for a sound isn't what you need in the moment. Sleep Cycle's connection lets you just ask — the agent starts playback of science-backed sleep sounds designed to help...
- [Slevomat](https://neotask.ai/integrations/slevomat): Slevomat runs deals on gifts, travel, and seaside holiday packages. Instead of browsing categories yourself, describe the kind of experience or getaway you're after — a weekend gift idea, a beach holiday for two — and the agent pulls...
- [SlidesGPT MCP Server](https://neotask.ai/integrations/slidesgpt-mcp-server): Building a slide deck from scratch is one of those tasks that eats an afternoon. SlidesGPT turns a prompt or source material into a styled PowerPoint deck directly — describe the content, structure, or source document you want turned into...
- [Smallpdf](https://neotask.ai/integrations/smallpdf): Smallpdf connects agent workflows to Smallpdf's PDF tools, letting document conversion and cleanup tasks happen as part of a normal conversation with Neotask instead of a separate trip to the Smallpdf site.
- [SmartBear MCP Server (BugSnag remote)](https://neotask.ai/integrations/smartbear-mcp-server-bugsnag-remote): When something breaks in production, the first step is usually checking BugSnag for what's crashing and how often. This connection lets the agent look up error reports and stability data from BugSnag directly, so you can ask "what's our...
- [SmartBear MCP Server (Swagger remote)](https://neotask.ai/integrations/smartbear-mcp-server-swagger-remote): Keeping track of API definitions across a growing SwaggerHub catalog gets messy fast. This connection lets the agent search OpenAPI specifications hosted in SwaggerHub, review different versions, and check documentation for a specific...
- [SmartBear MCP Server (Zephyr remote)](https://neotask.ai/integrations/smartbear-mcp-server-zephyr-remote): Test coverage and execution status live in Zephyr, tied to Jira's test management data. This connection lets the agent look up test cases, test cycles, and execution results directly, so you can ask "how much of the login flow is covered...
- [SmartCustomer](https://neotask.ai/integrations/smartcustomer): SmartCustomer pulls together how customers actually feel about a business from the reviews and feedback already sitting out there.
- [Smarty MCP Server](https://neotask.ai/integrations/smarty-mcp-server): Bad addresses cost money: failed deliveries, bounced mail, blocked checkouts. Smarty fixes that at the source. Give Neotask a messy or partial address and it verifies, standardizes, and geocodes it through Smarty before it ever reaches a...
- [SMTP2GO MCP server](https://neotask.ai/integrations/smtp2go-mcp-server): Instead of logging into SMTP2GO to check whether a message actually went out, ask Neotask directly. It can send transactional email or SMS through your SMTP2GO account, check delivery status on something you already sent, and pull sending...
- [SNCF Connect : Trains & Trajet](https://neotask.ai/integrations/sncf-connect-trains-and-trajet): Planning a train trip in France or across Europe usually means opening SNCF Connect and clicking through filters one at a time. Tell Neotask the departure city, destination, and date instead.
- [Snowplow MCP server](https://neotask.ai/integrations/snowplow-mcp-server): Behavioral event pipelines drift: schemas get out of sync, tracking plans go stale, and nobody notices until a dashboard breaks.
- [Softonic](https://neotask.ai/integrations/softonic): Rather than browsing Softonic's catalog yourself, describe what kind of software you need and Neotask searches it for you by category, platform, or rating, and can surface the best-rated or most popular pick for Windows, Mac, Android, iOS,...
- [Solo 401k Calculator](https://neotask.ai/integrations/solo-401k-calculator): Solo 401(k) contribution limits depend on your income, age, and business structure, and getting the math wrong means leaving money on the table or over-contributing.
- [SonicJobs](https://neotask.ai/integrations/sonicjobs): Instead of searching job boards one at a time, ask Neotask for roles matching a title, company, or location, and it queries SonicJobs across millions of listings from many employers in a single pass, then points you to the original listing...
- [Soul Family AI](https://neotask.ai/integrations/soul-family-ai): Soul Family AI gives you access to guided meditations and spiritual teachings from featured teachers like Kryon and The Stargate Experience.
- [Sourcegraph MCP server](https://neotask.ai/integrations/sourcegraph-mcp-server): Finding where something is actually defined across a large codebase usually means switching to Sourcegraph and running a search yourself.
- [Spaceship](https://neotask.ai/integrations/spaceship): Before registering a domain, ask Neotask to check it through Spaceship: availability, pricing across different TLDs, and alternative name suggestions if your first choice is already taken.
- [SparkSpot](https://neotask.ai/integrations/sparkspot): Finding parking near a destination usually means circling the block or opening yet another app. Tell Neotask where you're headed and it searches SparkSpot for available private parking spaces nearby, compares the options, and can reserve...
- [Speechify](https://neotask.ai/integrations/speechify): Turn pasted text, an article, or an uploaded PDF into natural-sounding speech by asking Neotask, which hands it to Speechify to generate the audio.
- [Spinify](https://neotask.ai/integrations/spinify): Sales reps and managers can check where they stand without opening a separate leaderboard app. Ask Neotask for your current standing, a live leaderboard for an active competition, or a coaching report showing who's on track versus at risk,...
- [Splice](https://neotask.ai/integrations/splice): Describe the sound you're after in plain language, a warm analog bassline or glitchy percussion, and Neotask searches Splice's royalty-free sample catalog for it, can generate a stack of complementary sounds for the project, and hands you...
- [Spreedly Docs MCP Server](https://neotask.ai/integrations/spreedly-docs-mcp-server): When you're building against Spreedly's payment orchestration platform, ask Neotask instead of searching the docs site yourself.
- [Stable Baseline](https://neotask.ai/integrations/stable-baseline): Stable Baseline gives you a shared knowledge base for structured documents and project work. Ask Neotask to draft a document, insert a diagram (Mermaid, BPMN, D2, PlantUML, or Vega), manage a project plan with tasks and phases, or track an...
- [Stack Overflow MCP Server](https://neotask.ai/integrations/stack-overflow-mcp-server): Rather than searching Stack Overflow in a browser tab, ask Neotask the question directly. It searches real Stack Overflow questions and answers, surfaces the accepted solution, and can pull details on tags, users, or reputation from across...
- [Standvirtual](https://neotask.ai/integrations/standvirtual): Describe the car you're looking for in plain language (make, budget, mileage, whatever matters to you) and Neotask searches Standvirtual's Portuguese car marketplace for matching listings, compares vehicles side by side, and refines...
- [Statsig](https://neotask.ai/integrations/statsig): Product teams can manage experiments without opening the Statsig console every time. Ask Neotask to inspect a feature gate, change a rollout percentage, toggle a flag off, or review which configs changed recently, and it acts directly...
- [Steadily](https://neotask.ai/integrations/steadily): Enter a rental property address and ask Neotask for a landlord insurance estimate. It gets an instant estimated monthly and annual premium from Steadily, and you can continue on Steadily's site afterward to customize coverage and bind the...
- [Steer Astro](https://neotask.ai/integrations/steer-astro): Set up your personal profile once, and Neotask can pull your Vedic birth chart, current planetary transits, dasha period timing, and daily panchang from Steer Astro whenever you ask, whether the question is about career timing, a...
- [Sticklight](https://neotask.ai/integrations/sticklight): Describe the application you want to build in plain language, and Neotask uses Sticklight to generate a working web project (structure, interface, and core logic already in place) that you can then manually edit, adjusting layouts,...
- [Storyblok MCP Server](https://neotask.ai/integrations/storyblok-mcp-server): Storyblok splits content into stories, components, and assets, and most of the day-to-day work in it is finding the right story and checking what's actually live versus what's still a draft.
- [StraighterLine Degree Planner](https://neotask.ai/integrations/straighterline-degree-planner): Planning a degree usually means juggling four things at once: how much time you have, how much it costs, what your major actually requires, and whether the plan is realistic given your pace.
- [Strapi Docs MCP server](https://neotask.ai/integrations/strapi-docs-mcp-server): Self-hosted Strapi has a lot of surface area — content-type builder, admin API, content API, and a permissions system that trips people up constantly.
- [Streak](https://neotask.ai/integrations/streak): Streak turns Gmail itself into a CRM — deals, contacts, and pipeline stages live next to your inbox instead of in a separate app, and emails get logged to the right record automatically.
- [Strive PDF Generator](https://neotask.ai/integrations/strive-pdf-generator): Strive turns a plain-language description — or actual LaTeX — into a formatted PDF, including graphs and math notation, which matters for anyone producing worksheets, exams, resumes, or papers that need to look properly typeset rather than...
- [StubHub](https://neotask.ai/integrations/stubhub): Finding a good seat at a fair price on StubHub usually means opening a dozen tabs to compare sellers, prices, and seat views for the same show.
- [Subdivide](https://neotask.ai/integrations/subdivide): Practicing complex rhythms usually means fighting with a metronome app's UI mid-session just to add a septuplet subdivision or mute one beat to isolate a pattern.
- [SubscriptionFlow](https://neotask.ai/integrations/subscriptionflow): Running a subscription business means watching billing, churn, and revenue workflows simultaneously, and none of those numbers are useful if you have to log into a dashboard every time you want a read on them.
- [Substack MCP Server](https://neotask.ai/integrations/substack-mcp-server): Publishing a newsletter is one thing; keeping up with subscribers is another, and Substack asks you to do both from its own dashboard.
- [Success.co](https://neotask.ai/integrations/success-co): Running EOS (Entrepreneurial Operating System) sessions well depends on everyone showing up with real numbers — scorecards, Rocks, Issues — not a vague sense of where things stand.
- [SueldoNeto](https://neotask.ai/integrations/sueldoneto): Spanish payroll deductions depend on more variables than most people want to track by hand: IRPF income tax brackets, Social Security withholdings, your autonomous community, dependents, and how your pay is split across the year.
- [Sumsub MCP server](https://neotask.ai/integrations/sumsub-mcp-server): Compliance review work in Sumsub means moving between identity verification cases, AML screening hits, and transaction monitoring alerts, often under time pressure to clear a queue.
- [Superhuman Mail](https://neotask.ai/integrations/superhuman-mail): Superhuman is built for speed, and connecting it to Neotask extends that speed into actual delegation: search the inbox, draft a reply that matches how you'd actually write to that specific person, check whether something's been read,...
- [Superserve MCP Server (hosted)](https://neotask.ai/integrations/superserve-mcp-server-hosted): Spinning up an isolated environment to run untrusted or experimental code usually means provisioning infrastructure by hand first.
- [SURI.CZ](https://neotask.ai/integrations/suri-cz): Comparing insurance offers in the Czech Republic means wading through coverage terms that are hard to compare side by side without a background in insurance.
- [SurveyMonkey](https://neotask.ai/integrations/surveymonkey): Building a survey in SurveyMonkey means adding pages and questions in the right order, thinking through the distribution plan, and then coming back later to check results — three separate visits to the dashboard for one survey.
- [Swyft Filings](https://neotask.ai/integrations/swyft-filings): Business formation and compliance filings are the kind of task that's easy to put off because starting the paperwork feels like its own project.
- [Sybill](https://neotask.ai/integrations/sybill): A sales rep juggling a full pipeline doesn't have time to re-listen to every call before a follow-up. Sybill already unifies call transcripts, deal summaries, MEDDIC scores, email, and CRM history — connect it to Neotask and you can ask...
- [Symbol](https://neotask.ai/integrations/symbol): Repeating the same client details, brand guidelines, or process notes at the start of every new chat wastes time and invites drift when someone paraphrases it slightly differently each time.
- [TalkGen](https://neotask.ai/integrations/talkgen): Turning a photo or an avatar into a lip-synced talking video used to require actual video editing software. TalkGen does the generation itself — upload a photo or pick an avatar, choose a language, and get a lip-synced talking video back —...
- [Tally MCP server](https://neotask.ai/integrations/tally-mcp-server): Tally forms are simple to build but still mean logging into the dashboard to check whether responses are coming in or to pull the data out for analysis.
- [Tango](https://neotask.ai/integrations/tango): Federal contracting intelligence means searching opportunities and awards, then untangling which agency or vendor name actually maps to the entity you're researching — government naming conventions are not consistent.
- [Target](https://neotask.ai/integrations/target): Grocery-plus-everything-else shopping at Target usually means bouncing between the app for ideas, the site for the actual basket, and then deciding on the spot how you want the order fulfilled.
- [Tarot](https://neotask.ai/integrations/tarot): A tarot reading is a small, private ritual — shuffle, draw, and reflect — and this connection reproduces that mechanic rather than replacing it with something else.
- [Taskrabbit Booking Assistance](https://neotask.ai/integrations/taskrabbit-booking-assistance): Hiring help around the house usually starts with a browser tab and a form asking questions you don't have precise answers to yet — how many rooms, what kind of space, when do you need it done.
- [Tavily](https://neotask.ai/integrations/tavily): Tavily is a search API built specifically for feeding accurate, current information into AI agents rather than into a browser.
- [Taxdown](https://neotask.ai/integrations/taxdown): Spanish tax rules — IRPF brackets, which deductions apply to a freelancer versus a salaried worker, what counts as a fiscal obligation this quarter — are notoriously easy to get wrong without a gestor.
- [Teamwork.com](https://neotask.ai/integrations/teamwork-com): Teamwork.com connected to Neotask puts your project and task data in reach of the agent the same way any other connected app does — the agent can operate on real Teamwork.com data on your behalf instead of you switching over to the app to...
- [TeklifimGelsin](https://neotask.ai/integrations/teklifimgelsin): Comparing loan offers usually means opening several bank sites and re-entering the same amount and term over and over. TeklifimGelsin flips that: tell Neotask's agent the loan amount and repayment term you want, and it pulls offers from...
- [Telnyx API MCP](https://neotask.ai/integrations/telnyx-api-mcp): Telnyx's platform covers a lot of ground for a communications provider — phone numbers, messaging, programmable voice, wireless SIM connectivity, even AI inference — and normally that means juggling separate API calls for each piece.
- [Texas Electricity Plans](https://neotask.ai/integrations/texas-electricity-plans): Texas's deregulated electricity market means dozens of retailers competing on rate structures that are genuinely hard to compare at a glance — different contract lengths, different renewable mixes, teaser rates that don't hold.
- [The Knot](https://neotask.ai/integrations/the-knot): Building a wedding vendor team means comparing venues and vendors across photos, reviews, and pricing that live on dozens of different sites.
- [TheHotelsNetwork](https://neotask.ai/integrations/thehotelsnetwork): TheHotelsNetwork works differently from a typical booking site aggregator — instead of routing you through a middleman, it pulls verified prices, amenities, and policies straight from each hotel's own site.
- [Third Bridge](https://neotask.ai/integrations/third-bridge): Third Bridge's value is its library of expert interview transcripts — primary-source conversations with industry insiders that analysts use for diligence rather than relying on public filings alone.
- [ThomasCook](https://neotask.ai/integrations/thomascook): Package holidays live and die on comparing bundled flight-and-hotel deals against each other, which usually means several browser tabs open at once.
- [ThoughtSpot Spotter](https://neotask.ai/integrations/thoughtspot-spotter): ThoughtSpot Spotter is built around one idea: business teams shouldn't need a dashboard or a SQL query to get a validated answer out of their own data.
- [ThousandEyes MCP Server](https://neotask.ai/integrations/thousandeyes-mcp-server): Troubleshooting a network incident often means checking observability data, vulnerability advisories, and security signals in three different tools before you can even start reasoning about root cause.
- [Three.js 3D Viewer](https://neotask.ai/integrations/three-js-3d-viewer): Instead of describing what a 3D scene should look like and hoping the words land, Three.js 3D Viewer lets Neotask's agent actually generate and render it — interactive scenes, models, and animations built with the Three.js library and...
- [Thumbtack](https://neotask.ai/integrations/thumbtack): Home projects usually start with a vague sense of what needs doing and no idea who's actually good at it locally. Ask Neotask's agent about a repair, maintenance task, or improvement project and Thumbtack matches you with local...
- [TickerSage](https://neotask.ai/integrations/tickersage): TickerSage connects Neotask's agent to its service so market or ticker-related requests can be handled directly through the conversation rather than a separate lookup step.
- [Ticketmaster](https://neotask.ai/integrations/ticketmaster): Finding out what's actually playing near you — concerts, sports, theater — usually means checking Ticketmaster directly and scrolling listings.
- [Tigris Remote MCP Server](https://neotask.ai/integrations/tigris-remote-mcp-server): Tigris gives developers S3-compatible object storage that's globally distributed, and this connection lets Neotask's agent manage it directly — creating buckets, uploading and retrieving objects — without a developer writing separate...
- [tiket.com](https://neotask.ai/integrations/tiket-com): tiket.com covers a wide swath of travel booking in Indonesia — flights, trains, hotels, villas, and event tickets — in one marketplace.
- [Tillys](https://neotask.ai/integrations/tillys): Instead of scrolling a clothing site's category pages hoping something matches what's in your head, describe the style you want to Neotask's agent — a look, an occasion, a vibe — and Tillys returns personalized product suggestions and...
- [Tim Hortons](https://neotask.ai/integrations/tim-hortons): A simple but genuinely useful connection: ask Neotask's agent where the nearest Tim Hortons is and it finds nearby locations directly, without you opening a maps app or the Tim Hortons site yourself.
- [Tinman AI](https://neotask.ai/integrations/tinman-ai): Mortgage underwriting involves running the same scenario several ways — different debt-to-income structures, different qualifying income calculations, checking which lenders a loan actually fits.
- [to3D](https://neotask.ai/integrations/to3d): Turning a flat image into something usable in a 3D pipeline is normally a specialized modeling task. to3D lets you upload a 2D image to Neotask's agent and get back a simple 3D glTF model converted from it, with a preview so you can check...
- [Token Terminal](https://neotask.ai/integrations/token-terminal): Comparing blockchain protocols on fundamentals — revenue, fees, earnings, valuation multiples — usually means digging through inconsistent dashboards or writing your own SQL against raw onchain data.
- [Tomorrow.io Weather](https://neotask.ai/integrations/tomorrow-io-weather): Weather questions usually turn into three or four separate app checks — current conditions, the hourly forecast, air quality, and then a second city to compare against.
- [TomTom Maps MCP Server](https://neotask.ai/integrations/tomtom-maps-mcp-server): Location features usually mean wiring up a mapping API directly — geocoding, routing, traffic — before you can answer a single question about where something is or how to get there.
- [Tonita](https://neotask.ai/integrations/tonita): Shopping across dozens of stores for one item — the right jacket in the right size at a price you'll actually pay — usually means opening tabs for every retailer you can think of.
- [Toolzy](https://neotask.ai/integrations/toolzy): Buying a power washer or a tile saw for one weekend job rarely makes sense, but finding a neighbor who'll rent you theirs usually means digging through a local marketplace app.
- [Tourlane](https://neotask.ai/integrations/tourlane): Planning a custom multi-stop trip from scratch — the kind with a rough idea but no fixed plan — is exactly what Tourlane is built for.
- [TourRadar](https://neotask.ai/integrations/tourradar): Choosing a multi-day organized tour usually means comparing options across dozens of operators — river cruises here, safaris there, each with a different site and layout.
- [TourRadar MCP (ChatGPT connector)](https://neotask.ai/integrations/tourradar-mcp-chatgpt-connector): This is TourRadar's separate connector built for embedding its tour-search catalog into other platforms — travel agencies, booking platforms, and airlines that want to surface bookable multi-day tours without building that search...
- [Trade Me Property](https://neotask.ai/integrations/trade-me-property): Property hunting in New Zealand largely happens on Trade Me, but describing what you actually want — a three-bedroom place near a specific suburb, under a price cap, with a yard — usually means fiddling with filters on a map.
- [Trainline](https://neotask.ai/integrations/trainline): For a quick "what time is the next train to Manchester" question, opening a full booking app is overkill. Trainline's connection is built for exactly that quick lookup: timetables and departure times for routes across the UK and Europe,...
- [traktorpool](https://neotask.ai/integrations/traktorpool): Shopping for a used tractor or piece of farm machinery is its own specialized search — general marketplaces don't have the listing depth, and Traktorpool exists specifically for it.
- [Transloadit hosted MCP server](https://neotask.ai/integrations/transloadit-hosted-mcp-server): Building a file processing pipeline — resize this image, transcode that video, then push the result to S3 — usually means writing and maintaining the orchestration yourself.
- [Trash Panda - Food Scanner](https://neotask.ai/integrations/trash-panda-food-scanner): Reading an ingredient label and actually understanding what's in it — which additives matter, what's an ultra-processed filler, what's hidden sugar under a different name — is not something most labels make easy.
- [Travel Impact Model MCP server](https://neotask.ai/integrations/travel-impact-model-mcp-server): Estimating a specific flight's greenhouse gas emissions accurately — not a generic per-mile average, but figures broken down by direct-route, typical-route, and Scope 3 reporting — requires Google's Travel Impact Model.
- [Travelenie](https://neotask.ai/integrations/travelenie): Most travel tools stop helping once you've booked — Travelenie is built to keep going through the actual trip, handling itinerary details, bookings, and day-to-day organization as things change along the way.
- [TravelOS Synxis MCP Server](https://neotask.ai/integrations/travelos-synxis-mcp-server): Building a hotel booking tool or AI assistant that needs real availability and rate data means integrating with a hotel distribution system, and Synxis is one of the major ones behind hotel inventory worldwide.
- [Tray Headless MCP](https://neotask.ai/integrations/tray-headless-mcp): Tray's orchestration platform is where a lot of connector-based automations and internal APIs already live for teams that use it — the gap is triggering and managing those workflows from a conversation instead of the Tray console.
- [Tredict](https://neotask.ai/integrations/tredict): Building a training plan for running, cycling, or swimming and then getting it onto your watch usually means switching between a planning tool and your device's own app.
- [TRIGGERcmd](https://neotask.ai/integrations/triggercmd): TRIGGERcmd's whole purpose is running pre-configured commands on your own computer from somewhere else — your phone, a browser, or now, a chat conversation.
- [Trimble SketchUp](https://neotask.ai/integrations/trimble-sketchup): Starting a SketchUp model from a blank canvas is often the slowest part of a design — even a rough starting point saves real time.
- [Tripbtoz](https://neotask.ai/integrations/tripbtoz): Searching hotels by neighborhood rather than a fixed pin on a map — near a specific landmark, in a particular district — is the kind of search Tripbtoz is set up for, returning a simple list you can actually compare instead of a cluttered...
- [Trivago](https://neotask.ai/integrations/trivago): Trivago's whole model is comparing hotel prices across different booking providers for the same dates and location, so you're not stuck checking one site's rate in isolation.
- [Tropic](https://neotask.ai/integrations/tropic): Negotiating a software or AI vendor contract without knowing what similar companies actually pay puts procurement and finance teams at a real disadvantage.
- [True Sky](https://neotask.ai/integrations/true-sky): Astrology calculations — natal charts, transits, synastry between two charts, composite charts, solar and lunar returns — depend heavily on which house system and zodiac system you use, including Vedic and sidereal options that change the...
- [Truffle Journal](https://neotask.ai/integrations/truffle-journal): A good line from an AI conversation is easy to lose track of once the chat scrolls past it. Truffle Journal is built to save those excerpts and your own notes on them as you go, so they're there to recall later instead of buried in a chat...
- [Trustoo](https://neotask.ai/integrations/trustoo): Finding a plumber, electrician, cleaner, or photographer usually means checking several review sites separately before you're confident enough to actually call someone.
- [Tubi](https://neotask.ai/integrations/tubi): Tubi is the free, ad-supported streaming service with a deep back catalog of movies and TV shows, and Neotask can reach it as a connected app inside a chat.
- [TUI](https://neotask.ai/integrations/tui): TUI is one of Europe's largest travel and tourism groups, running package holidays, flights, and hotel stays. This connection wires TUI's own agent-facing endpoint into Neotask so a travel request can be routed there directly from chat...
- [Tuist MCP server](https://neotask.ai/integrations/tuist-mcp-server): Tuist is the project generation and workflow tool a lot of iOS and macOS teams use to keep large Xcode projects manageable: modular targets, remote caching, and generated .xcodeproj files instead of hand-maintained ones.
- [Turkish Airlines](https://neotask.ai/integrations/turkish-airlines): Turkish Airlines flies one of the largest international networks out of Istanbul, and this connection lets Neotask search fares and check flight status the way you'd use turkishairlines.com: cabin class comparisons, real-time status, and...
- [Turo](https://neotask.ai/integrations/turo): Turo is a peer-to-peer car rental marketplace, real hosts renting out their own vehicles rather than a traditional rental counter.
- [Turso MCP server](https://neotask.ai/integrations/turso-mcp-server): Turso is a distributed, SQLite-compatible database platform built for edge deployments, and this connection puts its management surface behind conversation.
- [U.S. Gold Bureau Metal Spots](https://neotask.ai/integrations/u-s-gold-bureau-metal-spots): U.S. Gold Bureau's spot-price feed tracks live pricing for gold, silver, platinum, and palladium, and this connection puts that feed inside a Neotask conversation.
- [U.S. Secret Service Careers](https://neotask.ai/integrations/u-s-secret-service-careers): This connection reaches the U.S. Secret Service careers listings, letting Neotask search and filter open federal positions by role, location, and salary instead of you paging through the agency's own jobs site.
- [Uber](https://neotask.ai/integrations/uber): Uber's rides connection gives Neotask access to real-time fare estimates and ETAs between two U.S. locations, across the actual ride tiers Uber offers: UberX, UberXL, Comfort, Black, so you can compare before you ever open the app.
- [Uber Eats](https://neotask.ai/integrations/uber-eats): This is Uber Eats' own connection for ordering: search restaurants and specific dishes near a location by cuisine, craving, or menu item, filter by rating and delivery time, and continue to Uber Eats to place the order, all starting from a...
- [Ubersuggest](https://neotask.ai/integrations/ubersuggest): Ubersuggest is Neil Patel's keyword and SEO research tool, and this connection lets Neotask pull keyword ideas, search volume, and difficulty for any topic, plus a high-level traffic and backlink overview for a site: the kind of research...
- [UCP Shopping MCP](https://neotask.ai/integrations/ucp-shopping-mcp): This connection exposes share.eu's Shopify-backed storefront through its agent-facing commerce catalog, so Neotask can browse that store's products using natural language instead of you clicking through category pages, viewing item details...
- [Unified MCP Server](https://neotask.ai/integrations/unified-mcp-server): Unified.to is an integration layer that normalizes hundreds of third-party SaaS APIs into one interface, and this connection puts that same normalized surface behind Neotask, meaning one connection can read from and write to whichever...
- [Universal Commerce Protocol MCP](https://neotask.ai/integrations/universal-commerce-protocol-mcp): College Nation sells college-themed merchandise through a Shopify storefront, and this connection exposes that store's catalog through the Universal Commerce Protocol, an agent-facing shopping interface that lets Neotask browse products...
- [Unpack AI](https://neotask.ai/integrations/unpack-ai): Unpack AI is built for the moment you're stuck on a problem and don't have a clean question yet: it works by generating questions rather than answers, clustering them into themes, and helping you land on a working hypothesis.
- [Unthread](https://neotask.ai/integrations/unthread): Unthread is a support helpdesk built around conversations, and this connection lets Neotask search across those conversations, pull up one specific ticket with its custom fields, and report on SLA and time-series performance: the kind of...
- [Updater Internet Store](https://neotask.ai/integrations/updater-internet-store): This connection lets Neotask check which internet and TV plans are available at a given address and compare them on speed, price, and terms, essentially the comparison step people usually do across multiple provider sites, done in one...
- [Upsun MCP Server](https://neotask.ai/integrations/upsun-mcp-server): Upsun is a managed multi-cloud application platform with Git-driven configuration and preview environments, and this connection lets Neotask operate deployments on it directly: checking environments, managing configuration, and pulling...
- [UptimeRobot MCP server](https://neotask.ai/integrations/uptimerobot-mcp-server): UptimeRobot monitors websites, APIs, servers, cron jobs, and SSL certificates, and this connection puts that monitoring status inside Neotask: checking what's up or down, reviewing incidents, and managing alert integrations and status...
- [upword](https://neotask.ai/integrations/upword): upword tracks how often a client gets cited across AI answer engines, ChatGPT, Google AI Overview, Perplexity, and this connection surfaces that citation performance in Neotask, alongside recent content, incoming Reddit questions, and...
- [Upwork](https://neotask.ai/integrations/upwork): This connection lets Neotask describe a project and get matched with freelance talent on Upwork, or generate a draft job post from that same description: the groundwork step before you move to Upwork itself to review candidates and hire.
- [VakantieDiscounter](https://neotask.ai/integrations/vakantiediscounter): VakantieDiscounter is a Dutch travel site specializing in discounted package holidays, from beach destinations to city trips, and this connection lets Neotask browse those deals with pricing and images, useful for spotting a genuinely low...
- [Valpas Hotels](https://neotask.ai/integrations/valpas-hotels): Valpas certifies hotels that use bed bug prevention technology, and this connection searches that certified list by city, price, and style, pulling photos, amenities, ratings, and reviews so you can pick with more than just a price filter,...
- [Vancompare.com](https://neotask.ai/integrations/vancompare-com): Van insurance shopping usually means re-typing the same van and driver details into four or five separate insurer sites just to see who is cheapest.
- [Vantage](https://neotask.ai/integrations/vantage): Cloud bills sprawl across AWS, Azure, GCP, and whatever else a team has spun up, and pulling a single answer to "what are we actually spending, and where" normally means logging into each console separately.
- [Vectoryx - Claim management](https://neotask.ai/integrations/vectoryx-claim-management): Filing a building damage claim is normally a form-and-wait process: you enter the address, describe the incident, note when it happened and how urgent it is, then have no easy way to check where things stand without logging back in.
- [VEED Fabric](https://neotask.ai/integrations/veed-fabric): Making an AI talking-head video usually means picking a character, guessing at the credit cost, generating, and then checking back repeatedly to see if it's rendered.
- [Vendr MCP Server](https://neotask.ai/integrations/vendr-mcp-server): Negotiating a software contract without pricing benchmarks means guessing whether a vendor's quote is fair. Vendr connected to Neotask gives you access to its pricing-benchmark and marketplace data directly in conversation: hand the agent...
- [Venturu](https://neotask.ai/integrations/venturu): Shopping for a business to buy means filtering listings by location, price, revenue, and industry, then separately vetting whichever broker is attached to a listing you like.
- [Viator](https://neotask.ai/integrations/viator): Planning what to do on a trip usually turns into a dozen open tabs comparing tours and attractions by date and price. Viator connected to Neotask lets you search its whole catalog of tours, attractions, and experiences by destination,...
- [Virgin Atlantic](https://neotask.ai/integrations/virgin-atlantic): Virgin Atlantic's own connection is intentionally lean: it links Neotask's agent to Virgin Atlantic so requests about the airline can be routed through this integration rather than left unhandled.
- [VisitorsCoverage](https://neotask.ai/integrations/visitorscoverage): Travel insurance shopping means comparing coverage limits, exclusions, and prices across plans that all look similar at a glance.
- [Visualping](https://neotask.ai/integrations/visualping): Manually re-checking a competitor's pricing page or a job listing for changes is tedious and easy to forget. Visualping connected to Neotask lets you just ask it to watch a page (a competitor's site, a product's price, a job posting, a...
- [Vivin](https://neotask.ai/integrations/vivin): Vivin adds a visual presence to a chat rather than a purely functional data connection: it's an animated character that reacts with expressions matching what's being discussed in the conversation.
- [VON POLL IMMOBILIEN](https://neotask.ai/integrations/von-poll-immobilien): House hunting across a real estate site usually means re-entering the same city, budget, and room-count filters every time you refine a search.
- [Vonage Documentation MCP Server](https://neotask.ai/integrations/vonage-documentation-mcp-server): Integrating with Vonage Business Cloud usually means digging through separate docs for accounts, extensions, reports, users, and the Integration Suite, then cross-referencing the CLI reference for messaging or voice calls.
- [Voyage Privé](https://neotask.ai/integrations/voyage-prive): Flash-sale travel deals move fast and are easy to miss if you're checking manually. Voyage Privé connected to Neotask lets you search and filter its private-sale hotel deals and flight-plus-hotel packages by destination, dates, budget, and...
- [VroomVroomVroom - Car Hire](https://neotask.ai/integrations/vroomvroomvroom-car-hire): Comparing rental car prices in Australia usually means checking several providers' sites separately for the same destination and dates.
- [Vybit MCP Server](https://neotask.ai/integrations/vybit-mcp-server): Vybit turns notifications into distinct sounds rather than generic pings, and connected to Neotask you can set those up conversationally: create a custom sound-based notification or reminder, trigger it on demand, share it with someone...
- [W&B MCP Server](https://neotask.ai/integrations/w-and-b-mcp-server): Checking on machine learning experiments normally means opening the Weights & Biases dashboard and clicking through runs one at a time.
- [Waldo](https://neotask.ai/integrations/waldo): Keeping a real pulse on a brand's presence (paid ads running, mentions across the web, what audiences are actually saying, what's trending in the category) is normally scattered across separate monitoring tools.
- [Waldo Strategy MCP Endpoint](https://neotask.ai/integrations/waldo-strategy-mcp-endpoint): This is Waldo's dedicated strategy endpoint, focused specifically on AI-assisted brand research and monitoring: pull brand intelligence data across social platforms, ad activity, news coverage, and audience sources through Neotask, and use...
- [Wallector](https://neotask.ai/integrations/wallector): Deciding whether a piece of art or a collectible is worth buying usually means piecing together auction results, market commentary, and expert opinion from scattered sources.
- [Wallet Whistle](https://neotask.ai/integrations/wallet-whistle): Catching a real price drop on something you're tracking usually means checking a product page over and over. WalletWhistle connected to Neotask flips that around: ask about a tracked product and it searches for the biggest recent price...
- [Walmart](https://neotask.ai/integrations/walmart): Shopping Walmart normally means leaving whatever you're doing to browse its site or app. Connected to Neotask, you can browse Walmart's catalog and shop everyday low-priced items directly inside the conversation.
- [Wanderu](https://neotask.ai/integrations/wanderu): Comparing a bus or train trip across the US often means checking Greyhound, FlixBus, and Amtrak separately for the same route.
- [Wayground](https://neotask.ai/integrations/wayground): Turning a lesson into an assignable quiz normally means manually writing questions after the teaching content is already done.
- [WeatherPromise](https://neotask.ai/integrations/weatherpromise): Trip insurance for weather risk is normally an afterthought bought blind. WeatherPromise connected to Neotask lets you enter a destination, travel dates, and the trip's cost, and get back a personalized rain-protection offer for that...
- [Webex Meetings MCP Server](https://neotask.ai/integrations/webex-meetings-mcp-server): Scheduling a Webex meeting usually means opening the Webex app, picking a time, inviting people, and later hunting down the recording.
- [Webex Messaging MCP Server](https://neotask.ai/integrations/webex-messaging-mcp-server): Cisco Webex Messaging holds a lot of team conversation history that's genuinely painful to search by hand — scrolling through spaces looking for the one message with a decision in it.
- [Webex Vidcast MCP Server](https://neotask.ai/integrations/webex-vidcast-mcp-server): Vidcast is Cisco's async video tool — record a walkthrough, share it, and let people watch on their own time instead of scheduling a meeting.
- [Webex Workspaces MCP Server](https://neotask.ai/integrations/webex-workspaces-mcp-server): Webex Workspaces covers the physical side of Cisco Webex — the room devices, screens, and shared spaces in an office, plus how much they actually get used.
- [Webiny remote docs MCP server](https://neotask.ai/integrations/webiny-remote-docs-mcp-server): Webiny is a serverless CMS and application framework you deploy into your own AWS account, which means its docs cover a lot of ground — the admin app, the headless CMS, extension tooling, and how it all wires into AWS.
- [Wednesday.app](https://neotask.ai/integrations/wednesday-app): Finding the right creator for a campaign usually means scrolling through platforms manually and eyeballing follower counts and recent activity.
- [Wego](https://neotask.ai/integrations/wego): Wego is a travel search platform used across flights and hotels, and this connection is a thin integration point — a way for Neotask's agent to reach into Wego rather than a fully documented single-purpose tool.
- [WeightWatchers](https://neotask.ai/integrations/weightwatchers): This connection points at WeightWatchers' AI recipe endpoint, which is built specifically around GLP-1 medication users — people whose appetite and food tolerance have changed and who need recipes tailored to that.
- [Wesur](https://neotask.ai/integrations/wesur): Getting an insurance quote usually means filling out a long form before you even know if the price is in the right ballpark.
- [Whack](https://neotask.ai/integrations/whack): Property listing sites show asking prices, but what a home actually sold for is a different, more useful number — and it's usually locked away in registry data most buyers never check.
- [Whimsical](https://neotask.ai/integrations/whimsical): Turning a half-formed idea into a diagram people can actually look at usually means opening a whiteboard tool and dragging shapes around yourself.
- [Wikiloc](https://neotask.ai/integrations/wikiloc): Wikiloc's real value is its scale — millions of community-logged trails with waypoints and photos, which makes it a genuinely useful source for trip planning if you can search it well.
- [Windmill MCP server](https://neotask.ai/integrations/windmill-mcp-server): Windmill is an open-source workflow engine people use to run scripts, flows, and internal tools, often self-hosted or on their own cloud instance — which means managing it usually means jumping into its own UI.
- [Wix Site MCP](https://neotask.ai/integrations/wix-site-mcp): This connection runs against Perseuss, a company whose Wix-hosted site documents its cartonization, palletization, and freight packing optimization software — the kind of tooling warehouses and shippers use to figure out how to pack...
- [Wix Site MCP for berlin-group.org](https://neotask.ai/integrations/wix-site-mcp-for-berlin-group-org): The Berlin Group is a European standards body behind technical frameworks like NextGenPSD2 and openFinance, which underpin a lot of open banking and payments interoperability work across Europe.
- [Workable marketing site WebMCP](https://neotask.ai/integrations/workable-marketing-site-webmcp): This connection is a WebMCP integration into Workable's marketing site itself, rather than the recruiting platform's product features — a way for Neotask's agent to reach into that site directly when you ask something about Workable.
- [WorkOS Management MCP Server](https://neotask.ai/integrations/workos-management-mcp-server): WorkOS handles the enterprise authentication plumbing a lot of B2B apps rely on — SSO connections, directory sync, user management, and the audit logs that go with it.
- [Wyndham Hotels & Resorts](https://neotask.ai/integrations/wyndham-hotels-and-resorts): Booking a hotel usually means filtering through a search UI by hand — city, dates, then scrolling to check which properties actually allow pets or have a pool.
- [Yahoo! JAPAN Shopping](https://neotask.ai/integrations/yahoo-japan-shopping): Yahoo! JAPAN Shopping's catalog spans everything from daily necessities to fashion and electronics, and this connection lets Neotask's agent search across it directly and check out using a PayPay balance — the payment method a lot of...
- [Yango](https://neotask.ai/integrations/yango): Before you commit to a ride, it helps to know roughly what it'll cost and how long the car will take to arrive — and Yango gives you both without opening the app.
- [Yapay Zeka'nın Boyner'i](https://neotask.ai/integrations/yapay-zeka-n-n-boyner-i): Yapay Zeka'nın Boyner'i is Boyner's own AI shopping connection — Boyner being one of Turkey's well-known fashion and department-store retailers — and this connection is the general-purpose bridge into it.
- [Yelp](https://neotask.ai/integrations/yelp): Booking a restaurant reservation usually means checking a few sites to see who's actually got a table free tonight. With Yelp connected, Neotask's agent finds participating restaurants across the US and Canada and books an available...
- [YepCode](https://neotask.ai/integrations/yepcode): YepCode turns a block of Node.js or Python into something an agent can actually call: you define a reusable process, describe its inputs as a JSON Schema, and YepCode exposes it as a callable tool that runs in an isolated sandbox with npm...
- [Yoshi](https://neotask.ai/integrations/yoshi): Yoshi links your financial accounts — banking, investing, credit — into one place so balances, transactions, holdings, and liabilities are queryable in plain language instead of five separate app logins.
- [Zen Shopping](https://neotask.ai/integrations/zen-shopping): Zen Shopping is built for the moment you know roughly what you want but not where to buy it cheapest: describe the item and it compares prices across stores, surfaces recommendations tailored to what you described, and finds active deals.
- [Zeno MCP server](https://neotask.ai/integrations/zeno-mcp-server): Zeno is an internet radio and audio-streaming service, and this MCP server is the direct line Neotask uses to reach it as a connected app rather than a website you'd have to open yourself.
- [Zep Documentation MCP Server](https://neotask.ai/integrations/zep-documentation-mcp-server): Zep is a context-engineering platform that stores user, thread, and business context in temporal knowledge graphs, and this MCP server puts Zep's documentation directly in reach of Neotask's agent — including docs for Graphiti, Zep's...
- [Zillow](https://neotask.ai/integrations/zillow): Zillow gives you the national real estate listing data — homes for sale or rent, filtered by price, beds, baths, and specifics like parking or pet-friendliness — and Neotask's connection to it means you can search that inventory from a...
- [ZipRecruiter](https://neotask.ai/integrations/ziprecruiter): ZipRecruiter's job board is enormous, and searching it well usually means juggling filters for title, company, location, salary, distance, remote/hybrid, and employment type across multiple sessions.
- [Zocks](https://neotask.ai/integrations/zocks): Zocks is built for financial advisors managing a book of clients: it holds client contacts and meeting history, past and upcoming session details and transcripts, and AI-generated insights on client goals, concerns, and planning...
- [Zola](https://neotask.ai/integrations/zola): Zola is a wedding-planning and registry platform, and this connection puts it directly within reach of Neotask's agent so wedding-related tasks — registry items, planning details, vendor coordination — don't require jumping into Zola's own...
- [Zomato](https://neotask.ai/integrations/zomato): Zomato handles food delivery and restaurant discovery, and connecting it to Neotask means you can order from a restaurant, track where an order is, or browse menus by cuisine or craving without opening the Zomato app yourself.
- [ZOZO](https://neotask.ai/integrations/zozo): ZOZO is a fashion retail platform, and this integration connects Neotask's agent directly to it so a shopping request involving ZOZO's catalog can be handled inside the conversation instead of you browsing the ZOZO site yourself.
- [Zumba](https://neotask.ai/integrations/zumba): Zumba runs classes across formats — in person, livestream, and video on demand — and finding the right one usually means checking a schedule against a location or a specific fitness goal.
- [Zumper](https://neotask.ai/integrations/zumper): Zumper covers apartment, house, and condo rentals across the US and Canada, with filters for amenities like laundry, parking, or pet policy, plus local rent-comparison data to sanity-check whether an asking price is fair.
- [Zuora MCP Server (remote)](https://neotask.ai/integrations/zuora-mcp-server-remote): Zuora is subscription-billing infrastructure — orders, invoices, revenue workflows — and this remote MCP server puts that directly in reach of Neotask's agent, alongside access to Zephr's entitlement and digital subscription management...
- [ZZAIM](https://neotask.ai/integrations/zzaim): ZZAIM is aimed squarely at teachers working from Korea's official K-12 curriculum standards: it lets you search those standards by subject and grade, then design aligned lesson plans, worksheets, and slides in the same conversation.

## Documentation Text



### [Account Access](https://neotask.ai/docs/account-access)

## Check License And Account Status

1. Open the account menu.
2. Select **Settings**.
3. Open **License** to confirm the active plan and account.
4. Open **Models** when the issue concerns a connected subscription or provider.

![License plan, status, seat count, and expiration](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/license-plan-1280.webp)

Use this guide when you cannot sign in, cannot access the correct workspace, are unsure whether you count as a member, or need billing-related access help.

---

## Guest vs Member Support

### You are a guest if:

- you have not finished signing up
- your payment did not complete
- you do not yet have an active workspace
- your phone number is not yet linked to a member or employee record

### You are a member if:

- you can sign in to an active workspace
- you are an owner, admin, or viewer on a tenant
- you are an employee record tied to a company support flow
- your phone number is already saved as an approved member support route

If support cannot safely confirm membership from your calling context, the issue should stay in guest support until the workspace can be verified.

---

## Common Access Problems

### I cannot sign in

Check these first:

1. Confirm you are using the same email address that owns or was invited to the workspace.
2. Complete any Google or external sign-in step in the browser window if prompted.
3. Refresh the app or web session after sign-in completes.
4. If you were invited recently, confirm the invite was accepted on the same email address.

### I can sign in, but I do not see the right workspace

Possible causes:

- you signed in with the wrong email
- the workspace owner has not added you yet
- your membership changed and the app needs a refresh

Try:

1. Sign out and back in with the expected email.
2. Ask the workspace owner to confirm your role.
3. Refresh the app after the role change is saved.

### I am an employee, but support does not recognize my number

If your employee record exists but your phone number changed, support may need to verify the company another way before discussing member-only details. Updating the saved employee phone number restores direct member routing on future calls.

---

## Billing Access Problems

### Payment failed during sign-up

Use guest support for:

- card declines
- checkout interruptions
- pricing questions before activation
- a charge attempt that did not create an active workspace

### I paid, but the plan did not update

Try these steps:

1. Refresh the app or dashboard.
2. Wait a few minutes for billing state to sync.
3. Confirm the charge completed in the billing portal or card history.
4. If the workspace still has the old state, contact support with the billing email and approximate charge time.

### I need invoice or billing follow-up

Be ready with:

- workspace email
- billing email if different
- the plan you expected
- the approximate charge date

---

## Membership Roles

Tenant membership is generally organized as:

- owner
- admin
- viewer

If you can sign in but lack permission to change billing, channels, or support-sensitive settings, the issue may be role-related rather than a broken account.

---

## When Support Needs Escalation

Support should escalate instead of guessing when:

- a payment succeeded but no workspace was created
- a member can authenticate but has no valid tenant access
- an employee support record is missing or clearly stale
- a billing or access change appears to have partially synced

For product-setup help after access is restored, continue with [MCP Auth & OAuth Setup](./mcp-auth-oauth.md) or [App Setup Troubleshooting](./app-setup-troubleshooting.md).

---

### [Agents](https://neotask.ai/docs/agents)

## Start And Inspect Agent Work

1. Open **Agents**.
2. Select the agent you want to use.
3. Open **Chat** and describe the result you need.
4. Expand activity to see which apps and tools the agent used.
5. Open **Overview** for scheduled work, session totals, and recent activity.

![Main agent chat with app activity visible](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/chat-main-1280.webp)

## What Are Agents?

Agents are AI-powered assistants that can think, act, and communicate on your behalf. Unlike simple chatbots, Neotask agents are autonomous entities that reason through problems, execute tasks, and interact across multiple platforms.

Open **Overview** when you need scheduled work, session totals, and recent activity in one place.

![Agent overview with activity and scheduled work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

Key characteristics of agents:

- **Identity:** Each agent has a unique name, emoji, and personality that distinguishes it from others.
- **Workspace:** Every agent operates within its own workspace, keeping files, sessions, and context separate.
- **Skills:** Agents can be equipped with modular skills that extend their capabilities (see [Skills](skills.md)).
- **Channels:** Agents connect to communication platforms like Slack, Discord, and email to send and receive messages (see [Channels](channels.md)).
- **Independence:** Multiple agents can work independently on different tasks, or collaborate as a team to tackle complex workflows.

### Agent Capacity by Plan

| Plan | Concurrent Agents |
|------|-------------------|
| **Free** | Main chat agent only; custom agent creation is locked |
| **Individual** | Unlimited agents with full platform access |
| **Business** | Unlimited agents, plus company operations and employee seats |
| **Enterprise** | Unlimited agents with multi-company orchestration at scale |

On the Free plan, the main supported agent surface is interactive chat, boards, and coding workflows within the Free quota: first `10` messages/events, then `5` messages/events per user-local calendar day. Free users can connect apps and provider keys, but they cannot add custom agents or run autonomous company agents. On every paid plan, agent count is not capped: Individual gives one person the full platform, while Business and Enterprise add company operations and large-scale parallel orchestration, all coordinated through the platform's team orchestration engine. See [Billing & Plans](billing.md) and [Free Plan](free-plan.md) for details on how usage works.

## Creating an Agent

Creating additional custom agents requires a paid plan. Free users can use the main chat agent, boards, and coding subpage, but the Create Agent flow is locked behind upgrade.

To create a new agent:

1. Navigate to **Agents** in the dashboard.
2. Click **Create Agent**.
3. Provide the following:
   - **Name:** A descriptive name for the agent (e.g., "Support Bot", "DevOps Runner").
   - **Workspace:** Select an existing workspace or create a new one. The workspace determines where the agent stores files and maintains state.
   - **Emoji:** Choose an emoji to visually identify the agent across the dashboard.

### Agent Templates

Neotask provides pre-configured templates to get you started quickly. Each template comes with recommended skills, model settings, and system prompts tailored to a specific use case:

| Template | Description |
|----------|-------------|
| **Assistant** | General-purpose helper for everyday tasks, Q&A, and productivity |
| **Developer** | Code generation, debugging, repository management, and CI/CD workflows |
| **Researcher** | Deep research, summarization, and information gathering across the web |
| **Social Media** | Content creation, scheduling, and engagement across social platforms |
| **Data Analyst** | Data processing, visualization, and reporting |
| **Home Automation** | Smart home device control, routines, and monitoring |
| **Voice Assistant** | Speech-to-text and text-to-speech conversational workflows |
| **Tutor** | Educational content delivery, quizzes, and adaptive learning |
| **Campaign Manager** | Marketing campaign planning, execution, and analytics |

You can always customize a template after creation or start from scratch with a blank agent.

## Agent Configuration

### Model Override

By default, agents use the system-wide model configured in your Neotask settings. You can override this on a per-agent basis to assign a specific AI model.

- Open the agent's **Settings** tab.
- Under **Model**, select a provider and model from the dropdown.
- If set to **System Default**, the agent will follow whatever model is configured globally.

This is useful when certain agents need the highest reasoning tier (e.g., Claude Opus 4.7 or GPT-5.5 Pro for architecture tasks) while others can use a lower-cost frontier model such as Claude Sonnet 4.6 or GPT-5.4 for routine formatting and drafting.

### Spend Limits

Control costs by setting a daily spending cap per agent.

- Navigate to the agent's **Settings** tab.
- Under **Spend Limit**, enter a maximum daily cost in USD.
- When the agent reaches its limit, it will stop making API calls for the remainder of the day.
- Spend limits reset at midnight UTC.

Auto-enforcement ensures that no single agent can run up unexpected costs, which is especially important when running hundreds of agents on Enterprise or when agents run autonomously on schedules.

### Safe Mode

Safe mode runs the agent inside an isolated execution sandbox. When enabled:

- File system access is restricted to the agent's workspace directory.
- Network requests are limited to explicitly allowed domains.
- Shell commands execute in a contained environment.
- The agent cannot affect other agents or system-level resources.

Safe mode is recommended for agents that run untrusted code, interact with external APIs, or operate with minimal supervision.

## Agent Sessions

Each agent maintains conversation sessions that track the full interaction history.

### Session Types

| Type | Behavior |
|------|----------|
| **Main** | A persistent session that carries forward across interactions. The agent remembers prior context and can reference earlier messages. |
| **Isolated** | A fresh session created each time. No prior context is retained. Useful for stateless tasks or when you want the agent to start clean. |

### Chat History

Every session records a complete chat history with clearly labeled message roles:

- **User:** Messages sent by you or received from a connected channel.
- **Assistant:** Responses generated by the agent.
- **System:** System-level instructions, prompts, or configuration messages.
- **Tool:** Results returned from skill and tool executions.

### Thinking Mode

Enable thinking mode to see the AI's internal reasoning process before it produces a response. This is helpful for:

- Debugging unexpected agent behavior.
- Understanding how the agent arrives at a decision.
- Auditing complex multi-step reasoning chains.

When thinking mode is active, the agent's reasoning steps appear in a collapsible section above each response.

### Attachments

Agents can process attachments within a session:

- **Images:** Screenshots, photos, diagrams. The agent can analyze and describe visual content.
- **Files:** Documents, spreadsheets, PDFs. The agent can read, summarize, and extract data.
- **Code:** Source files and snippets. The agent can review, edit, and execute code.

## Boards & Live Artifacts

Agents also support **Boards**, which are live artifact pages built through chat and then saved as reusable dashboards.

Use an agent board when you want a custom operating surface tied to one agent’s own:

- workspace files
- memory
- MCP tools
- attached integrations

Boards are different from normal chat because the goal is not just to get one answer. The goal is to create a durable live page that you can reopen, refresh, publish, and activate live.

See [Boards & Live Artifacts](boards-live-artifacts.md) for the full guide.

## Agent Tools

Agents come equipped with a set of built-in tools organized by category. These tools are always available and do not require installation.

### File System

| Tool | Description |
|------|-------------|
| **Read** | Read the contents of a file from the agent's workspace |
| **Write** | Create or overwrite a file in the agent's workspace |
| **Edit** | Make targeted edits to an existing file using find-and-replace |

### Execution

| Tool | Description |
|------|-------------|
| **Shell Commands** | Execute shell commands (bash, zsh, etc.) within the agent's environment |
| **Process Management** | Start, stop, and monitor running processes |
| **Subprocess Agents** | Spawn child agents to handle subtasks, enabling parallel and delegated workflows |

### Web

| Tool | Description |
|------|-------------|
| **Browser Automation** | Control a headless browser to navigate pages, fill forms, click elements, and capture screenshots |
| **HTTP Requests** | Make HTTP/HTTPS requests to APIs and web services with full control over method, headers, and body |

### Scheduling

| Tool | Description |
|------|-------------|
| **Cron Jobs** | Schedule recurring tasks using cron expressions (e.g., run a report every Monday at 9 AM) |
| **Webhooks** | Register webhook endpoints that trigger agent actions when called by external services |

## Agent Models

Neotask supports multiple AI providers and models. You can assign any supported model to any agent.

### Supported Providers and Models

**Anthropic**

| Model | Best For |
|-------|----------|
| Claude Opus 4.7 | Maximum reasoning depth for architecture, planning, and critical decisions |
| Claude Sonnet 4.6 | Frontier coding, analysis, documentation, and day-to-day agent work |

**OpenAI**

| Model | Best For |
|-------|----------|
| GPT-5.5 / GPT-5.5 Pro | Frontier reasoning, complex planning, and high-accuracy generation |
| GPT-5.4 | General frontier work across coding, analysis, and drafting |
| GPT-5.3 Codex | Coding-focused implementation and multi-file changes |

**Google**

| Model | Best For |
|-------|----------|
| Gemini 3 Pro | Frontier long-context analysis when enabled for the workspace |

**OpenRouter**

| Model | Best For |
|-------|----------|
| Frontier model access | Workspace-approved frontier models routed through OpenRouter when configured |

### Smart Task Routing

When an agent is configured to use smart routing, Neotask automatically selects the best frontier model for each task based on complexity and type. Routing no longer uses legacy light/medium models such as Claude Haiku, DeepSeek, GPT-4.1 Nano, or Gemini Flash Lite. The routing table below shows the current frontier-only assignments:

| Task Type | Model | Tier |
|-----------|-------|------|
| Architecture and critical decisions | Claude Opus 4.7 or GPT-5.5 Pro | Frontier |
| Complex Coding | GPT-5.5, GPT-5.4, or GPT-5.3 Codex | Frontier |
| Complex Analysis | Claude Opus 4.7, Claude Sonnet 4.6, or GPT-5.5 | Frontier |
| Standard Coding | Claude Sonnet 4.6 or GPT-5.4 | Frontier |
| Code Review | Claude Sonnet 4.6, GPT-5.4, or GPT-5.3 Codex | Frontier |
| Testing | GPT-5.4 or Claude Sonnet 4.6 | Frontier |
| Research | Claude Opus 4.7, Claude Sonnet 4.6, or GPT-5.5 | Frontier |
| Documentation | Claude Sonnet 4.6 or GPT-5.4 | Frontier |
| Data Extraction | GPT-5.4 or Claude Sonnet 4.6 | Frontier |
| Formatting | GPT-5+ or Claude Sonnet 4.6, using the lightest approved frontier path | Frontier |
| Summarization | GPT-5+ or Claude Sonnet 4.6, using the lightest approved frontier path | Frontier |
| Simple Tasks | GPT-5+ or Claude Sonnet 4.6, using the lightest approved frontier path | Frontier |
| Boilerplate | GPT-5.3 Codex, GPT-5.4, or Claude Sonnet 4.6 | Frontier |

**Tier definitions:**

- **Frontier:** Tasks run on the approved frontier pool only: Claude Sonnet 4.6 level and up, Claude Opus 4.7, GPT-5 level and up, and approved Codex GPT-5 variants.
- **Cost control:** For simpler work, Neotask chooses the least expensive approved frontier route rather than falling back to non-frontier models.

Smart routing helps optimize your usage costs by matching the right frontier model to the right task automatically. This is especially valuable on Enterprise plans where hundreds of agents may be running concurrently.

---

### [Neotask API Reference](https://neotask.ai/docs/api-reference)

Complete API reference for the Neotask platform. This document covers every available endpoint, authentication methods, error handling, and rate limits.

---

## Authentication

All API requests require authentication via one of the following methods:

- **Bearer JWT**: Pass the header `Authorization: Bearer <token>`. Tokens are obtained from the login endpoints described below.
- **License HMAC**: Device-bound HMAC-SHA256 signing used by the Neotask desktop application. The client signs requests with a secret derived from the license key and device fingerprint.
- **Tenant Header**: Include `x-tenant-id: <tenantId>` for any tenant-scoped endpoint. This header identifies which tenant (workspace) the request applies to.

Many endpoints combine multiple authentication requirements. For example, "Tenant + Admin" means the request must include both a valid JWT (belonging to a user with the Admin role) and the `x-tenant-id` header.

---

## Auth Endpoints

Endpoints for user authentication across web, iOS, and desktop platforms.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/auth/login` | Authenticate with email and password. Returns a JWT on success. | None |
| POST | `/api/auth/google` | Authenticate via Google OAuth on the web. Expects a Google authorization code. | None |
| POST | `/api/auth/google-ios` | Authenticate via Google OAuth on iOS. Expects an iOS-specific authorization payload. | None |
| POST | `/api/auth/google-token` | Verify a Google access token and return an Neotask JWT. | None |
| GET | `/api/auth/apple/start/` | Start the web Sign in with Apple redirect flow. Accepts `mode=signin\|onboarding` and optional `next`. | None |
| POST | `/api/auth/apple/callback` | Sign in with Apple form-post callback. Exchanges the authorization code and creates a short-lived result ticket. | None |
| POST | `/api/auth/apple/result/` | Consumes a short-lived Apple result ticket and returns the Neotask JWT/session payload. | None |
| POST | `/api/auth/apple/notifications` | Apple server-to-server notification receiver for email relay and consent/account changes. | Apple JWS |
| POST | `/api/auth/apple` | Direct-token Apple Sign-In compatibility endpoint. Expects an Apple identity token. | None |
| GET | `/api/auth/me` | Return the authenticated user's profile (name, email, avatar, roles). | JWT |
| DELETE | `/api/auth/account` | Permanently delete the authenticated user's account and all associated data. | JWT |

---

## Dashboard Auth Endpoints

Authentication endpoints specific to the Neotask dashboard, including license-key login and two-factor authentication (TOTP).

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/dashboard/auth/login` | Authenticate with a license key. Returns a JWT on success, or signals that TOTP is required. | None |
| POST | `/api/dashboard/auth/totp-login` | Complete login with the purchaser email plus a TOTP code or backup code when two-factor authentication is enabled. | None |
| GET | `/api/dashboard/auth/me` | Return the current license information and feature entitlements for the authenticated dashboard user. | JWT |
| POST | `/api/dashboard/totp/setup` | Generate a TOTP secret and QR code for setting up two-factor authentication. | JWT |
| POST | `/api/dashboard/totp/verify` | Verify a TOTP code to confirm and enable two-factor authentication on the account. | JWT |
| POST | `/api/dashboard/totp/recovery-codes/regenerate` | Replace all recovery codes with eight one-time v2 codes after confirming a live TOTP code. | JWT |

---

## Billing Endpoints

Manage subscriptions, checkout sessions, and in-app purchases through Stripe and RevenueCat.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/billing/checkout` | Create a Stripe Checkout session for the Pro plan. Returns a checkout URL. | Tenant + Admin |
| POST | `/api/billing/checkout/enterprise` | Create a Stripe Checkout session for the Enterprise plan. Returns a checkout URL. | Tenant + Admin |
| GET | `/api/billing/status` | Return the current plan, message count, and usage limits for the tenant. | Tenant + Viewer |
| POST | `/api/billing/portal` | Create a Stripe Customer Portal session for managing payment methods and invoices. | Tenant + Admin |
| POST | `/api/billing/agent-addon` | Purchase an additional agent slot add-on for the tenant. | Tenant + Admin |
| POST | `/api/billing/webhook` | Stripe webhook handler. Validates the Stripe signature and processes subscription lifecycle events. | Stripe Signature |
| POST | `/api/billing/revenuecat-sync` | Verify an iOS in-app purchase via RevenueCat and sync entitlements to the tenant. | Tenant + Admin |
| POST | `/api/billing/revenuecat-webhook` | RevenueCat webhook handler. Processes subscription events originating from the App Store. | Bearer |

---

## Chat Endpoints

Send and receive messages, manage sessions, and poll for asynchronous job results.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/chat/init` | Auto-provision a tenant, agent, and session for the authenticated user. Use this to bootstrap a new user's chat environment in a single call. | JWT |
| POST | `/api/chat/send` | Send a message to the agent. The message is processed asynchronously; a job ID is returned immediately. | Tenant + Admin |
| GET | `/api/chat/jobs/:jobId` | Poll the status and result of an asynchronous chat job by its ID. | Tenant + Viewer |
| GET | `/api/chat/messages` | Fetch all messages for a given session. Pass the session ID as a query parameter. | Tenant + Viewer |
| GET | `/api/chat/sessions` | List all chat sessions for the tenant. | Tenant + Viewer |
| POST | `/api/chat/sessions` | Create a new chat session. | Tenant + Admin |

Every `POST /api/chat/send` caller must send an `Idempotency-Key` header containing 8-200 printable ASCII characters without spaces and reuse that exact key when retrying the same turn. A missing key returns `428 IDEMPOTENCY_KEY_REQUIRED`; reusing a key with different session, company scope, message, or attachments returns `409 IDEMPOTENCY_KEY_REUSED`. A retry of an accepted request returns the original `202` job and user-message identity with `Idempotency-Replayed: true`.

---

## Agent and Activity Endpoints

Browse agents, sessions, tools, files, skills, channels, scheduled jobs, and teams associated with the tenant.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET | `/api/activity/agents` | List all agents belonging to the tenant. | Tenant |
| GET | `/api/activity/sessions` | List sessions. Supports filtering by `agentId`, `status`, and date range query parameters. | Tenant |
| GET | `/api/activity/sessions/:id/entries` | Get all entries (messages and events) for a specific session. | Tenant |
| GET | `/api/activity/tools` | List all tools available to the tenant's agents. | Tenant |
| GET | `/api/activity/files` | List all files associated with the tenant's agents. | Tenant |
| GET | `/api/activity/skills` | List all skills associated with the tenant's agents. | Tenant |
| GET | `/api/activity/channels` | List all channels associated with the tenant's agents. | Tenant |
| GET | `/api/activity/cron-jobs` | List all scheduled (cron) jobs for the tenant. | Tenant |
| GET | `/api/activity/teams` | List all teams within the tenant. | Tenant |

---

## Skills Endpoints

Manage the skill catalog, install and configure skills, and publish custom
skills. The legacy metadata-only custom registration endpoints are retired;
use the [durable custom skills and plugins flow](neotask-desktop/custom-skills-and-plugins.md)
so Neotask can verify source bytes and bind the intended target.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET | `/api/skills/catalog` | Browse the full skill catalog. Returns all published skills available for installation. | Tenant |
| GET | `/api/skills/list` | List the skills currently installed by the user. | Tenant + Viewer |
| POST | `/api/skills/sync` | Sync skills from the filesystem. Use this after deploying new skill files to refresh the registry. | Tenant + Admin |
| POST | `/api/skills/register` | Retired. Returns `410 legacy_skill_registration_retired`; use Create Skill or a supported desktop import. | Tenant + Admin |
| DELETE | `/api/skills/register/:skillId` | Retired. Returns `410 legacy_skill_registration_retired` and preserves legacy metadata for audit and explicit re-import. | Tenant + Admin |
| POST | `/api/skills/:skillId/publish` | Publish a skill to make it available in the catalog. | Tenant + Admin |
| POST | `/api/skills/:skillId/unpublish` | Unpublish a skill, removing it from the catalog. | Tenant + Admin |
| GET | `/api/skills/all` | List all skills including unpublished ones. Admin-only endpoint for skill management. | Tenant + Admin |
| PUT | `/api/skills/:skillId/env-schema` | Define or update the environment variable schema for a skill. This controls which configuration fields are presented to users. | Tenant + Admin |
| POST | `/api/skills/:skillId/configure` | Save configuration values for an installed skill. | Tenant + Viewer |
| DELETE | `/api/skills/:skillId/configure` | Remove all configuration for a skill, resetting it to defaults. | Tenant + Viewer |

---

## Channel Endpoints

Link, configure, enable, and disconnect messaging channels (e.g., Slack, Discord, WhatsApp).

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET | `/api/channels/` | List all channels and their current statuses. | Tenant + Viewer |
| GET | `/api/channels/:channel/status` | Get the detailed status of a specific channel. | Tenant + Viewer |
| POST | `/api/channels/:channel/link` | Start the linking process for a channel. Returns a link URL or session token depending on the channel type. | Tenant + Admin |
| GET | `/api/channels/:channel/link/wait` | Long-poll for the completion of a channel linking flow. Returns once the channel is successfully linked or the operation times out. | Tenant + Admin |
| GET | `/api/channels/:channel/jobs/:jobId` | Get the result of a channel-specific asynchronous job. | Tenant + Admin |
| POST | `/api/channels/:channel/enable` | Enable a linked channel so the agent begins receiving messages from it. | Tenant + Admin |
| POST | `/api/channels/:channel/disable` | Disable a channel without unlinking it. The agent stops receiving messages but the channel connection is preserved. | Tenant + Admin |
| POST | `/api/channels/:channel/disconnect` | Fully disconnect and unlink a channel. | Tenant + Admin |
| GET | `/api/channels/:channel/config` | Get the current configuration for a channel. | Tenant + Admin |
| POST | `/api/channels/:channel/config` | Update the configuration for a channel. | Tenant + Admin |

---

## Google OAuth Endpoints

Initiate and manage Google OAuth flows for connecting Google services to Neotask.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET/POST | `/oauth/google/start` | Initiate a Google OAuth flow. Accepts optional parameters to specify which Google services to request access to. | Optional |
| GET | `/oauth/google/services` | List all Google service definitions supported by Neotask, grouped by category. | None |
| GET | `/oauth/google/callback` | OAuth callback handler. Google redirects here after the user grants or denies access. | None |
| GET | `/oauth/google/status` | Poll the status of an in-progress OAuth flow. Returns whether the flow is pending, completed, or failed. | None |

### Available Google Services

Neotask supports connecting to 25 Google services, organized by category:

| Category | Services |
|----------|----------|
| **Core** | Gmail, Calendar, Drive, Docs, Sheets, Slides, Forms, Keep, Tasks |
| **Communication** | People/Contacts, Chat, Meet, YouTube, Photos |
| **Location** | Places API, Routes/Directions, Business Profile |
| **Specialized** | Classroom, Play Developer, AdSense, Google Ads |

---

## Provider Keys Endpoints

Manage third-party API keys (e.g., for LLM providers). Keys are encrypted at rest and can be resolved (decrypted) on demand.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| PUT | `/api/provider-keys/:provider` | Save or update an API key for the specified provider. The key is encrypted before storage. | HMAC + Auth |
| GET | `/api/provider-keys/` | List all saved provider keys. Key values are masked (only the last four characters are shown). | Auth |
| DELETE | `/api/provider-keys/:provider` | Remove a saved API key for the specified provider. | HMAC + Auth |
| GET | `/api/provider-keys/:provider/resolve` | Decrypt and return the full API key for the specified provider. | Auth |
| GET | `/api/provider-keys/mode` | Get the current key mode (user-provided vs. platform credits) and the remaining credit balance. | Auth |

### Allowed Providers

The `:provider` path parameter must be one of the following values:

- `anthropic`
- `openai`
- `openai-codex`
- `openrouter`
- `google-ai`

---

## Usage Endpoints

Track usage, query analytics, monitor spending, and manage budgets.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/usage/ingest` | Ingest a usage data record (e.g., token counts, model used). | Auth |
| GET | `/api/usage/query` | Query usage analytics with flexible filters (date range, model, agent). | Auth |
| GET | `/api/usage/summary` | Get a usage summary for a given period. Pass `period` as a query parameter (e.g., `day`, `week`, `month`). | Auth |
| GET | `/api/usage/analytics` | Retrieve detailed usage analytics including breakdowns by model and agent. | Auth |
| GET | `/api/usage/spending` | Get spending data for a given period. Pass `period` as a query parameter. | Auth |
| GET | `/api/usage/budget` | Get budget information including the total limit, amount used, and amount remaining. | Auth |
| POST | `/api/usage/cron-runs` | Record the execution of a scheduled (cron) job for usage tracking. | Auth |
| POST | `/api/usage/budget-snapshot` | Capture a point-in-time snapshot of the current budget state. | Auth |

---

## Onboarding Endpoints

Guide new users through initial setup, including selecting a model provider and configuring an API key.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/onboarding/init` | Initialize the onboarding configuration for the authenticated user. Creates default settings and returns the onboarding state. | JWT |
| GET | `/api/onboarding/models` | List available model providers and their supported models. This endpoint is public and does not require authentication. | None |
| POST | `/api/onboarding/setup` | Configure a provider API key during onboarding. Validates the key before saving. | Tenant + Admin |
| POST | `/api/onboarding/complete` | Mark onboarding as complete for the current user. | Dashboard Auth |

---

## Tenant Endpoints

Create and manage tenants (workspaces), store encrypted secrets, and create sessions and jobs directly.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/tenants/` | Create a new tenant (workspace). Returns the tenant ID and default settings. | JWT |
| GET | `/api/tenants/:tenantId` | Get details for a specific tenant, including settings and member count. | JWT |
| POST | `/api/tenants/secrets` | Store an encrypted secret associated with the tenant. Used for service integrations. | Tenant + Admin |
| POST | `/api/tenants/sessions` | Create a new session within the tenant. | Tenant + Admin |
| POST | `/api/tenants/jobs` | Create a new asynchronous job within the tenant. | Tenant + Admin |

---

## Memory Endpoints

Manage agent memory configuration, files, and content. Memory allows agents to persist information across sessions.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET | `/api/memory/config` | Fetch the current memory configuration (enabled state, retention policy, limits). | Auth |
| POST | `/api/memory/config` | Update memory settings such as retention period and storage limits. | Auth |
| GET | `/api/memory/files` | List all memory files stored for the agent. | Auth |
| GET | `/api/memory/files/:filename` | Retrieve the contents of a specific memory file by filename. | Auth |
| DELETE | `/api/memory/files/:filename` | Delete a specific memory file by filename. | Auth |
| GET | `/api/memory/content` | Retrieve stored memory content. | Auth |
| POST | `/api/memory/content` | Store new memory content for the agent. | Auth |

---

## Google Accounts Endpoints

Manage Google accounts that have been connected to Neotask via OAuth.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| GET | `/api/google-accounts/` | List all connected Google accounts with their service scopes and status. | Auth |
| POST | `/api/google-accounts/:accountId/activate` | Activate a connected Google account, enabling the agent to use its authorized services. | Auth |
| DELETE | `/api/google-accounts/:accountId` | Remove a connected Google account and revoke its stored tokens. | Auth |

---

## Contact and Support Endpoints

Submit and manage support contact requests.

| Method | Endpoint | Description | Auth |
|--------|----------|-------------|------|
| POST | `/api/contact/` | Submit a contact form message. Rate limited to 5 requests per 15 minutes per IP address. | None |
| GET | `/api/contacts/` | List all contact submissions with pagination support. Admin-only. | Admin |
| PATCH | `/api/contacts/:id` | Update the status of a contact submission (e.g., mark as resolved). | Admin |
| DELETE | `/api/contacts/:id` | Delete a contact submission. | Admin |

---

## Rate Limits

Neotask enforces rate limits on certain endpoints to protect service stability. When a rate limit is exceeded, the API returns a `429 Too Many Requests` response.

| Endpoint Category | Limit |
|-------------------|-------|
| Contact form | 5 requests per 15 minutes per IP |
| Login attempts | 10 requests per 15 minutes per IP |
| Tracking/analytics | 30 requests per 60 seconds per IP |

---

## Error Responses

All error responses follow a consistent JSON format:

```json
{
  "error": "Error message description",
  "status": 400
}
```

### Status Codes

| Code | Meaning |
|------|---------|
| 400 | **Bad Request** -- The request was malformed or contained invalid parameters. |
| 401 | **Unauthorized** -- Authentication is missing or the provided token is invalid. |
| 402 | **Payment Required** -- The tenant's credits have been depleted. Upgrade your plan or add credits to continue. |
| 403 | **Forbidden** -- The authenticated user does not have sufficient permissions for this action. |
| 404 | **Not Found** -- The requested resource does not exist. |
| 429 | **Rate Limited** -- Too many requests. Wait and retry after the period indicated in the response headers. |
| 500 | **Internal Server Error** -- An unexpected error occurred on the server. If this persists, contact Neotask support. |

---

### [App Setup Troubleshooting](https://neotask.ai/docs/app-setup-troubleshooting)

Use this guide when Neotask is installed but setup is blocked by the app, gateway, voice, channels, automations, or app connections.

---

## First Triage

Before diving into a specific app or workflow, identify which layer is failing:

- desktop app install or first launch
- gateway or local runtime
- app connection or auth
- channels and message delivery
- automation and scheduled tasks
- voice and microphone setup

---

## Desktop App Setup

### The app will not launch

Check:

1. The installer finished successfully.
2. You are opening the installed app, not the disk image.
3. Your operating system did not block the app after download.

If the app opens and then immediately fails, move to the gateway section below.

### The app opens, but onboarding is blocked

Common causes:

- license or sign-in issue
- no active internet connection
- a required browser auth step was never completed
- the gateway did not finish starting

Start with:

1. confirm your workspace access
2. confirm your license or plan state
3. confirm the gateway is healthy

---

## Gateway and Local Runtime

### The gateway failed to start

This usually means the local runtime did not initialize correctly.

1. Fully quit and reopen Neotask once.
2. Open **Settings → Infrastructure → Connection**.
3. Check the **Local gateway** stage and use **Start** or **Restart** under **Gateway controls**.
4. Open **Settings → Infrastructure → Health** and click **Run check**.
5. Review **Settings → Advanced → System report** for the first live system check that needs attention.
6. Open **Settings → Advanced → Troubleshoot** and click **Check system** if the issue remains.
7. Read the result or open **Show technical output**.
8. Click **Apply repairs** when it becomes available.
9. Wait for the follow-up check before retrying setup.

![Gateway connection path and Gateway controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-connection-1280.webp)

The Health screen checks runtime responsiveness, workload, connected channels, and recent stability.

![Runtime Health with responsiveness, workload, channels, and stability checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-health-full-1280.webp)

When Connection and Health identify an installation problem, continue to the Doctor for the repair step.

![Troubleshoot settings used to check and repair the Gateway](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

### The gateway disconnects after startup

Possible causes:

- a local runtime crash
- a conflicting local service
- security or permission issues on the device
- stale local state after an update

If tasks, voice, or app tools all stop at once, gateway health is the first thing to verify.

1. Refresh **Settings → Infrastructure → Connection**.
2. Run **Settings → Infrastructure → Health**.
3. Review **Settings → Advanced → System report**.
4. Use **Settings → Advanced → Troubleshoot** when a repair is needed.

---

## App Connection Problems

Open **Apps** from the agent workspace and find the affected service. The catalog shows which apps are connected and which ones are still available to add.

![Connected and available apps in the agent workspace](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

### The app stays on Pending

Possible causes:

- the browser OAuth flow never finished
- the provider callback never returned cleanly
- the app is waiting on credentials or validation

Fixes:

1. re-open the provider and finish setup again
2. verify the callback URL and scopes
3. confirm the app does not also require a custom instance URL

Use the searchable picker only when the service has not been added yet.

![Searchable app picker for adding a service](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

### The app shows Error

This usually means:

- bad credentials
- wrong provider app configuration
- wrong instance URL
- token refresh failure

Go to [MCP Auth & OAuth Setup](./mcp-auth-oauth.md) and re-run the setup path for that provider.

### The app looks connected, but tasks still fail

This often means the saved auth state and the runtime state disagree.

Capture:

- the provider name
- whether the app says connected, expired, or error
- whether the failure happens in the main agent, a tenant, or a company workflow
- the exact task error

---

## Channels and Delivery

### A channel is linked, but messages do not arrive

Check:

1. the channel account is still connected
2. the target account, room, or thread is correct
3. permissions or scopes were not removed
4. the channel is enabled for the right workspace or company flow

### Voice or phone delivery is not reaching the right place

Check:

1. the saved phone number is correct
2. the route is tied to the right tenant or company
3. the caller is using the same recognized number expected by the workspace

If member-only routing fails because the number is not recognized, support should fall back to guest support or manual verification rather than guessing.

---

## Automation and Scheduled Tasks

### A scheduled task did not run

Check:

1. the schedule is still enabled
2. the task has the required apps connected
3. the task is not blocked on auth or approval
4. the delivery route still exists

### A task ran, but nothing was delivered

Possible causes:

- the task succeeded internally but had no valid delivery target
- the target channel or route was disconnected
- the task was blocked by a missing app auth state

---

## Voice and Microphone

### The microphone is not detected

Check:

1. operating-system microphone permissions
2. the selected input device
3. whether another app is holding the microphone

### Voice activation does not trigger

Check:

1. wake mode configuration
2. the selected shortcut or wake phrase
3. whether microphone permissions were granted

---

## When Support Should Escalate

Escalate instead of repeating generic troubleshooting when:

- the same provider fails after a correct reconnect
- billing is active but account access is still broken
- a task appears ready but company execution still reports missing auth
- member caller recognition is clearly wrong for a saved employee or member route
- the app state and the runtime state contradict each other

Related guides:

- [Support & Routing](./support.md)
- [Account Access](./account-access.md)
- [MCP Auth & OAuth Setup](./mcp-auth-oauth.md)
- [Troubleshooting](./troubleshooting.md)

---

### [Auto, Autonomous Companies](https://neotask.ai/docs/auto-companies)

## See The Operating Loop

The company workspace keeps the goal, plan, schedule, active run, and result connected. Use these screens to move from company context to the work happening now.

1. Open **Goals** to review active outcomes.
2. Open **Plan** to inspect the generated work and schedule.
3. Open **Overview** for live work and completed results.
4. Open a run to inspect its source, apps, and decisions.

![Company Plan showing generated work and the active schedule](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v6/2026-08-11-r1/ui/active-plan-1280.webp)

Auto gives you a company workspace inside Neotask where planning, approvals, files, integrations, content work, routing, and ongoing autonomous execution all live together.

This page is the starting point. If you want the full walkthroughs, use the related docs at the bottom of this page.

---

## What Auto Is

An Auto company workspace lets you:

- onboard a real business or a new idea
- generate and refine a company plan
- rehearse one task at a time in **Train** before trusting it live
- connect apps and Google Workspace at the company level
- review approvals before sensitive work runs
- upload SOPs and reference material
- run recurring company tasks on a heartbeat
- manage company chat, files, channels, employees, memory, SEO, and usage in one place

Auto is designed for company operations, not just personal task automation.

---

## Who Can Use Auto

Company workspaces require a plan that includes company access.

In the current product, that means:

- **Business** for one active company workspace
- **Enterprise** for larger company coverage and additional expansion

If your current plan does not include company workspaces, Neotask shows an upgrade flow instead of letting you create the company.

For current plan details, see [Billing & Plans](./billing.md).

---

## Where Auto Lives

Open the desktop app and click **Auto** in the main navigation.

From there you can:

- create a new company
- reopen an existing company
- return to the company dashboard
- move from the dashboard into one company’s full detail view

---

## The Core Auto Flow

Most companies go through the same broad lifecycle:

1. **Create the company**
   Start from an existing website, a manual description, or an AI brainstorm.
2. **Build the profile**
   Neotask fills in the company profile, business context, and initial operating direction.
3. **Connect knowledge and tools**
   Add SOPs, choose apps, and connect Google Workspace where needed.
4. **Set autonomy and heartbeat**
   Decide how much freedom the company has and how often it should check for work.
5. **Review the generated plan**
   Approve the starting plan, then keep refining it from the company dashboard.
6. **Complete setup in the company dashboard**
   Finalize tasks, integrations, channels, approvals, and any remaining setup items.
7. **Run the company**
   Use the company controls to start, pause, resume, or trigger the company heartbeat.

The **Goals** view connects active outcomes to the work created for them.

![Company goals with progress and linked work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/loop-goal-1280.webp)

The **Plan** view records the generated work before it reaches the schedule.

![Generated company plan](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/loop-plan-1280.webp)

---

## What You Can Manage After Setup

Once a company exists, the company dashboard becomes the main operating surface.

Open a live run to see its current step, connected apps, and progress.

![Active company run](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/loop-run-1280.webp)

Completed runs keep the result with the work that produced it.

![Completed company run with delivery status, source checks, owner notification, cost, and timing](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r7/company-completed-run-detail-1280.webp)

The current dashboard includes these tab areas:

- **Overview**
- **Activity**
- **Chat**
- **SOPs**
- **Approvals**
- **Plan**
- **Tasks**
- **Train**
- **Files**
- **Boards**
- **SEO & Blogs**
- **Apps**
- **Employees**
- **Profile**
- **Integrations**
- **Channels**
- **Skills**
- **Memory**
- **Usage**
- **Settings**

Outside the tab bar, the same company view also includes:

- a left-side controls rail
- quick stats
- a mini calendar
- a **Nightly Rundown** action

---

## What Auto Is Best At

Auto works best when you want one place to manage:

- recurring operational work
- approvals and safety controls
- company-scoped app access
- company-scoped Google access
- task scheduling and task refinement
- reference files and generated files
- SEO and content operations
- employee and channel routing
- usage visibility and operational status

---

## What To Read Next

- [Auto, Onboarding & Plan Setup](./auto-companies-onboarding.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Training Mode](./auto-companies-training.md)
- [Auto, Operations, Controls & Settings](./auto-companies-operations-controls-and-settings.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)
- [Desktop App](./desktop-app.md)
- [Billing & Plans](./billing.md)

---

### [Auto, Approvals & Safety](https://neotask.ai/docs/auto-companies-approvals-and-safety)

Approvals are the control layer that keeps a company useful without forcing every action to run unattended.

This guide explains the current approval experience from a user point of view.

---

## What The Approvals Surface Does

The **Approvals** tab serves two jobs:

- it shows the items currently waiting on you
- it lets you decide how much future work should run automatically

That means the tab is both:

- an approval inbox
- a policy control surface

---

## The Three Main Approval Modes

The company approval settings currently use three mode cards:

### Fully Supervised

Use this when you want the most control.

This mode is best when:

- the company is new
- the plan is still changing often
- your team wants to review everything closely

### Semi-Supervised

Use this when you want a mix of automation and manual review.

This mode is best when:

- the company is active but still needs oversight
- you want routine work to move faster
- you still want stricter control over more sensitive actions

### Fully Autonomous

Use this when you want the least approval friction.

This mode is best when:

- the company is stable
- the plan is well-shaped
- the connected tools and routing are already trustworthy
- you are comfortable tightening or loosening specific app rules afterward

If you are unsure, start with **Semi-Supervised** and tighten from there.

---

## What You Can Review In The Queue

The queue is where you review pending items that need an answer before the company can continue.

![Company approval queue showing pending actions and their owners](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/approval-queue-1920.webp)

That normally includes:

- approval cards waiting for approve or deny
- higher-friction situations that need a more explicit response
- items tied to one company, not your whole workspace

If a company seems stuck, the **Approvals** tab is one of the first places to check.

Open an item to read the requested action, the app it will use, the owner, and the review window before choosing **Approve** or **Deny**.

![Approval detail showing the proposed action and review information](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/approval-detail-1280.webp)

---

## Per-App Rules

The current approval surface also supports per-app behavior, not just one company-wide mode.

That means you can keep the company in one overall mode, then make a specific app:

- follow the default
- auto-approve all
- require all
- use custom rules

This is useful when one company depends on a tool you trust for routine work, but you still want stricter handling everywhere else.

---

## Action-Level Overrides

Inside expanded app rows, the current approval UI can go narrower than the whole app.

That lets you keep:

- one app mostly automatic
- but one action inside that app more strict

or the reverse:

- one app mostly strict
- but a few repetitive actions more automatic

If a company feels over-blocked or under-protected, this is usually where to refine it.

---

## Where Approval Behavior Also Shows Up Elsewhere

Approval behavior is not limited to one tab.

You will also feel it in:

- the company controls rail, through the company autonomy mode
- task editing, where one task can carry stricter expectations than another
- training, where meaningful rehearsal checkpoints and preview approvals can still ask for guidance before a live task is trusted
- channel and employee routing, when the company needs a real person involved
- nightly review and follow-up flows, where unresolved items surface again

So the clean operating pattern is:

- use the controls rail to set the broad level
- use **Approvals** to tune the real rule behavior

---

## When To Tighten The Company

Move toward **Fully Supervised** or stricter per-app rules when:

- the company is new
- sensitive actions are being prepared
- your routes or channels are not fully ready
- you are still validating the quality of the plan

---

## When To Loosen The Company

Move toward **Semi-Supervised** or **Fully Autonomous** when:

- routine tasks are stable
- connected apps are correct
- employees and channels are set up
- the company is producing the outcomes you expect without frequent corrections

---

## Best Practices

- Start new companies more conservatively than mature ones.
- Use per-app rules before flipping the whole company into the loosest mode.
- Re-check approvals after major plan edits, new channels, or new app connections.
- Treat **Approvals**, **Tasks**, **Channels**, and **Employees** as one operating system, not four unrelated tabs.

---

## Training Preview Approvals

Training mode can still surface approval-style moments so you can verify how a task would behave before you trust it live.

That is useful when you want to rehearse:

- outbound outreach
- browser-based submission flows
- CRM updates
- channel delivery steps
- other sensitive actions that still need human judgment

The important difference is that training is for rehearsal and guidance. Use it to shape the task’s behavior before loosening live approval rules.

---

## Related Docs

- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Training Mode](./auto-companies-training.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)

---

### [Auto, Apps, Integrations & Files](https://neotask.ai/docs/auto-companies-apps-integrations-and-files)

This guide covers the company-level connection and knowledge surfaces that users most often confuse:

- **Apps**
- **Integrations**
- **SOPs**
- **Files**
- related company data tabs such as **Channels**, **Employees**, **Profile**, **Skills**, **Memory**, and **Usage**

---

## The Core Split

The company dashboard intentionally separates connection surfaces.

### Apps

Use **Apps** for company-scoped app connections and app recommendations.

This is the right place for app-style providers and company app groups.

### Integrations

Use **Integrations** for company-scoped Google Workspace services and similar built-in company integrations.

This is the right place for:

- Gmail
- Google Calendar
- Google Docs
- Google Sheets
- Google Drive

### SOPs

Use **SOPs** for operating guidance and reference material that teaches the company how you want it to work.

### Files

Use **Files** for the documents and artifacts the company already has or creates over time.

---

## Apps: What You Manage There

The current **Apps** tab is the company’s main app-management surface.

![Company Apps page showing the connected app catalog](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/apps-catalog-1920.webp)

Use it to:

- review recommended apps for the company
- connect or reconnect company-scoped apps
- organize app groups
- reuse those app groups when you assign tools to tasks

This is where app connections belong when the company needs a provider like a business tool, sales tool, support tool, or workspace tool that is not part of the built-in Google integration lane.

---

## Integrations: What You Manage There

The **Integrations** tab is the company’s Google Workspace surface.

Use it to:

- connect company email access
- connect company calendar access
- connect company document access
- reconnect an existing Google account for that company
- choose whether the company should use one already connected account or add a new one

If a company plan depends on Google tools, this is the place to make sure they are attached to the company correctly.

![Company Integrations page showing connected Google services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/google-services-1920.webp)

---

## SOPs: What Belongs There

The **SOPs** tab is for the reference material that defines how the company should operate.

Typical examples:

- playbooks
- internal procedures
- checklists
- brand or process guidance
- operational reference docs

This is the best place to upload the knowledge you want the company to follow consistently.

Training mode can reuse that company guidance, but the files created during task training are reviewed inside **Train**, not managed as general SOP uploads.

### How SOPs Get Created And Kept Current

SOPs are origin-agnostic. You can upload or write one yourself (including dictating one with **Create with Voice**), and an agent can also propose an SOP based on how the company works. Anything an agent proposes arrives as a pending draft that a person reviews, approve, edit, or discard, before it becomes active guidance, with a diff to show what changed. For the full review and publish flow, see [Auto, SOP Authoring & Review](./auto-companies-sops-authoring.md).

### How Agents Pick Which SOPs To Follow

You do not assign SOPs to tasks. When an agent starts work, the company automatically attaches the SOPs most relevant to that task, matched by meaning, scoped to your company, and updated as soon as you publish. For how that works, see [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md).

---

## Files: What Belongs There

The **Files** tab is for the broader company file library.

The current tab supports:

![Company Files page showing folders and delivered work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/company-files-1920.webp)

- daily view
- filesystem view
- search
- sort and filter controls
- preview
- download

This is where users usually go when they want to review:

- generated files
- uploaded documents
- company file history
- website-related files
- reference exports produced during onboarding or later runtime work

This is different from the training bundle review. Training creates task-specific guidance files for one task run, while **Files** is the broader company file library.

---

## Managed Website Support In Files

If a company is using a generated or managed website, the **Files** tab becomes even more important.

That surface can include website-related state such as:

- preview and live website links
- current website status
- domain state
- website file state
- website source editing and related website management controls
- status/error visibility and save/redeploy actions

So for companies that need a website, **Files** is not just a document browser. It becomes part of the website operations surface too.

---

## Other Related Company Data Tabs

These tabs complete the same operating system even though they are not the primary connection surfaces:

### Employees

Use **Employees** for real people tied to the company and for the routing details that help the company know who to involve.

### Channels

Use **Channels** for the delivery endpoints the company can use, including messaging routes and company call-routing surfaces.

### Profile

Use **Profile** for company identity, website details, logo, company stage, and other core profile fields.

### Skills

Use **Skills** for the company’s skill-management surface.

### Memory

Use **Memory** for long-lived company context, including boot memory, categories, semantic memory indicators, and generated memory files.

This is where the user refines persistent company context rather than only reviewing one live chat or one task.

### Usage

Use **Usage** for company-level cost and activity measurement over time.

---

## Quick Decision Rule

Use this rule when you are not sure where to go:

- if you are connecting a company app or reviewing company app recommendations, use **Apps**
- if you are connecting Gmail, Calendar, Docs, Sheets, or Drive for one company, use **Integrations**
- if you are uploading operating guidance, use **SOPs**
- if you are rehearsing one task and reviewing the guidance files it generated, use **Train**
- if you are trying to preview, download, search, or review the company’s files and generated artifacts, use **Files**

---

## Related Docs

- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Onboarding & Plan Setup](./auto-companies-onboarding.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Training Mode](./auto-companies-training.md)
- [Auto, Operations, Controls & Settings](./auto-companies-operations-controls-and-settings.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, SOP Authoring & Review](./auto-companies-sops-authoring.md)
- [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md)

---

### [Auto, Company Dashboard](https://neotask.ai/docs/auto-companies-dashboard)

Once a company is created, almost all day-to-day work happens in that company’s dashboard.

This guide explains the current dashboard layout and what each tab is for.

---

## The Two Main Views

Auto has two main company views:

- the **company dashboard list**, where you see all companies
- the **company detail view**, where you manage one company in depth

The company list helps you compare companies quickly. The company detail view is where you actually operate one company.

---

## Company Cards In The Dashboard List

Each company card shows a summary of the company at a glance, including:

- the company name
- current status
- task count
- file count
- approvals waiting
- latest activity

Depending on the company state, the label can show things such as:

- **Setup**
- **Plan Pending**
- **Finalizing**
- **Plan Setup**
- **Ready**
- **Running**
- **Paused**
- **Error**

Open a card to move into the full company dashboard.

---

## The Company Detail Layout

The open company view is a three-part operating surface:

![Company Mission Control showing live runs and company-wide operating metrics](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/mission-control-1920.webp)

### Left Side: Controls

The controls rail is where you manage the company runtime.

This area includes:

- start, pause, resume, stop, and trigger-now controls
- current autonomy tier
- heartbeat settings
- daily budget controls
- sub-agent auto-approve
- last heartbeat and quick runtime stats

If the plan still needs setup, this area also shows the notice that the plan setup must be finished before autonomy can start.

### Center: Tabs

The center area is the main company workspace. This is where the tab bar lives.

### Right Side: Quick Context

The right side shows quick stats and the mini calendar. On the **Overview** tab, this area also works together with the overview widgets.

At the top of the company tab area, you can also open **Nightly Rundown**.

The surrounding company shell also includes shared surfaces such as the notification bell, the floating company chat widget, widget customization overlays, activity detail overlays, and tutorial overlays.

---

## The Current Company Tabs

The current company dashboard includes these tabs:

### Overview

Use **Overview** for the company’s high-level operating picture.

This tab includes:

- a time-window toggle
- overview stat cards
- customizable widgets
- widget pinning
- add-widget and customize-layout controls
- activity detail drill-ins from widgets
- quick navigation into deeper tabs

Overview is the best first stop when you want to understand what is going on without reading the full activity feed.

### Activity

Use **Activity** for recent operational events.

This tab is built for:

- live company activity
- recent company activity
- clearer event readability
- faster troubleshooting before you open a deeper surface

### Chat

Use **Chat** when you want to talk directly inside the company workspace.

This is the company’s manual orchestration chat surface and it is separate from the general Agents chat area.

### SOPs

Use **SOPs** for the company’s uploaded playbooks and guidance documents.

This is where you manage operating instructions that help the company behave the right way.

### Approvals

Use **Approvals** for both:

- the current approval queue
- the company’s approval settings

This is where you review pending approval items, requires-action items, and the policy that determines how much can run automatically.

### Plan

Use **Plan** for the company’s current operating plan.

This tab is the main place to:

- review the approved plan
- inspect the schedule view
- refine tasks and requirements
- keep completing setup after onboarding

### Train

Use **Train** to rehearse one task until it becomes reliable enough to trust live.

This tab is built for:

- opening one real task at a time in a training workspace
- answering preflight questions before the first run
- responding to meaningful checkpoints during the run
- coaching the task in the training input
- reviewing the generated bundle files and diffs
- promoting the current working revision with **Bundle It**
- running **Train Again** to verify the task stays in the improved shape

### Tasks

Use **Tasks** when you want task-by-task management instead of the broader plan view.

This tab is built for:

- opening one task at a time
- reviewing runtime status
- editing assigned apps, integrations, channels, and employees
- adjusting task-level settings

### Files

Use **Files** for the company’s generated files and uploaded files.

This tab includes:

- search
- sorting and filtering
- daily view
- filesystem view
- preview, download, and file actions
- managed website preview and live links
- website status and last-error visibility
- custom-domain management
- website source editing and save/redeploy actions

If the company is using a generated website, this tab also becomes the main place to review managed website files and related website state.

### SEO & Blogs

Use **SEO & Blogs** for content work inside the company.

This tab includes:

- **Headlines**
- **Articles**
- **Usage**
- SEO settings

It is the company’s main content planning and publishing-prep surface.

### Boards

Use **Boards** for the company’s custom dashboard and insight-board surfaces.

This is where the company can move from the standard overview into more tailored board views when that company needs a more specific operating layout.

Boards are best when the company needs:

- a custom live page for one decision or workflow
- a multi-widget artifact instead of one generic overview widget set
- a live page built from connected apps, company files, and internal company context

See [Boards & Live Artifacts](./boards-live-artifacts.md) for the full guide.

### Apps

Use **Apps** for company-scoped app access.

This tab is where you manage:

- company app recommendations
- company app connections
- company app groups
- task assignment flows tied to those app groups

### Employees

Use **Employees** for real people tied to the company workspace.

This is where you manage:

- employee records
- contact details
- routing notes
- employee-channel relationships

### Profile

Use **Profile** for the company’s identity and core business information.

This tab is where you:

- edit profile details
- manage company stage
- manage build/manage mode
- upload or remove the logo
- re-run website analysis when needed

### Integrations

Use **Integrations** for company-scoped Google Workspace access and related built-in services.

This tab is separate from **Apps**.

Use it when the company needs:

- Gmail
- Google Calendar
- Google Docs
- Google Sheets
- Google Drive

This is also where the user can decide whether to reuse an already connected Google account for the company or add a different one.

### Channels

Use **Channels** for message and delivery routing.

This tab is where you:

- add channels
- edit channel bindings
- review phone-call routes
- manage delivery targets tied to the company

### Skills

Use **Skills** for the company’s skills surface.

This tab uses the same core skill-management experience used elsewhere in the app, but in the context of the open company.

### Memory

Use **Memory** for longer-lived company context.

This tab includes:

- boot memory
- memory categories
- semantic memory status
- auto-generated memory files
- generated memory file preview/open flows

### Usage

Use **Usage** for company-specific usage and cost review.

This tab includes:

- time-window switching
- summary cards
- charted usage history
- counter breakdowns

### Settings

Use **Settings** for company-level configuration panels that apply across the whole workspace.

This tab includes:

- heartbeat settings
- approval settings
- adaptive learning
- notification preferences
- public live settings

It is the right place to update company-wide behavior without editing the plan itself.

---

## Nightly Rundown

The **Nightly Rundown** button lives above the tab content, not inside a normal tab.

Use it for a company summary that helps you:

- review what happened
- see open issues
- jump into the right tab next

It is especially useful when the user wants one compact review of the company instead of opening several tabs separately.

---

## Important Shared Surfaces Outside The Tabs

Some company features are not cleanly “one tab only.”

The current company dashboard also includes:

- the left controls rail
- the right-side quick stats and mini calendar
- **Nightly Rundown**
- the notification bell
- the floating company chat widget
- overview widget customization overlays
- activity detail modals

That means users sometimes describe a company feature by where it appears on the screen rather than by a tab name.

---

## A Good Operating Pattern

For most companies, the cleanest routine is:

1. check **Overview**
2. open **Activity** if something looks off
3. clear **Approvals**
4. refine **Plan** or **Tasks**
5. use **Apps**, **Integrations**, **Channels**, and **Employees** to finish missing operating dependencies
6. use **Files**, **SEO & Blogs**, **Memory**, and **Usage** for the company’s longer-running operating trail

---

## Related Docs

- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Onboarding & Plan Setup](./auto-companies-onboarding.md)
- [Auto, Operations, Controls & Settings](./auto-companies-operations-controls-and-settings.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)

---

### [Auto, Leads & Prospecting](https://neotask.ai/docs/auto-companies-leads)

Leads is where a company builds its own prospect lists. You describe the kind of company or person you want to reach in plain English, Neotask researches them, and the result is a structured, reviewable table saved to that company.

Every list belongs to one company. Leads captured for one company are never visible to another, and never leave the tenant that captured them.

This guide covers the first run, day-to-day searching, reading and exporting a list, how billing works, and how your autonomous agents can use the same system.

---

## Where Leads Lives

Open a company, then select the **Leads** tab.

What you see depends on whether that company has used Leads before:

- a company that has never run a search opens the **first-run setup**
- a company that has opens the **Leads workspace**

Leads is per company. Turning it on for one company does nothing to the others.

---

## First Run: Setting Up Lead Finding

The first time you open Leads for a company, Neotask offers to set up lead finding for it. This is a full-screen flow, and it exists so your very first search already knows who the company is and who it sells to.

The setup moves through these stages:

1. **Preparing**: Neotask reads what the company already knows about itself: its profile, its plan, its goals, and anything captured during company onboarding.
2. **Review**: you see the prospect profile it drafted: what the company sells, who it sells to, and the kind of buyer worth targeting. Everything here is editable. Correct it now and every later search starts from the corrected version.
3. **Running**: your first search runs immediately using that profile.
4. **Complete**: the workspace opens with your first lead table in it.

### You Can Leave

Once the first search starts, it runs on Neotask servers, not in the window. You can close the page, switch companies, or quit the app. The search keeps going and the results are waiting when you return.

Because of that, there is no cancel button during the first search. The action offered is **Run in background**, which simply lets you leave.

### Correcting The Profile Later

The profile is not frozen after setup. Say what changed in the chat and the next search uses the correction:

> We stopped selling to agencies. Focus on in-house marketing teams at B2B SaaS companies.

---

## The Leads Workspace

The workspace has three panels.

### 1. Conversation rail (left)

Every lead conversation for this company, newest first. Select one to reopen it with its full history and the lists it produced.

Starting a new conversation does **not** start the company over. The company keeps one lead workspace, so a new conversation reuses the same prospect profile and the same captured history. Ask for something different, not something disconnected.

The rail collapses when you want the table full width.

### 2. Results pane (middle)

Five views, described in detail below.

### 3. Chat (right)

Where you describe what you want. The message box is always available, so you can queue a refinement while a search is still running.

Above the input, suggestion chips offer useful next actions based on what the company already has. They float above the composer and never push the conversation around.

---

## The Five Views

### Overview

Orientation for the whole company:

- **Searches**: how many lead searches this company has run
- **Lead tables**: how many captured result tables it holds
- **Prospect rows**: total rows across every table
- **Unique leads**: rows after duplicates are collapsed
- **Recent searches**: the latest runs and their outcomes
- **How lead finding is billed**: the current allowance and what a search draws from

### Searches

Every search this company has run, with its brief and its state:

| State | Meaning |
|---|---|
| **Running** | working now; results appear as they are captured |
| **Completed** | finished and captured to a lead table |
| **Succeeded** | finished and returned results |
| **Failed** | stopped on an error; nothing captured |
| **Cancelled** | stopped by you or an agent |

Select a search to see its brief, its runs, and the table it produced.

### Prospects

Every captured lead for the company in one place, across all lists. This is the view to use when you want to answer "do we already have this person?" before spending anything on a new search.

### Lead tables

The captured result tables, newest first. One finished search produces one table.

### Runs

The individual execution attempts behind each search, including continuations. Useful when a search was refined several times and you want to see what each pass added.

---

## Running A Search

Describe what you want in plain English. Specific briefs produce better tables than broad ones.

**Weak brief**

> Find me some leads

**Strong briefs**

> Find heads of marketing at B2B SaaS companies in the US with 50 to 200 employees that recently raised a Series A.

> Find operations managers at independent dental practices in Texas with more than three locations.

> Find agency owners in the UK who publish case studies about paid social and have a team page listing at least five people.

If a brief is too vague to run well, Neotask tightens it with you first rather than spending on a weak search.

### Refining Instead Of Restarting

To adjust a search, continue it rather than starting a new one. Continuing keeps the rows already captured and builds on them.

> Same list, but only companies with in-house engineering teams.

> Add a verified work email for each of these rows.

> Drop anyone without a LinkedIn profile, then find twenty more like the rest.

Starting fresh when you meant to refine discards context and costs more.

### While A Search Runs

- Results appear as they are captured; you do not wait for the whole run
- You can leave the page, the company, or the app
- You can queue the next instruction in chat
- You can stop a running search from the Searches view

---

## Reading A Lead Table

Lead tables use the same table as Tasks and Goals: the same sorting, the same column behavior, the same horizontal scrolling for wide tables. Scroll sideways to reach every column; nothing is hidden.

**Click any row to open the full lead detail.** The table shows a summary. The detail panel shows everything captured for that lead, including fields too long for a cell.

### Columns You May See

Research returns whatever it finds, so tables vary. Neotask recognizes common fields and renders them appropriately:

| Group | Fields |
|---|---|
| Person | name, title, photo |
| Company | company, website, about |
| Contact | email |
| Social | linkedin, twitter, facebook, crunchbase |
| Assessment | fit, signal, evidence, confidence, why, status |

Anything unrecognized still appears as a normal column. Nothing is dropped for not matching a known shape.

### Duplicates

If a lead already exists in another list for this company, the row is flagged with an information icon showing where else it appears.

**Duplicates are flagged, never removed.** A row stays in the list it was captured in, because a list is a record of what a specific search found. Removing rows would silently rewrite that record.

When you want a clean file, use **Export CSV without duplicates**.

---

## Exporting

Two options from a lead table:

- **Export CSV**: downloads the table. Column order and column labels are exactly as stored, so the file matches what you reviewed.
- **Export to Company Files**: saves the table into that company's Files, where agents, tasks, and other company work can use it.

Use Company Files when the list should feed further automation. Use CSV when it is going to a person or another tool.

---

## Billing

**The first 25 leads each month are included.** After that, delivered leads draw from your Neotask credit balance.

The Overview panel always shows the current position: how many are included, how many are used, how many remain, and when the allowance resets. Trust that panel over any number written here, because the allowance is set on our servers and can change without an app update.

Two details worth knowing:

**The allowance is included leads, not a cap.** Running past it does not stop lead finding. It bills.

**Lead finding always uses Neotask credits.** It runs on Neotask servers, so it uses Neotask credits in every AI mode, including when your assistant is running on your own sign-in or your own API key. Bringing your own model key changes where your *assistant* runs. It does not change where lead research runs.

A search that returns nothing still represents work performed, so usage is metered on the work rather than counted per row. Cost per lead is therefore an average that moves with how hard a brief is to satisfy. Tight, well-specified briefs are cheaper per useful lead than broad ones.

---

## Using Leads With Agents

Leads is fully available to autonomous agents and company orchestrators, per company and per item. An agent can check what the company already holds, run a search, refine it, and read the results, all inside the same governed system you use.

### Free reads

These never spend. They only read this company's own records:

| Tool | Purpose |
|---|---|
| `leads_overview` | orientation: counts and recent searches |
| `leads_prospect_search` | search already-captured leads across every list |
| `leads_lists` | list the captured lead tables |
| `leads_list_read` | read one captured table |
| `leads_search_status` | check searches, poll a running one |

### Spending actions

These do real work and **require human approval every time**:

| Tool | Purpose |
|---|---|
| `leads_search_start` | start a brand new search |
| `leads_search_continue` | continue an existing search: refine or enrich it |

`leads_search_cancel` stops a running search. Rows already captured are kept.

The read tools and the spending tools are separate by design, so an agent can look before it spends and can never reach a spending action through a read.

### Expected agent behavior

A well-behaved agent checks stored records before starting anything new, tightens a vague brief before running it, and continues an existing search instead of starting a second one for the same goal. The built-in lead-finding skill enforces exactly that sequence.

### Automating It

Because the tools are available to the planner, lead finding can run as scheduled or triggered work:

> Every Monday, check our lead lists and find twenty new prospects matching our best-fit profile, then add them to Company Files.

The spend approval still applies. A scheduled search asks before it spends.

---

## What Leads Does Not Do

Leads builds and enriches lists. It is deliberately not an outreach tool.

- It does not send email, messages, or connection requests
- It does not publish anything or write public copy
- A captured list is an internal record for your team to review

Export the list and use your own outreach tooling to contact anyone on it.

---

## Common Questions

**Does a new conversation start the company over?**
No. One company keeps one lead workspace, one prospect profile, and one captured history. A new conversation reuses all of it.

**Can I close the app while a search runs?**
Yes. Searches run on Neotask servers. Close the page or quit the app and the results are waiting when you return.

**Why is a lead in two lists?**
Because two searches both found it. It is flagged, not removed, so each list stays a true record of what that search found. Use **Export CSV without duplicates** for a clean file.

**Why did a search return fewer rows than I asked for?**
The brief was narrower than the available data. Broaden one constraint at a time, continuing the same search rather than starting over.

**Does my own API key make lead finding free?**
No. Lead finding runs on Neotask servers and uses Neotask credits in every AI mode.

**Can I change the monthly included amount?**
It is set on our servers, not in the app. The Overview panel always shows the current figure.

---

## Related

- [Auto, Company Dashboard](auto-companies-dashboard.md): where the Leads tab lives
- [Auto, Apps, Integrations & Files](auto-companies-apps-integrations-and-files.md): where exported lists land
- [Auto, Approvals & Safety](auto-companies-approvals-and-safety.md): how spend approvals work
- [Billing](billing.md): credits and balances
- [Usage Rates](usage-rates.md): how usage is metered

---

### [Auto, Onboarding & Plan Setup](https://neotask.ai/docs/auto-companies-onboarding)

## Follow The Company Setup Flow

1. Open **Companies** and select **Add company**.
2. Enter the company website or choose the manual setup path.
3. Review the website analysis and generated company profile.
4. Check the proposed plan, apps, channels, and budget.
5. Finish setup when the ready screen confirms the company workspace was created.

This guide explains how company onboarding works today and what happens before a new company is ready to run.

---

## Before You Start

To create a company workspace, you need a plan that includes company access.

Free users may be able to walk through company onboarding or preview the generated plan, but Free mode does not allow the company to actually run tasks, automations, heartbeats, or company operations. After onboarding, a Free account remains gated on the company Plan/upgrade flow until it has at least an Individual subscription.

You should also know which of these you already have:

- a live company website
- SOPs or internal reference files
- the apps your company needs
- Google Workspace accounts you want connected at the company level

If you do not have all of that yet, you can still continue. The wizard supports partial setup and the company dashboard lets you finish the rest later.

---

## Starting The Wizard

Open **Auto** and choose the new company action from the dashboard.

The first onboarding choice is whether this is:

- an **existing company**
- a **new company**

That decision changes the early steps of the wizard.

---

## Existing Company Path

Use the existing company path when you already have a real business site or real operating context.

This path starts with:

1. **Company**
2. **Website**

In the website step, you enter the company URL and Neotask analyzes the site to build the initial company profile.

![Company website entry during onboarding](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/origin-website-entry-1280.webp)

The current website flow can also validate the site and guide the user through sitemap-related issues before continuing.

![Website analysis and company research](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/origin-analysis-1280.webp)

This is the fastest path when the company already has a public site and clear operating context.

---

## New Company Paths

If you choose a new company, the wizard asks how you want to begin.

Current options are:

- **Description**
  Best when you already know what the company should do and want to describe it directly.
- **Brainstorm**
  Best when the company is still being shaped and you want to talk it through with AI first.

### Description Path

The description path asks for a rough description of the company or product, then turns that into the company profile.

Review the generated profile before continuing. Correct the company description or operating context here so the first plan starts from accurate information.

![Generated company profile ready for review](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/origin-profile-1280.webp)

This path also supports the current **Make a website** option, so a new company can be described first and still continue into website creation later.

### Brainstorm Path

The brainstorm path opens a chat-style planning step where you describe the business idea, target market, offer, and positioning. Neotask then turns that conversation into a company profile draft.

This path is especially useful for:

- pre-launch companies
- service businesses still refining positioning
- new product or brand concepts

---

## The Current Wizard Step Order

After the entry-specific steps, the current onboarding flow continues with the same later stages:

1. **SOPs**
2. **Apps**
3. **Integrations**
4. **Autonomy**
5. **Heartbeat**
6. **Plan**

### SOPs

Upload playbooks, SOPs, reference documents, and internal guidance. This step is optional, but it improves planning and later task execution.

### Apps

Choose the apps and MCP tools the company is expected to use. This helps the generated plan attach the right tools to the right tasks.

### Integrations

Connect company-scoped Google Workspace access. This step is separate from general app selection and is used for services such as:

- Gmail
- Google Calendar
- Google Docs
- Google Sheets
- Google Drive

If you do not connect them here, you can still connect them later inside the company dashboard.

In the current flow, users can also choose whether to reuse an already connected Google account for the company or add a different account.

### Autonomy

Choose the company’s operating mode:

- **Fully Supervised**
- **Semi-Supervised**
- **Fully Autonomous**

You can change this again later.

### Heartbeat

Set how often the company heartbeat should run and how the company should watch for approvals, replies, or new incoming work.

### Plan

The onboarding flow ends on the plan step. This is where you review the generated company plan and confirm the starting setup.

![Initial company plan](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/origin-plan-1280.webp)

After onboarding completes, Neotask opens the new company on the **Plan** tab so you can keep refining the plan, connect any missing tools, and finish setup before starting autonomy.

---

## What Happens After You Finish

After the wizard completes successfully, Neotask:

![Company workspace ready after onboarding](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/origin-ready-1280.webp)

- creates the company workspace
- keeps the selected apps and integrations attached to the draft company context
- opens the new company directly on the **Plan** tab

At that point, you can keep working inside the company dashboard to:

- refine tasks
- connect missing apps
- link channels
- assign employees
- review approval policy
- start or pause the company heartbeat

---

## If The Company Is Not Ready To Start Yet

A newly created company may still need setup in the dashboard before autonomous execution starts.

That usually means one or more of these still needs attention:

- task setup is incomplete
- required apps are not connected yet
- channels or employee routing still need review
- the plan needs one more round of edits
- heartbeat settings still need confirmation

This is normal. Onboarding gets you into the company cleanly; the company dashboard is where you complete the operating setup.

---

## Best Practices

- Use the website path when the company already has a live site.
- Use the brainstorm path when the business still needs shaping.
- Upload SOPs early if the company has real operating rules.
- Connect Google Workspace during onboarding if email, calendar, docs, or sheets are part of the plan.
- Treat the onboarding plan as the starting operating plan, then refine it inside the company dashboard before turning the company loose.

---

## Related Docs

- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Operations, Controls & Settings](./auto-companies-operations-controls-and-settings.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)

---

### [Auto, Operations, Controls & Settings](https://neotask.ai/docs/auto-companies-operations-controls-and-settings)

## Review The Current Plan Before Changing Controls

1. Open the company and select **Plan**.
2. Review scheduled and dynamic work.

![Company Plan showing scheduled work and operating controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v6/2026-08-11-r1/ui/active-plan-1280.webp)

3. Select **Tasks** to inspect the recurring company work created by the plan.

![Company Tasks showing recurring work, schedules, apps, run status, and spend](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-11-r1/tasks-recurring-1280.webp)

4. Use the controls rail to change budget, heartbeat, autonomy, or quiet hours.
5. Open **Approvals** to confirm which actions still pause for review.

![Company Approvals showing the pending review queue](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/approval-queue-1440.webp)

This guide covers the company operating surfaces that sit around the main tab bar and the company settings panels that control runtime behavior.

These areas matter because they determine whether a company can start, how often it acts, how much it can do automatically, who gets notified, and what becomes visible outside the company workspace.

---

## The Left Controls Rail

Every open company has a left-side controls rail.

This is the main runtime control area for:

- start
- pause
- resume
- stop
- trigger now

It also shows the current company status and key operating signals such as:

- the current autonomy tier
- heartbeat timing
- last heartbeat
- plan-setup readiness
- budget posture

If a company is not ready to run yet, this rail is often where the blocking state is most obvious.

---

## Autonomy Tiers

The current controls rail uses three runtime tier labels:

- **Supervised**
- **Balanced**
- **Autopilot**

These are quick operating shortcuts for how freely the company should run.

If a company feels too locked down or too loose, this is one of the first places to check.

![Autonomy controls for company work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-autonomy-1280.webp)

For stricter approval tuning, use the dedicated **Approvals** tab or the approval settings panels in **Settings**.

## Approval Guardrails

Open **Approvals** to review the company policy and any app-level exceptions. These rules determine which actions pause before an external system changes.

![Approval and safety guardrails](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-guardrails-1280.webp)

---

## Heartbeat Controls

The heartbeat controls decide how often the company checks for work and keeps its recurring automation moving.

Heartbeat behavior appears in two places:

- the left controls rail for quick changes
- the **Settings** tab for the fuller settings experience

If a user says:

- “the company is not doing anything”
- “it is not checking often enough”
- “I need to change how often it runs”

the heartbeat settings are the right place to look.

---

## Continue Work After Restarts

The company **Settings** tab includes a **Continue work after restarts** toggle.

This decides whether interrupted work for this company starts again by itself after an app update, an operating-system restart, a crash, or a power loss. Interrupted scheduled runs resume, interactive threads continue where they stopped, and coding sessions reopen in their own branch working copy.

The account-wide setting in **Settings** -> **Automations** is the default for every company, and it is on. The company toggle overrides it for this company only.

Two behaviors are worth knowing:

- Changing the account-wide toggle re-syncs every company to the new value and clears company overrides.
- A company differs again only after you change the toggle inside that company's settings.

Work that is waiting for approval is never resumed automatically. It keeps waiting for a decision.

For the full behavior, see [Continue Work After Restarts](./continue-work-after-restarts.md).

---

## Budget Controls

The controls rail also includes company budget enforcement.

This area lets users understand:

- today’s spend
- the daily limit
- whether the daily limit is enforced
- whether the company should auto-pause when the limit is hit

This is especially important when a company appears healthy from an activity point of view but is not continuing to run as expected.

---

## Sub-Agent Auto-Approve

The controls rail also includes the sub-agent auto-approve toggle.

This is a company-level shortcut that affects whether delegated helper runs can move with less manual friction.

If a company is heavily using orchestrated or delegated work, this setting can materially change how much operator involvement is required.

---

## Overview Widgets And Quick Context

The company dashboard is not only a tab strip. It also includes shell-level summary surfaces around the tabs.

Those include:

- quick stats beside the main tab content
- the mini calendar
- overview widgets
- overview widget pinning
- overview widget add/customize flows
- activity detail overlays opened from overview widgets

The **Overview** tab is still the main place for these, but the surrounding shell and right-side context are part of the daily operating experience too.

---

## Nightly Rundown

The company view also includes a **Nightly Rundown** action outside the normal tab list.

Nightly Rundown is for the end-of-day or start-of-day company summary experience. It helps users review:

- today’s company summary
- gateway review summary
- tomorrow’s confirmed schedule
- suggested unscheduled work
- pending approvals
- unresolved requires-action items

This is useful when the user wants a compact operational review instead of opening several tabs separately.

---

## Notification Surfaces

Company notifications are not limited to the activity feed.

The current company experience also includes:

- the notification bell
- in-app notification history
- company notification preferences
- nightly rundown delivery preferences

Inside the company **Settings** tab, users can configure company-level notification behavior such as:

- in-app notifications
- desktop push notifications
- email notifications
- Slack delivery
- Discord delivery
- nightly rundown timing

If someone says “I’m not being notified,” the answer is usually in the company notification settings, not in the global app settings.

---

## Public Live Visibility

The company **Settings** tab also includes **Public Live** controls.

These settings let users decide whether the company appears on the public live board and how aggressively public activity is redacted.

That includes controls such as:

- whether the company appears publicly at all
- the public display name
- the redaction mode
- whether aggregate overview metrics include the company
- whether the company name is shown publicly
- whether revenue metrics are allowed into the public aggregate view

Public Live is about external visibility. It is not the same thing as internal activity history.

---

## Adaptive Learning Settings

The current company **Settings** tab also includes adaptive learning controls.

This area is for users who want to tune how the company learns from prior approvals, workflows, memory patterns, and orchestrator hints.

From a user point of view, the important ideas are:

- whether adaptive learning is on
- whether the company should only observe, suggest, or require review
- whether learned changes should show up in activity, memory, or nightly review

This is more advanced than the basic approval mode, but it is still part of normal company operations.

---

## Floating And Shared Company Surfaces

Some company experiences are shared across multiple tabs instead of belonging to one tab alone.

That includes:

- the floating company chat widget
- the notification bell
- tutorial overlays
- widget customization overlays
- activity detail modals

If a user says “something popped up over the company dashboard,” it may belong to this shared layer rather than one tab.

---

## Quick Decision Rules

- If the user needs to start, pause, resume, or trigger the company: use the left controls rail.
- If the user needs heartbeat timing or runtime frequency: use the controls rail or the heartbeat panel in **Settings**.
- If the user needs to review the company at a glance: use **Overview** and the right-side quick context.
- If the user wants the daily summary: use **Nightly Rundown**.
- If the user wants notification behavior: use **Settings** -> notification preferences.
- If the user wants public visibility controls: use **Settings** -> public live.
- If the user wants advanced learning behavior: use **Settings** -> adaptive learning.
- If the user asks why interrupted work did or did not start again: use **Settings** -> continue work after restarts.

---

## Related Docs

- [Continue Work After Restarts](./continue-work-after-restarts.md)
- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)

---

### [Auto, SOP Authoring & Review](https://neotask.ai/docs/auto-companies-sops-authoring)

SOPs are the operating guidance your company follows. This page covers where SOPs come from, how a draft becomes a published SOP, and the review controls you use along the way.

In the current product, an SOP can be created two ways: you can upload or write one yourself, or an agent can propose one for you. Either way, a human reviews and publishes it before the company treats it as guidance.

---

## What An SOP Is

An SOP is a piece of operating guidance that teaches the company how you want it to work.

Typical examples:

- playbooks
- internal procedures
- checklists
- brand or process guidance
- operational reference docs

Published SOPs become the operating guidance your agents follow when they do work.

---

## Where SOPs Live

Open the company and go to the **SOPs** tab.

![SOP procedure library showing published and reviewable procedures](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-procedure-library-1280.webp)

This is the single place to:

- review your published SOPs
- review SOPs an agent has proposed
- create a new SOP yourself
- approve, edit, or discard a pending draft

---

## Who Proposes vs Who Approves

There are two roles in the SOP flow, and they stay separate.

- **Agents propose.** As your company works and learns, an agent can draft an SOP that captures how a recurring task should be handled and propose it to you. A proposed SOP arrives as a **pending draft**, it is not active yet.
- **A human approves.** You always review a pending draft before it counts. Nothing an agent proposes becomes operating guidance until a person publishes it.

This keeps you in control of the company’s playbook. Agents can suggest how the company should operate, but the operating guidance only changes when you say so.

---

## The Review Workflow

When an SOP is waiting for you, open the **SOPs** tab and review the pending draft.

For each draft you can:

- **Approve / publish**, accept the draft as written and make it active operating guidance.
- **Edit**, adjust the wording or content first, then publish your edited version.
- **Discard**, reject the draft so it does not become guidance.

### Reviewing changes with a diff

When a draft updates an SOP you already have, the review shows a **diff**, a side-by-side view of what changed between the current SOP and the proposed version. The diff makes it easy to see exactly what an agent wants to add, remove, or reword before you accept it, so you are never approving a change blind.

A brand-new SOP is reviewed the same way; the diff simply shows it as new content.

![Pending SOP revision showing the proposed diff before publication](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-pending-diff-1280.webp)

---

## Creating An SOP Yourself

You do not have to wait for an agent to propose one. From the **SOPs** tab you can author your own SOP at any time.

### Create with Voice

You can also dictate an SOP by talking. Use **Create with Voice** (also shown as **Tell Me About It**) to describe the procedure out loud, and Neotask turns what you say into a drafted SOP. You can keep talking to refine it, then review and publish it like any other draft.

Voice authoring is just another way to get to a draft, once it exists, it goes through the same review and publish step.

---

## Uploaded vs Agent-Authored

SOPs are origin-agnostic. It does not matter whether you uploaded a document, wrote it yourself, dictated it with voice, or an agent proposed it, once it is published, it is treated the same way as company operating guidance.

This means:

- you can build your playbook from your own existing documents
- agents can fill in the gaps by proposing SOPs based on how the company actually works
- everything published lives together in the **SOPs** tab as one operating guide

---

## What Happens After You Publish

Once an SOP is published, it becomes active operating guidance for the company.

From that point on, your agents draw on it when they do relevant work, the most relevant SOPs are automatically found and attached to an agent’s context when it starts a task. You do not have to point an agent at the right SOP; the company surfaces the guidance that fits the work at hand.

For how that matching works, see [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md).

---

## Quick Decision Rule

- want to add operating guidance yourself → open **SOPs** and create one (typing or **Create with Voice**)
- an agent proposed an SOP → open **SOPs**, review the draft (use the diff if it updates an existing one), then approve, edit, or discard it
- want to know how agents pick which SOPs to follow → see [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md)

---

## Related Docs

- [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Training Mode](./auto-companies-training.md)
- [Auto, Autonomous Companies](./auto-companies.md)

---

### [Auto, Smart SOP Retrieval](https://neotask.ai/docs/auto-companies-sops-semantic-retrieval)

## Check Which Procedure Was Used

1. Open **SOPs** and select the procedure.
2. Review its active version and source history.
3. Open the related run when you need to confirm which evidence was retrieved.
4. Expand the cited source before changing the procedure.

![SOP source review with the active procedure and supporting material](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-source-review-1280.webp)

When your agents do work, they follow your company’s SOPs. Smart SOP Retrieval makes sure they follow the *right* ones, the SOPs that actually relate to the task in front of them, automatically.

This page explains what that means, why it helps, and what you need to know about how it behaves.

---

## What Smart Retrieval Does

Every time an agent starts work, an autonomous task, an agent chat, or a coding action, Neotask automatically finds the SOPs that are most relevant to what the agent is about to do and attaches them to the agent’s context.

The agent does not work from a wall of every SOP you have. It works from the handful that fit the current task, so the guidance it reads is the guidance that matters.

The run preserves the evidence the agent received. Expand an evidence entry to read the source text and confirm why it was used.

![Expanded memory evidence used by an agent](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r2/memory-evidence-1280.webp)

This happens on its own. You do not have to tag an SOP, point an agent at it, or pick which guidance applies, the company does that matching for you.

---

## How This Is Different From Before

The earlier behavior matched SOPs by keywords and simply attached a fixed top set. That worked when an SOP happened to share words with the task, but it missed guidance that was clearly related yet worded differently.

Smart retrieval matches by **meaning** instead of just words. It understands what an SOP is about and what the task is about, then finds the SOPs whose *meaning* is the closest fit, even when they do not share the same phrasing.

In practice that means:

- a task can pull in the SOP that covers it, even if the wording is different
- agents are not flooded with every SOP at once, they focus on the most relevant ones
- guidance you wrote in your own words still gets found for the right work

---

## Why It Helps

- **Better focus.** Agents read the guidance that fits the task, not a long list of everything.
- **Fewer misses.** Relevant SOPs are found by meaning, so the right one shows up even when the words don’t line up.
- **No manual matching.** You publish your SOPs and Neotask handles which ones apply to which work.

---

## What It Covers

- **All your SOPs, whatever their origin.** It covers SOPs you uploaded, SOPs you wrote, SOPs you dictated with voice, and SOPs an agent proposed and you published. Where an SOP came from does not change whether it can be retrieved.
- **Only your company’s SOPs.** Retrieval is isolated to your company. A company only ever sees and uses its own SOPs, never another company’s.

---

## When New SOPs Become Available

When you publish an SOP, it becomes available to smart retrieval right away. The matching index updates on publish, so newly published guidance can be found on the very next task, there is no long wait for it to “take effect.”

If you change an existing SOP and publish the update, the new version is what gets matched going forward.

---

## If The Matching Service Is Briefly Unavailable

Smart retrieval is designed so you never lose guidance. If the meaning-based matching is momentarily unavailable, the company falls back to the prior keyword-based selection instead. Your agents still get SOPs attached, they just use the older matching method until the smarter one is back.

In other words, there is no “SOPs stopped working” state. The worst case is a brief return to the previous behavior, then back to smart matching automatically.

---

## What You Need To Do

Nothing extra. Smart retrieval is on by default and runs automatically. Your only job is to keep your SOPs published and current, see [Auto, SOP Authoring & Review](./auto-companies-sops-authoring.md) for how SOPs are created, reviewed, and published.

---

## Related Docs

- [Auto, SOP Authoring & Review](./auto-companies-sops-authoring.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Autonomous Companies](./auto-companies.md)

---

### [Auto, Training Mode](https://neotask.ai/docs/auto-companies-training)

## Open A Training Session

1. Open the company and select **Train**.
2. Choose a task or completed run to use as training material.
3. Review the source steps and the proposed reusable procedure.
4. Accept the bundle only after its sources and instructions are correct.

![Company training workspace with source material and training status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/idx-train-1280.webp)

Training mode is the company dashboard surface for rehearsing one real task until it becomes reliable enough to trust live.

This guide explains what training mode does, what Neotask creates during training, and how to train a task efficiently without fighting the system.

---

## What Training Mode Is

The **Train** tab is not a generic chatbot and it is not a throwaway sandbox.

It is a guided rehearsal lane for one specific company task.

That means a training session is always tied to:

- one company
- one plan task
- one task outcome you are trying to improve

Training mode exists so you can:

- start from a vague or imperfect task
- watch how it behaves
- coach it before it becomes scheduled or trusted
- review exactly what changed
- keep the improvements for the next training run and the later live run

The goal is not to “train the model.” The goal is to train the task’s reusable operating guidance.

---

## Where To Find It

Open **Auto**, open a company, then click **Train** in the company tab bar.

The Train tab automatically loads the company’s current task list. You do not create a separate training-only task list.

That matters because training is supposed to improve the same real tasks that later run live.

---

## What A Fresh Training Session Looks Like

A true first-time training session starts with no accepted bundle and no staged bundle for that task.

The flow should feel like this:

1. open a real task in **Train**
2. answer a short preflight
3. run the baseline once
4. answer any meaningful checkpoints during the run
5. coach the task in the training input
6. click **Add to Training**
7. review the generated changes
8. click **Bundle It**
9. run **Train Again** until the result becomes consistent

You should not see a pile of training files before the first training pass has actually created them.

---

## What The Main Training Surfaces Mean

The training workspace is built around three operator jobs:

### Activity Feed

This is where you watch the task operate.

Use it to understand:

- what the task tried to do
- where it became uncertain
- what approvals it would have needed
- what the run produced

### Training Input

This is where you coach the task.

Use it to tell the task:

- what was wrong with the last run
- what you want it to prioritize next time
- what must never be invented
- how you want tone, routing, approvals, or outputs handled

### Bundle Review

This is where you review the training artifacts that Neotask generated from the run and your guidance.

This is the critical trust surface because it shows:

- what changed
- which files were added or revised
- what is only staged
- what is already accepted

---

## The Three Main Training Phases

### Preflight

Before the first meaningful run, Neotask asks short setup questions.

Use preflight to define:

- the desired outcome
- the required output shape
- the tone or style
- hard safety boundaries
- what to do when information is missing

Good preflight answers make the first run much closer to correct.

### Checkpointed Execution

During the run, Neotask can pause at meaningful action boundaries.

That means it may stop to ask about:

- a planned outreach action
- a browser workflow branch
- a CRM writeback decision
- an approval-sensitive next step
- a situation where more than one path looks plausible

It should not pause for every tiny click.

The right mental model is:

- pause on decisions
- not on every micro-step

### Post-Run Consolidation

After the run, you review what training generated.

This is where **Add to Training** and **Bundle It** matter.

That final review is mandatory because the system is supposed to show you the exact guidance it wants to keep.

---

## What Add To Training Does

**Add to Training** stages a working bundle for that task.

That working bundle is the system’s current proposal for how the task should improve.

It is not fully trusted yet. It is the staged version you review before accepting it.

After you click **Add to Training**, you should expect to see:

- newly created or updated guidance files
- a working summary
- a working diff
- a clear separation between staged changes and already accepted guidance

---

## What Bundle It Does

**Bundle It** promotes the current working bundle into the accepted bundle.

That accepted bundle becomes the task’s active training guidance for:

- the next **Train Again** run
- the later live version of the same task

This is the point where training stops being a draft and starts becoming durable task behavior.

---

## What Files Training Creates

Training mode works by creating reusable guidance files for the task.

The exact file set can grow over time, but the core training bundle typically includes these kinds of files:

### Routine

The routine is the task’s operating playbook.

It usually captures:

- the task purpose
- the exact outcome contract
- the order of sections or steps
- execution rules
- approval wording
- cadence or operating logic

### Output Template

The output template defines what the finished answer should look like.

This is where Neotask can lock in things like:

- required headings
- section order
- field structure
- placeholder format

### QA Checklist

The QA checklist is the “before returning this result, verify these things” layer.

This is useful for making a task reliably check itself before it finishes.

Typical examples:

- required headings are present
- the approval sentence is exact
- no fake CRM fields were invented
- the response stayed concise

### Placeholders

Placeholder files tell the task what to do when real data is missing.

This prevents the task from making up details just to complete the shape.

Examples include placeholder tokens for:

- names
- emails
- owners
- dates
- stage IDs

### Additional Guidance Files

Depending on the task, Neotask can also create or evolve other bundle files later.

Examples include:

- memory notes
- browser-specific guidance
- supporting rule sheets
- future task-specific helper artifacts

The important operator rule is simple:

you are reviewing reusable task guidance, not just one answer transcript.

---

## Where This Information Comes From

This is the point that most users need clarified at least once:

the long training documents you see are usually **not** something the user typed by hand.

They come from a combination of:

- the original task definition
- your preflight answers
- your checkpoint answers
- your coaching messages
- the task’s baseline run
- the model’s attempt to turn all of that into reusable guidance

So if you open a trained task and see a long routine or checklist, that does not mean someone manually authored every line in the training pane. It means training has already materialized those instructions into task files.

---

## Working vs Accepted Bundles

Training mode keeps two bundle states separate on purpose.

### Working Bundle

The working bundle is the currently staged proposal.

Think of it as:

- “what this next revision would keep if I accepted it”

### Accepted Bundle

The accepted bundle is the currently trusted task guidance.

Think of it as:

- “what the next rerun and live task already use”

If you are not sure what changed recently, compare working against accepted before bundling.

---

## How Training Improves The Next Run

Training improves a task because the accepted bundle is loaded again on the next run.

That means the next training run does not start from zero. It starts with the accepted routine, template, checklist, placeholders, and other task-linked guidance already attached to that task.

That is why a good training loop should look like:

- rough first run
- clearer second run
- consistent later runs

If a task keeps behaving exactly like the baseline after bundling, training is not actually being applied correctly.

---

## What Is Simulated In Training

Training mode is supposed to be a rehearsal lane.

That means the task can still surface:

- intended actions
- approval gates
- planned outreach
- planned writebacks
- browser steps

But the user experience should still feel like review and guidance first, not uncontrolled live execution.

The operator should be able to see:

- what would have happened
- what needs approval
- what still needs clarification

before trusting the task live.

---

## The Best Way To Train A Task

The best training runs start from a task that is slightly vague but has a clear business outcome.

Good examples:

- re-engage stale qualified leads
- prepare a renewal follow-up package
- review inbound support escalations and prepare a response plan

This is the recommended training pattern:

1. define the desired business outcome clearly
2. answer preflight in concrete terms
3. let the first run happen without overcorrecting too early
4. fix the biggest structural mistakes first
5. stage the bundle
6. review the files and diffs
7. bundle only when the guidance looks durable
8. rerun several times to verify consistency

Do not judge training by whether one lucky run looked good. Judge it by whether repeated reruns keep producing the right structure and decisions.

---

## Best Practices

- Start with one task, not ten at once.
- Train for one clear business outcome per session.
- Answer preflight in concrete business language, not vague wishes.
- Use checkpoints to explain decisions, priorities, and exceptions.
- Review the diff before bundling anything.
- Bundle only when the guidance looks reusable, not just impressive.
- Re-run the task several times after bundling.
- Treat training as a path to stable live execution, not as a separate permanent operating mode.

---

## What To Avoid

- Do not dump random text into training just to “give it more data.”
- Do not bundle a revision you did not actually read.
- Do not overfit to one narrow example if the live task has to handle broader cases.
- Do not let the task invent names, dates, IDs, or CRM values just because the template looks empty.
- Do not confuse the activity feed with the accepted task guidance. The accepted bundle is what the next run keeps.

---

## How To Tell Training Is Working

You should see this pattern:

1. the first run is incomplete, weak, or inconsistent
2. your guidance creates a meaningful working bundle
3. the files and diffs reflect the improvements you actually asked for
4. the accepted bundle becomes the new baseline
5. repeated reruns stay in the improved shape

That is the strongest sign that the task is truly learning the right behavior at the task level.

---

## How Training Relates To The Rest Of Auto

Training mode is not separate from the rest of the company dashboard. It works with the same company context.

Training is closely related to:

- **Plan**, because you are training real plan tasks
- **Tasks**, because task settings still matter
- **Approvals**, because simulated approvals and live approvals need to line up
- **SOPs**, because company guidance still shapes the task
- **Apps**, **Integrations**, **Channels**, and **Employees**, because the task still depends on the company’s real operating context

Use training when the task exists but is not trustworthy enough yet. Use the live task lane only after training proves the task is stable.

---

## Related Docs

- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Onboarding & Plan Setup](./auto-companies-onboarding.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Approvals & Safety](./auto-companies-approvals-and-safety.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)

---

### [Automation & Scheduling](https://neotask.ai/docs/automation)

Automate your Neotask agents with scheduled jobs, flexible delivery modes, and powerful session targeting.

---

## Scheduled Jobs (Cron)

Scheduled jobs let you automate agent actions on a recurring or one-time schedule. Define when a job runs, what it does, and where the results go.

Free-plan users cannot create, edit, run, skip, or delete their own scheduled jobs, cron jobs, reminders, or automations. They can use chat, boards, coding workflows, and connected apps within the Free message quota, but autonomous scheduling requires at least an Individual subscription.

The only Free-plan automation exception is the system-managed daily memory learning job named `Learn from Daily Activities`. It runs every day at 3:00 AM to maintain memory across agent and company workspaces. It is not a user-created automation.

Open **Agents**, select an agent, and choose **Automations** to see scheduled work and its current state.

![Automations page showing scheduled agent work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/automations-main-1280.webp)

---

## Schedule Types

| Type | Format | Example |
|------|--------|---------|
| **At** | Fixed datetime | Run once at Feb 28, 2026 9:00 AM |
| **Every** | Interval | Run every 30 minutes |
| **Cron** | Standard cron expression | `0 9 * * 1-5` (weekdays at 9 AM) |

### Timezone Support

- All schedules respect your configured timezone
- Cron expressions support a timezone parameter
- Stagger option available to offset execution times across multiple jobs

---

## Job Payloads

Each scheduled job executes a payload -- the action the agent performs when the job fires.

| Type | Description |
|------|-------------|
| **System Event** | Trigger an arbitrary event (text-based) |
| **Agent Turn** | Send a message to an agent with an optional model override |

---

## Delivery Modes

Control where the results of a scheduled job are sent after execution.

| Mode | Description |
|------|-------------|
| **None** | Internal execution only -- no output is delivered |
| **Announce** | Broadcast the result to a channel |
| **Webhook** | POST the result to an external URL |

---

## Session Targeting

Choose how conversation context is handled for each scheduled job.

| Mode | Description |
|------|-------------|
| **Main** | Shared session -- cumulative conversation history carries over between runs |
| **Isolated** | Fresh session per run -- clean slate each time the job executes |

---

## Job Operations

Neotask gives you full control over the lifecycle of your scheduled jobs:

- **Create, update, and delete** jobs at any time
- **Enable or disable** a job without deleting it
- **Force-run immediately** to bypass the schedule and execute on demand
- **Skip once** to delay a single execution, then automatically resume the normal schedule
- **View full run history** including status, duration, and errors for each execution

---

## Interrupted Runs

A scheduled run can be cut short by an app update, an operating-system restart, a crash, or a power loss. Neotask does not lose that run.

**Continue work after restarts** decides what happens next. It is on by default.

1. Open **Settings**.
2. Select **Automations**.
3. Turn **Continue work after restarts** on or off.

With it on, an interrupted run resumes after the restart, a bounded number of times. Work that already finished is never repeated, and blog posts and emails are protected against duplicates if a run is retried.

With it off, the run is recorded as **Interrupted by restart** in the activity feed and waits for you to run it again.

Runs that are waiting for your approval are never resumed automatically, and neither are runs that failed for a reason other than the restart.

The same **Automations** screen has **Launch Neotask at login**, which is off by default. Scheduled tasks keep running with the app closed, but company planning, new task creation, and the nightly and weekly reviews run only while the app is open. Turn this on so they continue right after a restart.

For the full behavior, see [Continue Work After Restarts](continue-work-after-restarts.md).

---

## Example Automations

Here are some common ways teams use Neotask scheduled jobs:

### Morning Briefing

- **Schedule:** Daily at 8:00 AM
- **What it does:** Agent summarizes the latest news and sends a digest to Slack

### Weekly Report

- **Schedule:** Every Monday
- **What it does:** Agent compiles analytics data and emails a summary to the team

### Data Sync

- **Schedule:** Every 4 hours
- **What it does:** Agent checks external APIs and updates the dashboard with fresh data

### Health Check

- **Schedule:** Every 15 minutes
- **What it does:** Agent monitors services and sends alerts on failure

---

### [Billing & Plans](https://neotask.ai/docs/billing)

Everything you need to know about Neotask pricing, how the platform works, and what you're paying for.

---

## How Neotask Works

Neotask is a **platform**, not a traditional SaaS with hard usage caps. You are paying for access to a unified agent orchestration system: the software, infrastructure, and compute that powers your AI agents across every channel, integration, and automation.

Unlike services that advertise "20x usage" or bundle a fixed number of agent runs, Neotask gives you the **full platform** with every feature included in your plan. Your subscription covers the software and an initial allocation of AI token usage. Any usage beyond that allocation is billed at transparent usage rates.

**What your subscription covers:**

- Full access to the Neotask desktop app, web dashboard, and all platform features
- Unified agent management across 14+ messaging channels
- Voice control, automation scheduling, smart routing, and team orchestration
- All 55+ built-in skills and tools
- All supported AI model providers (Anthropic, OpenAI, Google, DeepSeek, OpenRouter)

**What usage fees cover:**

- The actual AI token consumption (input and output tokens) your agents use when processing tasks
- Flat scheduled-task fees for real autonomous or automation runs that fire later from saved schedules
- Usage is metered transparently so you only pay for what your agents consume and the scheduled work they actually execute

---

## Plans

### Check Your Current Plan

1. Open **Settings**.
2. Select **License** under **General**.
3. Read the plan, agent count, status, and expiration date at the top of the screen.

The License screen is also where you can reopen either product walkthrough and reach the account data controls.

![License settings with plan, agent count, status, and expiration date](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/license-plan-settings-1280.webp)

### Free

- **Price:** $0/month + usage coverage
- **Message quota:** First `10` messages/events as a one-time intro allowance, then `5` messages/events per user-local calendar day.
- **Usage coverage:** Free users must cover hosted model/token usage with Neotask credits or their own provider keys. Local models on your computer are free and do not count against the message quota. Every account still starts with the standard `$10` starter credit grant.
- **Allowed surfaces:** Main chat mode, boards, coding workflows, app connection flows, and connected-app usage inside allowed chat/boards/coding interactions.
- **Locked surfaces:** Company operation, company tasks, user-created automations, scheduled jobs, cron jobs, reminders, and custom agent creation.
- **Company onboarding:** Free users can onboard or preview a company, but they cannot actually run company tasks or automations unless they upgrade to at least an Individual subscription.
- **Overage:** Free users get no overage. If credits or provider balance are exhausted, usage hard-stops until they add credits, fix provider billing, or upgrade.

For the detailed Free-plan contract, see [Free Plan](free-plan.md).

### Individual

- **Price:** $50/month + usage
- **What you get:** Personal access to the Neotask platform with no company or sector slots included.
- **Included setup tasks:** **25 one-time setup tasks** before flat `$1` scheduled-task fees begin.
- **Best for:** Solo users who want the full app and agent platform without company operations.

### Business

- **Price:** $100/month + usage
- **What you get:** Everything in Individual, plus **1 company or sector** and support for paid employee seats.
- **Employee seats:** Additional employee seats are **$50/month each** for base access. Giving a seat company access (collaborator or manager) is a separate **+$50/month upgrade**, so a company-access seat totals $100/month.
- **Included setup tasks:** **50 one-time setup tasks** before flat `$1` scheduled-task fees begin.
- **Best for:** One business or team that wants company operations without multi-company expansion.

### Enterprise

- **Price:** $200/month + usage
- **What you get:** Everything in Business, plus **2 companies or sectors included** and support for additional company or sector slots.
- **Additional companies or sectors:** **$100/month each**
- **Employee seats:** **$50/month each** for base access. Giving a seat company access (collaborator or manager) is a separate **+$50/month upgrade**, so a company-access seat totals $100/month.
- **Included setup tasks:** **50 one-time setup tasks** before flat `$1` scheduled-task fees begin.
- **Optional setup engagement:** Dedicated onboarding is available as a one-time add-on for customers who want hands-on rollout help.

All paid plans can still use either hosted credits or BYOK, depending on how you want billing and model access to work.

**All users start with $10 in one-time signup credits.** These credits cover your initial usage, but active users will likely need to top up or enable overage billing as their agents consume tokens. You can purchase additional credits at any time starting at $5, and auto top-up is available to keep your balance funded automatically. Usage-based billing is transparent: you see exactly what your agents consume, and you only pay for actual usage. For enterprise users running agents at scale, Neotask is actively automating your business, reducing costs, and making you money.

For example, a scheduled autonomy task that takes ~3 hours to complete and consumes $10 in provider usage would cost $11 total if it qualifies for a flat task fee: $10 for raw AI compute plus $1 for the scheduled task execution. If you are still inside your included setup-task allowance, the flat `$1` fee is waived, but token and provider usage still apply.

For the complete pricing breakdown, see the [detailed usage rate schedule](usage-rates.md).

---

## Scheduled Task Fees

Neotask now bills **$1 per real billable scheduled execution** on top of normal usage billing.

This applies to:

- autonomous company plan tasks
- autonomous company dynamic tasks
- saved agent automations that later fire on schedule
- schedules created in chat, but only when the saved schedule actually fires later

This does **not** apply to:

- heartbeats
- one-off chat turns
- tools inside a task
- approvals by themselves
- generic reminder or training crons unless they are upgraded into a billable automation lane

One scheduled execution is still just one `$1` fee, even if it performs many tool calls.

A scheduled execution whose model calls all ran on a local model on your computer is currently not charged. Its `$1` fee is taken when the run starts and credited back after the run's usage shows that every model call ran locally, usually within about half an hour. If the run itself fell back to a hosted model, the fee stays. Hosted model calls made by sub-agents the run starts are billed as normal usage and do not affect this fee.

---

## Included Setup Tasks

Paid plans include a one-time setup cushion so you can tune automations before flat task-fee deductions begin:

- **Individual:** 25 included setup tasks
- **Business:** 50 included setup tasks
- **Enterprise:** 50 included setup tasks

These included tasks only waive the flat `$1` scheduled-task fee. They do **not** waive provider usage, model-token usage, or any normal platform markup.

They are **not monthly resets**. They are a startup allowance.

---

## Credits, BYOK, And Billing Readiness

Neotask still uses the existing credits system for scheduled-task fees.

- **Hosted credits mode:** provider usage and scheduled-task fees are both funded through credits.
- **BYOK mode:** provider usage goes to your own provider account, but flat scheduled-task fees and any platform usage charges still use Neotask credits.

For autonomous companies and billable automations, Neotask requires billing readiness:

- saved payment method
- automatic top-up enabled
- daily budget configured

Automatic top-ups will **never** exceed the remaining daily budget headroom for that day.

### Add Credits Or Update Your Payment Method

1. Open **Settings**.
2. Select **Credit top-up** under **AI**.
3. Choose a preset amount or enter a custom amount.
4. Use **Update Payment Method** if the saved card needs to change.
5. Click the button showing the selected amount to complete the top-up.

The top of the Credit top-up screen shows available credits and unsettled overage before you buy more credits.

![Credit top-up settings with balance, purchase amounts, and saved payment method](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/credit-top-up-settings-1280.webp)

### Configure Auto Top-Up And A Daily Budget

1. Stay on **Settings → AI → Credit top-up**.
2. Turn on **Auto top-up when balance is low**.
3. Enter the balance threshold and the amount to add.
4. Set the daily budget cap for autonomous companies and billable agent automations.

The lower part of the same screen contains recent task-fee activity and the top-up payment trail.

![Auto top-up, daily budget, and task-fee history in Credit top-up settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/auto-top-up-settings-1280.webp)

---

## No Task-Fee Overage

Normal usage overage can follow its own plan rules, but scheduled task fees do **not** continue once your billable task-spend balance is gone.

After the included setup-task allowance is exhausted:

- if your balance is positive, a billable scheduled run can start
- if your balance is **$0 or negative**, that run is blocked before launch
- the schedule is **paused, not deleted**
- once credits are restored, the same automation or schedule becomes runnable again automatically

This is designed to keep autonomous work predictable instead of silently digging deeper into overage.

---

## Why This Model

Traditional AI agent services charge flat fees for limited runs or cap you at a fixed number of agents. Neotask takes a different approach:

- **You pay for the platform:** the unified orchestration layer, the desktop app, voice control, channel integrations, automation engine, and every tool and skill.
- **You cover your own token usage:** the actual AI compute your agents consume. This is metered transparently so you never pay for idle capacity or artificial limits.
- **Enterprise scales with you.** It includes multi-company operations, employee seats, and expandable company capacity without forcing separate subscriptions for every company slot.

This model keeps costs predictable for light users and scales efficiently for teams running hundreds of agents at once.

---

## Payment Methods

| Method | Details |
|--------|---------|
| **Credit/Debit Card** | Visa, Mastercard, and Amex via Stripe |
| **iOS In-App Purchase** | Available through the App Store (powered by RevenueCat) |
| **Enterprise Invoicing** | Custom invoicing available for Enterprise plans |

Card changes for hosted credits happen in **Settings → AI → Credit top-up**. The saved card appears inside the Add credits panel beside **Update Payment Method**.

![Saved payment method and Add credits controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/payment-method-settings-1280.webp)

---

## Managing Your Subscription

- **Open Settings** from the avatar menu in the upper-right corner
- **Use General → License** to check the current plan and status
- **Use AI → Credit top-up** to manage the saved payment method, add credits, and configure auto top-up
- **Use AI → Models** to choose how models are connected
- **Set a daily budget** before you rely on autonomous companies or scheduled automations
- **Cancel anytime.** You retain access until the end of your current billing period.
- **No long-term contracts** required

---

## API Keys & Provider Access

Neotask supports two modes for handling AI provider access:

### System Key Mode (Default)

- Use Neotask's managed API infrastructure
- Token usage is tracked and billed through your plan
- No setup required; works out of the box

### BYOK Mode (Bring Your Own Keys)

- Use your own API keys from supported providers
- **Supported providers:** Anthropic, OpenAI, Google AI, OpenRouter, DeepSeek
- Keys are stored with AES-256-GCM encryption
- Full control over which models you use
- Base token costs go directly to your provider, but a **20% platform fee** and any applicable **automation markup** are still deducted from your Neotask credit balance

---

## Budget Controls

Stay in control of your AI spending with built-in budget management:

- Set **daily spending limits** (global or per-agent)
- **Auto-shutoff** when budget is exceeded
- **Budget-aware auto top-up caps** so autonomous replenishment never exceeds the remaining daily budget headroom for that day
- **Task-fee launch blocking** when scheduled work reaches a zero or negative task-spend balance
- **Real-time usage tracking** so there are no surprises
- **Budget snapshots and history** for auditing
- **Configurable alerts** to notify you before limits are reached

---

## Usage Tracking

Neotask provides detailed visibility into your usage and costs:

Open **Usage** from the main navigation to review model activity and cost by day. Select a row for its provider, model, agent, and session details.

![Usage overview with model activity and cost](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

- **Token-level tracking:** input, output, cache read, and cache write tokens
- **Scheduled-task execution records** with execution ids, scope ids, and billing status
- **Task-fee credit-backs** if reconciliation later shows a run failed or became non-billable
- **Payment-linked allocation trails** that show which top-up lot funded which task fee
- **Downloadable billing PDFs** for task executions and payment-lot statements
- **Cost breakdown** by model, provider, agent, channel, and date
- **Hourly drill-down** for today's usage
- **30-day rolling usage charts** for trend analysis
- **CSV/JSON export** for external analysis and reporting

---

### [Blog](https://neotask.ai/docs/blog)

## Follow An Article From Draft To Publish

1. Open the company and select **SEO & Blogs**.
2. Open the pipeline to see the article's current stage.
3. Select an article to review its content, sources, and SEO checks.
4. Publish only after the review state is complete.

![SEO and blog production pipeline](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-13-r2/blog-pipeline-1280.webp)

Neotask blog publishes insights, tutorials, product updates, and guides about AI agents,
automation, and the Neotask platform.

---

## Accessing the Blog

- **From the website:** Click **"Blog"** in the top navigation bar at neotask.ai
- **Direct URL:** [neotask.ai/blog](/blog)
- **Localized URLs:** The blog is available in all 21 supported languages. Localized versions are
  at `neotask.ai/{lang}/blog`, for example:
  - Spanish: [/es/blog](/es/blog)
  - Japanese: [/ja/blog](/ja/blog)
  - German: [/de/blog](/de/blog)

---

## Browsing Articles

### Article Grid

The blog index displays articles in a responsive grid of cards. Each card shows:

- **Cover image:** article thumbnail
- **Headline:** article title
- **Category pill:** colored badge showing the article's category
- **Reading time:** estimated time to read the full article

12 articles are shown per page. Use the **pagination controls** at the bottom of the page to
navigate between pages.

### Filtering by Category

At the top of the blog index, you'll see **category pills**, one for each available category.

- Click any category pill to filter articles to that category only
- The URL updates to `/blog/category/{category-slug}` (e.g. `/blog/category/tutorials`)
- Click **"All"** to clear the filter and show all articles

### Filtering by Tag

Tags appear on individual article pages. Click any tag to view all articles with that tag.
Tag-filtered views are at `/blog/tag/{tag-slug}`.

### Searching Articles

A **search bar** appears at the top of the blog index.

- Type a keyword to search across article titles and content
- Results update as you type
- Clear the search to return to the full article list

---

## Reading an Article

Click any article card to open the full article at `/blog/{article-slug}`. The article page includes:

- **Full headline** and publication metadata at the top
- **Reading time estimate:** how long the article takes to read
- **Table of contents:** a sidebar on the right (desktop) listing all sections. Click any section
  to jump to it. Hidden on mobile.
- **Article body:** full content with formatted text, images, code blocks, and tables
- **Sources section:** external references cited in the article. If there are 3 or more sources,
  they are compacted into an inline paragraph to save space.

![Completed article with content, sources, and controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-13-r2/blog-article-1280.webp)

The SEO review records the article checks separately from the writing surface. Review it before publishing.

![SEO review for a completed article](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-13-r2/blog-seo-1280.webp)

---

## Multi-Language Support

Blog content is available in **21 languages**: English, Spanish, French, German, Italian,
Portuguese (BR), Russian, Chinese (Simplified & Traditional), Japanese, Korean, Hindi, Arabic,
Dutch, Polish, Indonesian, Thai, Vietnamese, Swedish, Turkish, and Ukrainian.

- Your language is auto-detected from your browser locale
- Switch languages using the site language selector
- Some articles may not be available in all languages; you'll see the English version as fallback

---

### [Boards & Live Artifacts](https://neotask.ai/docs/boards-live-artifacts)

## Open And Review A Board

1. Open **Boards**.
2. Select an existing artifact or start a new board from chat.
3. Review the live sources and values before publishing.
4. Use the artifact controls to save, refresh, publish, export, or restore a version.

![Revenue command center board with its artifact rail](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/work-boards-1280.webp)

Boards are live artifact pages that you can build through chat, refine in a draft editor, and keep as reusable operating surfaces for an agent or a company.

This guide explains what boards are, where to find them, how the chat-to-artifact flow works, what data sources they can use, and how to turn a rough idea into a saved live page that stays useful.

---

## What Boards Are

A board is not just a single chart.

It is a full live page that can contain one or more widgets such as:

- KPI blocks
- line charts
- area charts
- multi-line charts
- bar charts
- tables

Think of boards as custom live dashboards for the exact question you want answered.

Good board use cases include:

- revenue health for one company
- support backlog by owner and priority
- operating exceptions across connected tools
- calendar plus file-change activity in one place
- a personal agent dashboard for one workflow

Boards are meant to move beyond the default overview widgets when you need a more specific operating surface.

---

## Where To Find Boards

Boards exist in two places:

### Company Boards

Open **Auto**, open a company, then click **Boards**.

Use company boards when the artifact should use company-scoped context such as:

- connected company apps
- company files
- company SOPs
- company tasks and operating context

### Agent Boards

Open **Agents**, open one agent, then click **Boards**.

Use agent boards when the artifact should use agent-scoped context such as:

- the agent workspace
- agent memory
- agent-attached tools and MCP connections
- agent-specific files

The board system is the same in both places. The main difference is the scope of the data it can reach.

---

## What A Fresh Board Session Looks Like

A fresh board session starts in the **New artifact chat**.

The expected flow is:

1. describe the live page you want
2. let Neotask create the latest artifact candidate in chat
3. apply that result into a draft
4. review the live source wiring
5. preview the board
6. save it
7. publish a version when it is worth keeping
8. activate it live when the wiring is ready

The important mental model is:

- chat creates the candidate
- the draft editor turns that candidate into a structured board
- saved versions make it durable
- live activation makes it an actively refreshing artifact

---

## How The Chat Composer Works

The composer is the fastest way to create a board.

It is built for natural requests such as:

- “Make a live Stripe revenue artifact with KPIs, trend lines, and recent payments.”
- “Build an artifact that combines Google Calendar activity with company file changes.”
- “Create a live operations page with charts, tables, and exceptions from my connected data.”

The composer can use:

- connected apps
- MCP tools
- company files
- workspace files
- SOPs
- memory
- internal company context

You can also:

- attach files
- attach connected integrations into the prompt
- keep iterating in the same chat until the latest result is worth promoting into a draft

When the chat has produced a usable candidate, the composer shows that the latest artifact candidate is ready. That is the moment to move it into the draft lane and review it as a real board.

---

## The Difference Between Chat And The Saved Artifact

This is the part most users need clarified:

the chat is not the artifact itself.

The chat is the creation lane.

The saved artifact is the structured board record that can be:

- previewed
- validated
- saved
- published
- activated live
- favorited
- exported
- restored to older versions

If you only stay in chat, you have not actually saved the board yet.

---

## What Data Sources A Board Can Use

Boards can pull from four main source types.

### Connected App / MCP Tool Sources

Use this when a widget should fetch live data from a connected system.

Examples:

- Stripe payments
- CRM records
- calendar events
- issue trackers
- project tools

This is the most common source type for truly live artifacts.

### Company File Search

Use this when the board should search through company files and turn those results into a live widget.

Examples:

- recent reports
- uploaded documents
- generated website content
- operating exports

### Workspace Files

Use this when the board should read files from the agent workspace rather than company-managed files.

This is especially useful for agent-scoped boards.

### Manual Rows

Use this when you want to:

- prototype the layout
- mock up a board before wiring live data
- create a static widget intentionally

Manual rows are the easiest place to start when you want to shape the page first and connect live sources later.

---

## What Live Source Review Means

Open the board grid to check the values and charts that were produced from the selected sources.

![Board grid with live operational data](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/work-boards-grid-1280.webp)

Boards that rely on live data have to be checked before they should be trusted.

The **Live Source Review** panel exists for that reason.

It tells you whether the current board still has issues such as:

- missing connector bindings
- missing authentication
- missing tools
- blocked connector policy
- imported board bindings that still need to be reviewed

This matters because a saved board is only as reliable as the data sources it can actually resolve.

You should treat Live Source Review as the final reality check before publishing or activating a board.

---

## The Board Lifecycle

Boards move through three main states.

### Draft

Draft is the editable working state.

Use draft when you are still:

- shaping the layout
- fixing data wiring
- refining descriptions
- adjusting widgets

### Published

Published means you have kept a durable saved version.

This is useful when you want:

- a stable checkpoint
- a version you can restore later
- a clear review milestone before going live

Publishing creates version history.

### Live

Live means the published version is now the active live artifact.

A live board:

- re-fetches when you open it
- keeps re-fetching while it is open
- behaves like an active operating surface rather than a one-time static draft

That is why live boards are the right end state for dashboards you want to keep using regularly.

---

## Refresh Behavior

Boards do not all refresh the same way.

### Draft Or Saved-But-Not-Live Boards

These re-fetch on open and when you manually refresh them.

### Live Boards

These re-fetch on open and continue refreshing while the board is open.

So if the board is meant to stay current while you monitor it, activating it live is the right step.

---

## Save, Publish, And Activate Live

These three actions do different jobs.

### Save Artifact

Save keeps the current board as a durable artifact record.

Use save when:

- the board is worth keeping
- you want it in the artifact list
- you want to come back and keep refining it later

### Publish

Publish creates a real saved version.

Use publish when:

- the board has reached a stable milestone
- you want a version you can restore later
- you want a controlled checkpoint before making it live

### Activate Live

Activate Live makes the published version the live artifact.

Use this only when:

- the data wiring is ready
- the validation issues are resolved
- the board is trustworthy enough to operate as a real live page

If a board still has binding or connector issues, it should stay in draft or published review, not live.

---

## Versions And Restore

Boards keep version history after publish.

That means you can:

- see the version list
- understand which version is current
- understand which version is published
- understand which version is live
- restore an older version if a newer one made the board worse

This is important because boards are operating surfaces. You do not want to lose a working version just because a later experiment was wrong.

---

## Import, Export Package, And Export PDF

Boards support more than just in-place editing.

### Import File

Use import when you want to bring in a board draft from a package file.

This is useful for:

- moving a board definition between environments
- sharing a starting point
- restoring a packaged draft into the editor

### Export Package

Use export package when you want the board definition itself.

This is the right choice when the goal is:

- reuse
- transfer
- backup
- import into another environment later

### Export PDF

Use export PDF when you want a readable snapshot of the current rendered result.

This is useful for:

- sharing the current view
- reviews
- reporting
- archiving a static representation of the artifact

---

## Plan Limits And Saving Rules

Drafting and previewing are more flexible than saving.

Current save limits are:

- Free: no saved artifacts
- Pro: 5 saved artifacts
- Individual: 5 saved artifacts
- Business: 10 saved artifacts
- Enterprise: unlimited saved artifacts

That means a user can still explore the composer and preview ideas before committing to a saved artifact, but saving is plan-limited.

---

## What Makes A Good Prompt

The best boards start from a business question, not a chart type.

Good prompt ingredients:

- what the page should help you decide
- which systems it should use
- what metrics or signals matter
- what time window matters
- what exceptions or anomalies should stand out
- what shape the page should take

Good example:

“Build a live revenue health artifact for this company using Stripe and our company files. I want a KPI row, a weekly trend chart, recent payments, failed payments, and a short exceptions table.”

Weak example:

“Make a chart.”

The second prompt is too vague. The first tells Neotask what the artifact is for and what should appear on the page.

---

## Prompt Examples You Can Reuse

### Revenue And Finance

- “Build a live Stripe revenue artifact with KPIs, trend lines, and recent payments.”
- “Create a finance board that shows daily revenue, failed charges, refunds, and the five biggest anomalies this week.”
- “Make a cash-collection artifact with open invoices, overdue counts, and recent collections.”

### Operations

- “Build a live operations artifact for this company with task volume, exceptions, and recent approvals.”
- “Create an operations page that combines recent company activity with file changes and open blockers.”
- “Make a board that shows throughput, pending approvals, and anything stuck for more than 24 hours.”

### Sales And CRM

- “Create a pipeline health artifact with stage counts, stalled deals, and top owners by open revenue.”
- “Build a board for stale qualified leads that shows who has gone cold, how long they have been cold, and who owns them.”
- “Make a renewal board with upcoming renewals, risk signals, and accounts needing manual outreach.”

### Support

- “Build a support board with open ticket volume, oldest unresolved items, SLA risk, and recent escalations.”
- “Create a support artifact that groups urgent tickets by owner and highlights backlog spikes.”

### Calendar And Activity

- “Build an artifact that combines Google Calendar activity with company file changes.”
- “Create a live weekly planning page with meetings, deadlines, and recent operational changes in one board.”

### Content And Marketing

- “Make a content operations board with blog production status, pending approvals, recent headlines, and article throughput.”
- “Build a marketing artifact that shows campaign launch dates, recent content output, and exceptions that need review.”

### File-Driven Or Internal Boards

- “Create a board from company files that shows the newest uploaded reports, their dates, and the biggest changes.”
- “Make a live workspace artifact that tracks the most recent generated files and groups them by folder.”

### Personal Agent Boards

- “Build a personal research board for this agent with recent findings, source files, and next actions.”
- “Create an engineering board for this agent with repo activity, pending file changes, and the current working set.”

---

## Best Practices

- Ask for a page, not a single chart.
- Start with the decision the board should support.
- Name the systems it should use.
- Say what should stand out, not just what should be displayed.
- Use manual rows first if you want to prototype the shape before wiring live data.
- Review Live Source Review before publishing or activating.
- Save when the board is worth keeping, not just because the preview looked interesting once.
- Publish when you want a durable version checkpoint.
- Activate live only when the board is actually ready to stay connected to live sources.
- Use the advanced editor for cleanup and structural repair, not as the default creation path.

---

## What To Avoid

- Do not describe only a chart type without saying what business question the board should answer.
- Do not activate a board live before its connector or binding issues are resolved.
- Do not confuse the chat result with the saved artifact.
- Do not treat one good preview as proof that a live board is reliable.
- Do not ignore imported binding warnings after package import.

---

## How To Tell A Board Is Working Well

You should see this pattern:

1. the chat creates a strong candidate
2. the draft preview renders the right widgets
3. Live Source Review does not show blocking issues
4. the saved artifact is easy to reopen and understand
5. the published version is stable enough to keep
6. the live board refreshes with useful current data while it is open

That is when a board stops being a demo and becomes a real operating surface.

---

## When To Use Boards Instead Of Other Surfaces

Use **Boards** when you need a custom page for one decision or workflow.

Use **Overview** when the default company dashboard already answers the question.

Use **Files** when the real task is file browsing or file management, not a dashboard.

Use **Plan** or **Tasks** when the real job is changing the work itself, not visualizing the state of that work.

Use **Train** when the real job is improving task behavior rather than building a live artifact page.

---

## Related Docs

- [Agents](./agents.md)
- [Auto, Company Dashboard](./auto-companies-dashboard.md)
- [Auto, Autonomous Companies](./auto-companies.md)
- [Auto, Apps, Integrations & Files](./auto-companies-apps-integrations-and-files.md)
- [Auto, Training Mode](./auto-companies-training.md)

---

### [Channels](https://neotask.ai/docs/channels)

## What Are Channels?

Channels are the communication platforms your Neotask agents use to send and receive messages. By connecting a channel, you give an agent the ability to listen for incoming messages on that platform and respond in real time.

Key concepts:

- **Per-agent configuration** -- Each channel connection is configured at the agent level. One agent can be connected to Slack while another handles Discord, or a single agent can be connected to multiple channels simultaneously.
- **Bidirectional** -- Channels support both sending and receiving messages. Agents can respond to incoming messages, initiate outbound messages, or both.
- **Multi-account** -- You can connect multiple accounts for the same channel type (e.g., two different Slack workspaces or several email addresses).

## Supported Channels

| Channel | Auth Method | Features |
|---------|-------------|----------|
| **Slack** | OAuth | Full messaging, channels, threads, reactions, file uploads |
| **Discord** | Bot token | Server messaging, channels, DMs, embeds, reactions |
| **WhatsApp** | QR code | Text messages, media messages, group chats |
| **Telegram** | Bot token | Text messages, groups, channels, inline keyboards, file sharing |
| **Signal** | Local app | End-to-end encrypted messaging, group chats |
| **Microsoft Teams** | OAuth | Teams, channels, chats, adaptive cards, file sharing |
| **Google Chat** | OAuth | Spaces, direct messages, cards, threaded conversations |
| **SMS** | API key | Outbound and inbound text messaging via supported providers |
| **Email / Gmail** | OAuth | Send, receive, draft, reply, attachments, label management |
| **Matrix** | Token | Federated messaging, rooms, end-to-end encryption |
| **Line** | Token | Text messages, stickers, group chats, rich menus |
| **Zalo** | Token | Text messages, media sharing |
| **IRC** | Config | Server connections, channels, private messages |
| **iMessage** | Local | macOS-native messaging (requires macOS host) |

## Connecting a Channel

The connection process varies depending on the channel's authentication method.

1. Open **Agents** and select the agent that will use the channel.
2. Open its **Channels** tab.
3. Select **Add Channel**.
4. Choose the channel from the connection picker.

![Add Channel dialog showing the available messaging channels](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/channel-picker-1120.webp)

### OAuth Flow

Used by: **Slack, Microsoft Teams, Google Chat, Email / Gmail**

1. Open the agent's **Channels** tab.
2. Click **Connect** next to the channel you want to add.
3. You will be redirected to the service's authorization page.
4. Sign in with your account and grant Neotask the requested permissions.
5. After authorization, you are redirected back to the dashboard and the channel appears as connected.

For mobile platforms, a **QR code option** is available. Scan the QR code from your phone to complete the authorization without needing to sign in on your desktop browser.

### Bot Token

Used by: **Discord, Telegram**

1. Create a bot on the respective platform:
   - **Discord** -- Create an application at the Discord Developer Portal and generate a bot token.
   - **Telegram** -- Message [@BotFather](https://t.me/BotFather) on Telegram to create a bot and receive a token.
2. Open the agent's **Channels** tab and click **Connect** next to the channel.
3. Paste the bot token into the token field.
4. Click **Save** to establish the connection.

### QR Code

Used by: **WhatsApp**

1. Open the agent's **Channels** tab and click **Connect** next to WhatsApp.
2. A QR code will appear on screen.
3. Open WhatsApp on your phone, go to **Linked Devices**, and scan the QR code.
4. The connection is established once your phone confirms the link.

### API Key

Used by: **SMS**

1. Obtain an API key from your SMS provider (e.g., Twilio, Vonage).
2. Open the agent's **Channels** tab and click **Connect** next to SMS.
3. Enter your API key and any additional required fields (e.g., phone number, account SID).
4. Click **Save** to activate the channel.

### Token

Used by: **Matrix, Line, Zalo**

1. Generate an access token from the respective platform's developer settings or admin panel.
2. Open the agent's **Channels** tab and click **Connect** next to the channel.
3. Paste the token into the token field.
4. Click **Save** to establish the connection.

### Server Configuration

Used by: **IRC**

1. Open the agent's **Channels** tab and click **Connect** next to IRC.
2. Enter the server address, port, nickname, and (optionally) a password.
3. Specify the channels to join on connect.
4. Click **Save** to connect.

### Local Application

Used by: **Signal, iMessage**

1. Ensure the required application is installed and running on the host machine:
   - **Signal** -- Signal Desktop must be installed and linked to your Signal account.
   - **iMessage** -- Requires a macOS host with Messages.app configured.
2. Open the agent's **Channels** tab and click **Connect** next to the channel.
3. Neotask will detect the local application and establish a connection.
4. If the application is not detected, you will be prompted to install or configure it.

### Connection Testing

After connecting any channel, Neotask automatically runs a **latency probe** to verify the connection:

- A test message is sent and received through the channel's API.
- Round-trip latency is measured and displayed.
- If the test fails, an error message describes the issue (e.g., invalid token, expired OAuth grant, unreachable server).

You can re-run the connection test at any time from the channel's settings.

After the account is connected, select **Test Channel**. A working account shows **Connected** and **Passed** in its row.

![Slack channel accounts showing connected and passed test states](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/channel-tests-1280.webp)

## Channel Operations

Once a channel is connected, you can manage it from the agent's **Channels** tab.

### Enable / Disable

Toggle a channel on or off without disconnecting it. When disabled:

- The agent stops listening for messages on that channel.
- No outbound messages are sent through it.
- The connection credentials are preserved so you can re-enable at any time without reconfiguring.

### Limit Slack access

1. Select **Edit Settings** for the Slack connection.
2. Under **Channel access**, choose **All channels** or **Specific**.
3. Under **DM access**, choose **Everyone** or **Specific**.
4. Select the approved channels and people.
5. Select **Save & Exit**, or use **Save & Continue** to review the remaining Slack settings.

![Slack settings showing specific channel and direct-message access](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/slack-access-1280.webp)

### Test Connectivity

Run an on-demand connectivity test to verify the channel is reachable and responsive. The test:

- Sends a lightweight API request to the channel's service.
- Measures round-trip latency in milliseconds.
- Reports the result as **Connected** (with latency) or **Failed** (with error details).

### View Latency

Each connected channel displays its most recent latency reading on the **Channels** tab. This helps you identify slow or degraded connections at a glance. Latency is updated:

- Automatically after each connection test.
- Periodically in the background while the channel is active.

### Disconnect / Logout

To fully remove a channel connection:

1. Click **Disconnect** next to the channel.
2. Confirm the action in the dialog.
3. Neotask revokes or discards the stored credentials (OAuth tokens are revoked on the provider side where supported).
4. The channel is removed from the agent's active channels list.

To reconnect after disconnecting, follow the original connection steps again.

---

### [Compliance Overview](https://neotask.ai/docs/compliance-overview)

## Controls Available In The Product

Company administrators can review organization identity, directory sync, privacy choices, approval history, and session records from the product.

![Enterprise organization and directory sync status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/enterprise-sso-1280.webp)

Where Neotask stands on security and privacy compliance, in plain language. We build to four frameworks, **SOC 2**, **GDPR**, **ISO/IEC 27001**, and **HIPAA** (for healthcare customers, under a BAA). For technical detail see [Trust & Security](trust-and-security); for your data rights see [Data Privacy & Your Rights](data-privacy-and-your-rights).

## SOC 2

Neotask has built its controls to the **SOC 2** Trust Services Criteria, **Security, Availability, Confidentiality, Privacy, and Processing Integrity**. We maintain a complete control framework (access control, encryption and key management, change management, monitoring, incident response, backup/DR, vendor management, and more), each backed by working controls in our product and a documented evidence trail.

- **Status:** SOC 2 Type I readiness. Our controls are designed and implemented; a Type II observation period (controls operating over time) is the next step.
- **Reports:** A SOC 2 report is issued by an independent auditor. If you're evaluating Neotask and need our current report or readiness package, we can share it **under NDA**, contact **compliance@neotask.ai**.

## GDPR & data protection

Optional session-data sharing is controlled in **Settings → Privacy and Data**.

![Session data sharing control](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-consent-1280.webp)

We process personal data in accordance with the **EU General Data Protection Regulation (GDPR)** and apply the same protections globally:

- **Lawful, minimal processing**, we collect only what we need to provide the service.
- **Your rights**, access, export (portability), correction, deletion, and the ability to object or withdraw consent. See [Data Privacy & Your Rights](data-privacy-and-your-rights).
- **Subprocessors & transfers**, we maintain a published [subprocessor list](subprocessors) and put data-processing terms in place with our vendors.
- **Data Processing Agreement (DPA)**, a DPA is **available to customers on request** at **compliance@neotask.ai**.
- **Breach handling**, we maintain an incident-response process designed to meet GDPR's notification requirements.

## ISO/IEC 27001

We operate an **Information Security Management System (ISMS)** aligned to **ISO/IEC 27001:2022**. We maintain a full Statement of Applicability covering all 93 Annex A controls, a risk register and treatment plan, and the management-system documentation the standard requires (context, leadership, planning, support, operation, performance evaluation, and improvement).

- **Status:** ISMS built and Stage 1-ready. ISO 27001 certification is granted by an **accredited registrar** after a Stage 1 (documentation) and Stage 2 (implementation) audit plus an operating period; we are preparing for that external audit. We are **not yet certified**.
- **Documentation:** Our ISMS scope, Statement of Applicability, and readiness package can be shared **under NDA**, contact **compliance@neotask.ai**.

## HIPAA

Actions that require a person remain visible in the approval record with the requested operation and review controls.

![Approval detail and review controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-detail-1280.webp)

For healthcare customers, Neotask operates as a **HIPAA Business Associate** and will **sign a Business Associate Agreement (BAA)**. The HIPAA Security Rule safeguards (access control, encryption at rest and in transit, audit logging, integrity, transmission security, automatic logoff, and minimum-necessary handling) are built into the product. When a healthcare customer is onboarded under a BAA, their protected health information (PHI) is handled in a **dedicated, isolated, BAA-covered environment** with a BAA-covered AI provider, it never flows to non-BAA subprocessors.

- **Status:** HIPAA-aligned Business Associate posture, available to healthcare customers under a signed BAA. There is **no such thing as "HIPAA certified"**, HIPAA is enforced by regulation (OCR) and by the BAA contract, not by a certificate.
- **To engage:** contact **compliance@neotask.ai** to start a BAA and healthcare onboarding.

## Hosting & subprocessors

We rely on established infrastructure and service providers, each with their own security/compliance programs. The current list of subprocessors that may process customer data is published on the [Subprocessors](subprocessors) page and is kept up to date; material changes are communicated per your agreement.

## How to request documentation

| You need | Contact | Notes |
|---|---|---|
| SOC 2 report / readiness package | compliance@neotask.ai | Shared under NDA |
| ISO 27001 ISMS scope / SoA / readiness | compliance@neotask.ai | Shared under NDA |
| HIPAA BAA + healthcare onboarding | compliance@neotask.ai | For healthcare customers |
| Data Processing Agreement (DPA) | compliance@neotask.ai | For customers / prospects |
| Security questionnaire | security@neotask.ai | We'll complete standard questionnaires |
| Privacy / data-rights request | privacy@neotask.ai | See Data Privacy & Your Rights |

*This page describes our compliance posture; it is not itself a certification. SOC 2 reports are issued by an independent CPA firm; ISO/IEC 27001 certification is issued by an accredited registrar; GDPR compliance is demonstrated through our documentation and practices; and HIPAA is enforced by regulation and by the Business Associate Agreement, we can evidence each on request.*

---

### [Continue Work After Restarts](https://neotask.ai/docs/continue-work-after-restarts)

Neotask can pick your work back up after a restart. When this is on, interrupted scheduled runs start again, interactive agent threads continue where they stopped, and coding sessions reopen in the branch they were working in. You do not have to remember what was in flight.

The setting is on by default. It is account-wide, so it follows you to every device you sign in on, and it applies to every company. A single company can deviate when you want that company to behave differently.

## What It Is

An interruption is anything that stops Neotask before its work finishes:

- An app update.
- A Neotask update.
- An operating-system restart.
- A crash.
- A power loss.

With **Continue work after restarts** on, Neotask records what was interrupted, then resumes it after the restart. Work that already finished is never repeated. Scheduled runs resume a bounded number of times, so a run that keeps being interrupted does not retry forever.

With the setting off, Neotask still records what was interrupted. You see it in the activity feed, but nothing starts again on its own. You resume by opening the thread or the task yourself.

## Turn It On Or Off

1. Open the Neotask desktop app and select **Settings**.
2. Select **Automations**.
3. Turn **Continue work after restarts** on or off.

The choice is saved to your account and synced to every device you sign in on.

Changing this account-wide toggle re-syncs every company to the new value and clears any company override. A company differs again only after you change the toggle inside that company's own settings.

## Per-Company Override

One company can behave differently from your account setting.

1. Open the company dashboard.
2. Select **Settings**.
3. Turn **Continue work after restarts** on or off for this company.

This overrides the account-wide value for that company only. Every other company keeps the account setting.

A company override is useful when one company runs work you would rather restart by hand, for example while you are testing a new process.

## What Resumes And What Does Not

When the setting is on, Neotask resumes:

- **Scheduled task runs** that were interrupted. Each run resumes a bounded number of times, and work that already finished is never repeated.
- **Interactive agent threads.** The agent receives a short "Continue where you left off." nudge on the same session. Your own message is never sent a second time.
- **Coding sessions.** Each session reopens in its own branch working copy.

Neotask also cleans up after the interrupted session. Development servers and commands left running by a previous session are stopped or adopted safely, so nothing is left holding a port or running twice.

Blog posts and emails are protected against duplicates. If a run is retried, Neotask does not publish the same post or send the same email again.

By design, Neotask never resumes these on its own:

- **Anything waiting for your approval.** It keeps waiting for you. A restart never approves an action.
- **Runs that already finished.**
- **Runs that failed for a reason other than the restart.** Those need the underlying problem fixed, not another attempt.

## Launch Neotask At Login

**Launch Neotask at login** starts Neotask when you sign in to your computer. It is off by default and sits on the same **Automations** screen.

1. Open **Settings**.
2. Select **Automations**.
3. Turn on **Launch Neotask at login**.

Not everything needs the app open. Scheduled tasks and the Neotask service keep running with the app closed. Company planning, new task creation, and the nightly and weekly reviews run only while the app is open. With **Launch Neotask at login** on, that work continues right after a restart instead of waiting until you next open Neotask.

**Launch Neotask at login** is supported on macOS, Windows, and Linux.

## Windows Notes

During shutdown, Windows may briefly show **Neotask is saving agent work**. That message means Neotask is saving the state of work that is in progress so it can be resumed. Let it finish.

If Windows restarts to finish an update, it can reopen Neotask when you sign back in. This depends on your Windows sign-in options allowing apps to restart.

## Coding Sessions

Neotask Code keeps each coding session tied to its own place in your repository, so a restart never puts your work on the wrong branch.

- Every session remembers its repository, its branch, and its working copy.
- Every turn runs inside that session's working copy.
- If the working copy folder is missing, Neotask repairs it before running. If it cannot repair it, it stops with a clear message instead of running somewhere else.
- Working copy registrations that no longer point at a real folder are listed for you instead of hidden.
- Cleanup never deletes uncommitted work.
- A development server from a previous run is stopped or adopted safely. It is never left holding the port.

For more about branches and working copies, see [Neotask Code: Git and GitHub](/docs/neotask-desktop/neotask-code-git).

## Troubleshooting

### Find What Was Interrupted

The activity feed records both sides of a restart:

- **Interrupted by restart** marks work that stopped.
- **Resumed after restart** marks work that started again.

If you see **Interrupted by restart** with no matching **Resumed after restart**, check whether the setting is on for that company.

### A Thread Shows "Send A Message To Continue"

Some threads cannot be continued automatically, for example when the assistant session has expired. The thread shows **Send a message to continue** instead of resuming on its own. Send a message in that thread to continue working.

### Nothing Resumed After A Restart

1. Open **Settings**, select **Automations**, and confirm **Continue work after restarts** is on.
2. Open the company, select **Settings**, and confirm that company is not overriding the setting to off.
3. Check the activity feed for **Interrupted by restart** entries. Work that is waiting for your approval is recorded there but is never resumed automatically.
4. Confirm the app was open. Company planning, new task creation, and the nightly and weekly reviews need the app running. Turn on **Launch Neotask at login** so Neotask opens with your computer.

### A Run Stopped Retrying

Scheduled runs resume a bounded number of times. A run that keeps being interrupted stops retrying and stays in the activity feed for you to review. Run it again once the cause is fixed.

## Related

- [Automation & Scheduling](automation.md)
- [Neotask Desktop Application](desktop-app.md)
- [Auto, Operations, Controls & Settings](auto-companies-operations-controls-and-settings.md)
- [Neotask Code](/docs/neotask-desktop/neotask-code)
- [Neotask Code: Git and GitHub](/docs/neotask-desktop/neotask-code-git)
- [Troubleshooting](troubleshooting.md)

---

### [Neotask Web Dashboard](https://neotask.ai/docs/dashboard)

## Find Work Across The Dashboard

1. Open **Agents** for personal and team agent work.
2. Open **Companies** for company plans, activity, and approvals.
3. Open **Sessions** for conversation and tool history.
4. Open **Usage** for cost and model activity.

![Agent dashboard overview](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

## Overview

Neotask Web Dashboard is a browser-based control center for managing your agents, monitoring sessions, tracking usage, and configuring your workspace. Access the dashboard at your assigned dashboard URL after authenticating with your license key.

The dashboard provides a centralized interface for every aspect of your Neotask deployment, including agent configuration, session monitoring, team orchestration, usage analytics, and account settings.

---

## Authentication

### License Key Login

To access the dashboard, enter your license key on the login screen. License keys follow this format:

```
NT-XXXXXXXX-XXXXXXXX-XXXXXXXX-XXXXXXXX
```

After successful authentication, you receive a session token that grants access to the dashboard for the duration of your session.

### Two-Factor Authentication (TOTP)

Two-factor authentication is optional but strongly recommended for all accounts.

**Setting up two-factor authentication:**

1. Navigate to **Settings > Security**.
2. Select **Enable Two-Factor Authentication**.
3. Scan the displayed QR code with an authenticator app such as Google Authenticator, Authy, or any TOTP-compatible application.
4. Enter the 6-digit verification code from your authenticator app to confirm setup.
5. Download and securely store the backup codes provided. These codes allow you to regain access if you lose your authenticator device.

Once enabled, a 6-digit code from your authenticator app is required on every subsequent login.

To disable two-factor authentication, go to **Settings > Security** and select the option to disable it.

---

## Home Page

The Home page serves as your dashboard overview, presenting a high-level summary of your workspace activity at a glance.

Open a company to reach Mission Control, where live work and company metrics share one operating view.

![Company Mission Control dashboard](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/mission-control-1440.webp)

### Agent Spending Graph

A bar chart displaying costs broken down by agent. Use this to identify which agents are consuming the most resources.

### Session Summary

Displays counts of your sessions organized by status:

- **Active** -- Sessions currently in progress.
- **Failed** -- Sessions that encountered errors.
- **Completed** -- Sessions that finished successfully.

### Latest Sessions

A list of the 5 most recent sessions, including token usage for each. Select any session to view its full details.

### Model Usage Breakdown

Shows which AI models are being used across your agents, helping you understand model distribution and cost implications.

### Upcoming Tasks

A calendar view of your scheduled automations, giving you visibility into what jobs are coming up next.

### Quick Widgets

Compact widgets providing fast access to key functions:

- **Agents** -- Displays agent count with an auto-deploy toggle for enabling or disabling automatic agent deployment.
- **Usage** -- Displays usage summary with an auto-track toggle for enabling or disabling automatic usage tracking.
- **Sessions** -- Displays current session activity.

---

## Agents Page

The Agents page is where you manage all of your AI agents. It is organized into tabbed sub-pages, each focused on a different aspect of agent management.

### Overview Tab

The Overview tab presents a comprehensive look at your agents and their activity.

- **Agent Statistics** -- View each agent's name, assigned model, session count, total tokens consumed, and current status.
- **Spending Chart** -- Select an agent to see a detailed spending chart for that agent.
- **Session Breakdown** -- A breakdown of session activity for the selected agent.
- **Upcoming Jobs** -- A list of scheduled jobs associated with the selected agent.
- **Mini Calendar** -- A compact calendar highlighting dates with scheduled events.

### Chat Tab

The Chat tab provides a read-only view of agent conversations.

- **Session Sidebar** -- Browse a list of sessions in the sidebar and select one to view.
- **Message Bubbles** -- Conversation messages are displayed as chat bubbles with role indicators:
  - **User** -- Messages sent by the user.
  - **Assistant** -- Responses from the AI agent.
  - **System** -- System-level messages and instructions.
  - **Tool** -- Messages generated by tool calls.
- **Timestamps** -- Each message displays a relative timestamp (for example, "2 minutes ago").

### Skills Tab

The Skills tab allows you to browse, install, and manage agent skills.

- **Catalog** -- Browse all available skills. Each skill listing includes the version number, a description, and a link to its homepage.
- **My Skills** -- View your installed skills and use the toggle to enable or disable each one.
- **Category Organization** -- Skills are organized by category for easy browsing.
- **Search Filtering** -- Use the search bar to filter skills by name or keyword.

### Channels Tab

The Channels tab displays all connected communication channels.

- **Status Badges** -- Each channel displays a status badge indicating its current state: **Connected**, **Error**, **Pending**, or **Disconnected**.
- **Add New Channels** -- Use the add channel option to connect a new communication platform.
- **Filter by Type** -- Filter the channel list by channel type to find specific connections.

### Integrations Tab

The Integrations tab manages external service connections and API configurations.

**Filtering options:**

- All
- Google
- OAuth
- API Key

**Google Workspace services:**

The following Google Workspace services can be connected and managed:

- Gmail
- Google Calendar
- Google Drive
- Google Sheets
- Google Docs
- Google Meet

Each Google service can be toggled on or off individually.

**API configuration:**

- Configure API keys for third-party services.
- Set environment variables required by integrations.

### Automations Tab

The Automations tab provides tools for scheduling and managing automated jobs.

- **Calendar View** -- A calendar displaying all scheduled automation jobs.
- **Job List** -- A list view showing each job's name, schedule, current status, and next run time.
- **Run History** -- View the execution history of each automation with status indicators showing success or failure.
- **Create and Edit** -- Use the modal dialogs to create new automations or edit existing ones.

### Tools Tab

The Tools tab displays all available tools organized by category. Each tool can be individually toggled on or off.

**File System tools:**

- Read files
- Write files
- Edit files

**Execution tools:**

- Shell commands
- Process management
- Subprocess agents

**Web tools:**

- Browser
- HTTP requests

**Scheduling tools:**

- Cron jobs
- Webhooks

---

## Sessions Page

The Sessions page provides detailed monitoring of all agent sessions across your workspace.

### KPI Boxes

At the top of the page, key performance indicator boxes display:

- **Total Sessions** -- The total number of sessions.
- **Active** -- Sessions currently running.
- **Tool Calls** -- The total number of tool calls made across sessions.
- **Average Duration** -- The average length of a session.
- **Failed** -- The number of sessions that ended in failure.

### Token Usage Chart

A multi-line graph showing token usage over time, with a separate line for each agent. Use this to identify trends and spikes in token consumption.

### Breakdowns

- **Token Usage by Agent** -- A breakdown showing how tokens are distributed across your agents.
- **Tool Calls by Agent** -- A breakdown showing tool call volume per agent.

### Searchable Session Table

A table listing all sessions with the following columns:

| Column | Description |
|---|---|
| Agent | The agent that ran the session |
| Session Key | The unique session identifier |
| Tokens (Input) | Number of input tokens consumed |
| Tokens (Output) | Number of output tokens generated |
| Tokens (Total) | Combined input and output tokens |
| Cost | The monetary cost of the session |
| Started | When the session began |
| Status | Current session status |

### Session Detail Panel

Select any session from the table to open the detail panel, which displays:

- Full session metadata (agent, model, timestamps, token counts, cost).
- Complete session entries with role indicators for each message.

### Filters

Narrow down the session list using:

- **Status Filter** -- All, Active, Completed, or Failed.
- **Search** -- Search by agent name or session key.
- **Date Range** -- Predefined ranges including Today, Week, and Month.

---

## Teams Page

The Teams page enables multi-agent orchestration, allowing you to group agents into teams and run coordinated tasks.

### Layout

The Teams page uses a three-column layout:

1. **Agents List** -- All available agents that can be assigned to teams.
2. **Teams List** -- Your created teams with their configurations.
3. **Task Runs** -- History and status of task executions.

### Creating a Team

To create a new team:

1. Select **Create Team**.
2. Enter a **name** and **description** for the team.
3. Select the **agents** to include in the team.
4. Choose an **orchestration mode**:
   - **Parallel** -- All agents execute simultaneously.
   - **Sequential** -- Agents execute one after another in order.
   - **Hybrid** -- A combination of parallel and sequential execution.

### Task Runs

- Execute prompts against a team to start a task run.
- View the run history with status updates for each execution.

### Controls

- **Start/Stop Runs** -- Begin or halt task executions.
- **Delete Teams** -- Remove teams that are no longer needed.

---

## Usage Page

The Usage page provides comprehensive cost and token tracking across your entire workspace.

### KPI Boxes

Summary metrics displayed at the top of the page:

- **Total Tokens and Cost** -- Combined token count and total monetary cost.
- **Input Breakdown** -- Total input tokens and their associated cost.
- **Output Breakdown** -- Total output tokens and their associated cost.
- **Days Tracked** -- The number of days included in the tracking period.

### Cost Chart

A detailed cost breakdown over time. Use the range selector to adjust the view:

- Today
- Day
- Week
- Month
- Year
- Custom date range

### Model Breakdown

Displays the top 8 models ranked by cost, showing how spending is distributed across different AI models.

### Token Distribution

A percentage breakdown of input tokens versus output tokens, giving you insight into the balance of your token consumption.

### Agent Breakdown

Displays the top 8 agents ranked by token usage, with corresponding cost figures for each agent.

### Daily Table

A tabular view of usage data over the last 30 days with the following columns:

- Date
- Tokens
- Input tokens
- Output tokens
- Cost

### Export

Select **Export** to download your usage data as a CSV file for external analysis or record-keeping.

---

## Downloads Page

The Downloads page provides access to the Neotask desktop application.

### Available Downloads

- **macOS** -- Desktop app download available now.
- **Windows** -- Coming soon.

### Quick Start Installation Guide

1. Download the desktop app for your platform.
2. Install the application on your machine.
3. Log in with your license key to get started.

---

## Settings

Access Settings from the avatar dropdown menu in the navigation bar.

### Account

View your account details:

- **Tenant ID** -- Your unique workspace identifier.
- **License Status** -- Whether your license is active.
- **Plan** -- Your current subscription plan.
- **Expiry Date** -- When your license expires.
- **Agent Capacity:** Agent count is not capped on any plan; this reflects your current plan's capabilities.

### Security

Manage authentication and security settings:

- Enable or disable TOTP two-factor authentication.
- View and regenerate backup codes.

### About

View version information for the Neotask dashboard and related components.

### Language

Select your preferred display language. The dashboard supports 20 languages.

---

## Onboarding Wizard

New users are guided through an 11-step onboarding wizard to set up their workspace.

### Step 1: Select Agent Template

Choose from 9 available agent templates to get started with a preconfigured agent setup.

### Step 2: Configure Agent

Set your agent's **name** and select the AI **model** it will use.

### Step 3: Set Daily Budget

Define a daily spending limit to control costs for your agent.

### Step 4: Connect Channels

Connect your agent to communication platforms. 10 platforms are available for integration.

### Step 5: Connect Google Workspace

Link up to 12 Google Workspace services, including Gmail, Calendar, Drive, Sheets, Docs, and Meet.

### Step 6: Developer and Code Features

Configure developer-oriented features and code execution capabilities for your agent.

### Step 7: Built-in Skills Overview

Review the 55+ built-in skills available for your agent and select which ones to enable.

### Step 8: Automations Setup

Set up scheduled automations and recurring jobs for your agent.

### Step 9: Voice Control Configuration

Configure voice control settings to interact with your agent using voice commands.

### Step 10: Notifications

Set up notification delivery through Discord webhooks to stay informed about agent activity.

### Step 11: Download Desktop App

Download and install the Neotask desktop application for direct access from your computer.

---

### [Data Privacy & Your Rights](https://neotask.ai/docs/data-privacy-and-your-rights)

## Find The Data Controls

1. Open **Settings → Privacy and Data**.
2. Review the main session-data control.
3. Review company-specific controls when your account contains company workspaces.
4. Save only after the selected scope is correct.

![Session data sharing control](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-consent-1280.webp)

Neotask is built so you stay in control of your data. This page summarizes the personal data we process and the rights you can exercise. The full legal detail is in our [Privacy Policy](https://neotask.ai/privacy-policy). For our security controls see [Trust & Security](trust-and-security).

## What we process

- **Account data**, your email, name, and login identifiers.
- **Billing data**, subscription and payment references (processed by Stripe; we don't store full card numbers).
- **Product content**, the prompts, tasks, files, and configurations you create in the app.
- **Connected-service data**, only what you authorize when you connect an integration (e.g. Google, Slack, X).
- **Usage & diagnostics**, feature usage and error diagnostics, with personal data minimized and IP addresses anonymized.

We collect only what we need to run the service, and we do not sell your personal data.

## Your rights (GDPR Articles 15, 22)

You can exercise these rights at any time:

| Right | What it means | How |
|---|---|---|
| **Access** | Get a copy of the personal data we hold about you | Request export below |
| **Portability** | Receive your data in a portable format | Export below |
| **Rectification** | Correct inaccurate data | Edit in-app or contact us |
| **Erasure ("right to be forgotten")** | Delete your account and associated data | In-app account deletion, or contact us |
| **Restriction / Objection** | Limit or object to certain processing | Contact us |
| **Withdraw consent** | Turn off optional processing (e.g. analytics) | Tracking honors Do-Not-Track / Global Privacy Control; manage in settings |

### How to exercise them

- **In the app:** delete your account and connected data from **Account → Danger Zone → Delete Account**; export your data from your account settings.
- **By email:** **privacy@neotask.ai**. We verify your identity and respond within **one month** (extendable for complex requests, as GDPR allows).

## Deletion & retention

- When you delete your account, we remove your personal data across all of our systems and revoke the access tokens for any services you connected, except where we must retain limited records to meet legal, tax, or security obligations.
- We keep operational data only as long as needed and apply documented retention periods; financial and audit records follow their required statutory retention.

## Consent & tracking

Company workspaces have their own data controls below the account-level choice. Review the company list before changing a company-specific setting.

![Company-specific data controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-company-controls-1280.webp)

- Analytics and non-essential tracking respect **Do-Not-Track** and **Global Privacy Control** signals and are not collected when you opt out.
- We anonymize IP addresses before any geo-lookup and minimize what location data we keep.

## International transfers & subprocessors

We use the vendors listed on the [Subprocessors](subprocessors) page. Where data is transferred outside your region, we rely on appropriate safeguards (such as Standard Contractual Clauses). A Data Processing Agreement is available to customers on request at **compliance@neotask.ai**.

## Contact

- Privacy / data-rights requests: **privacy@neotask.ai**
- Security: **security@neotask.ai**
- Compliance / DPA / reports: **compliance@neotask.ai**

---

### [Neotask Desktop Application](https://neotask.ai/docs/desktop-app)

## Check The Desktop Connection

1. Open **Settings**.
2. Select **Gateway**.
3. Confirm the connection path and health status.
4. Open **Preferences** before changing the local runtime or launch behavior.

![Desktop Gateway connection path](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-connection-path-1280.webp)

Neotask desktop application is your central hub for managing AI agents, voice control, and automated workflows. It runs natively on macOS, Windows, and Linux, with a built-in local gateway that keeps your data on your machine.

---

## Installation

### macOS

1. Download the `.dmg` installer from your Neotask account dashboard.
2. Open the DMG file and drag **Neotask** into your **Applications** folder.
3. Launch Neotask from Applications or Spotlight.

Neotask supports both **Intel** and **Apple Silicon** (M1/M2/M3/M4) Macs with a universal binary. No Rosetta translation required on Apple Silicon.

### Windows

1. Download the `.exe` installer from your Neotask account dashboard.
2. Run the installer (built with NSIS).
3. Choose your installation directory or accept the default (`C:\Program Files\Neotask`).
4. Follow the on-screen prompts to complete installation.
5. Launch Neotask from the Start Menu or desktop shortcut.

### Linux

1. Download the `.AppImage` file from your Neotask account dashboard.
2. Make the file executable:
   ```bash
   chmod +x Neotask-*.AppImage
   ```
3. Run the AppImage:
   ```bash
   ./Neotask-*.AppImage
   ```

No installation or root privileges required. The AppImage is self-contained and runs on most modern Linux distributions.

---

## License Activation

When you first launch Neotask, you will be prompted to enter your license key.

### License Key Format

License keys follow this format:

```
NT-XXXXXXXX-XXXXXXXX-XXXXXXXX-XXXXXXXX
```

where each `X` is an alphanumeric character. You receive your license key via email after purchase.

### Device Binding

Each license key is bound to **one device**. Activating your license on a new device will deactivate it on the previous one. If you need to transfer your license, deactivate it from the current device first through **Settings > License**.

### Offline Grace Period

Neotask requires periodic online validation. If your machine goes offline, the application continues to work for up to **72 hours** before requiring a connection to revalidate.

### Auto-Revalidation

Your license is automatically revalidated every **6 hours** in the background. This process is silent and requires no user interaction as long as you have an active internet connection.

### License States

| State | Description |
|---|---|
| **Valid** | License is active and in good standing. Full access to all features. |
| **Expired** | License term has ended. Renew your subscription to restore access. |
| **Revoked** | License has been revoked by the issuer. Contact support for details. |
| **Blocked Version** | Your application version is no longer supported. Update to the latest version to continue. |

---

## Interface Overview

### Dashboard Home

The home screen provides an at-a-glance summary of your Neotask workspace:

- **Metrics**: Token usage, session count, active agent count, and budget consumption for the current billing period.
- **Agent Overview**: A list of your agents with their current status (running, idle, or stopped).
- **Recent Sessions**: Quick access to your most recent agent sessions and conversations.

### Navigation

The main navigation menu includes the following sections:

| Section | Description |
|---|---|
| **Home** | Dashboard with metrics, agent overview, and recent activity. |
| **Agents** | Create, configure, and manage your AI agents. |
| **Sessions** | View session history, transcripts, and logs. |
| **Teams** | Manage team members, roles, and shared agent access. |
| **Usage** | Detailed usage analytics, token consumption, and cost breakdowns. |

### System Tray

Neotask runs in the system tray (menu bar on macOS) when minimized. The tray icon includes a **gateway status indicator**:

- **Green**: Gateway is running and healthy.
- **Yellow**: Gateway is starting up or recovering.
- **Red**: Gateway is stopped or has encountered an error.

Right-click the tray icon for quick actions: open the dashboard, start/stop the gateway, or quit the application.

### Theme

Neotask uses a **dark theme** interface designed for extended use with reduced eye strain.

---

## Gateway

The gateway is a bundled local runtime that powers agent execution, tool calls, and voice processing. It runs entirely on your machine; no cloud servers are involved in gateway operations.

### How It Works

- The gateway binds exclusively to **loopback** (`127.0.0.1`), meaning it is only accessible from your local machine. There is **no external network exposure**.
- All agent execution, tool orchestration, and local processing happen through the gateway.

### Controls

You can manage the gateway from the dashboard or the system tray:

| Action | Description |
|---|---|
| **Start** | Launch the gateway process. |
| **Stop** | Gracefully shut down the gateway. |
| **Restart** | Stop and immediately restart the gateway. |

### Auto-Recovery Watchdog

Neotask includes a watchdog process that monitors gateway health. If the gateway crashes or becomes unresponsive, the watchdog automatically restarts it. No user intervention is required.

Beyond process crashes, Neotask also detects a **stalled or black-holed connection**, a gateway that is still running but has stopped responding, through a continuous connection-liveness check. When this happens, the app surfaces a clear "reconnecting" state and re-establishes the connection on its own instead of waiting indefinitely. Requests that were in flight return a prompt "gateway unavailable, retrying" rather than hanging. In-progress AI responses and pending approvals are never interrupted by this recovery.

### Health Monitoring

The gateway continuously reports its health status. You can view the current state in the system tray icon or on the dashboard home screen. Health checks include process status, memory usage, and responsiveness.

---

## Settings & Preferences

All user-configurable settings are accessible from **Settings** in the navigation menu.

Use the Gateway preferences screen when you need to change how the local runtime starts or connects. Check the displayed connection path before changing a port or launch setting.

![Gateway preferences in the desktop app](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-preferences-1280.webp)

| Setting | Description | Default |
|---|---|---|
| **Language** | Interface and voice language. 21 languages supported. | English |
| **Voice** | Select a text-to-speech voice from the ElevenLabs voice library. | System default |
| **Wake Word** | Custom phrase that activates voice listening (e.g., "Hey Neotask"). | "Hey Neotask" |
| **Wake Mode** | How voice activation is triggered: **Always-listening** (wake word), **Keyboard shortcut**, or **Off**. | Off |
| **Activation Sound** | Play an audio cue when voice activation is triggered. | On |
| **Safe Mode** | Require explicit approval before executing sensitive actions (file deletion, deployments, sending messages). | On |
| **Daily Budget** | Maximum daily spend in USD across all agent activity. | No limit |
| **Auto-Shutoff** | Automatically stop all agents when the daily budget is exceeded. | Off |
| **Preferred Code Editor** | Code editor to open when file editing is requested (e.g., VS Code, Cursor, Sublime Text). | VS Code |
| **Default AI Model** | The AI model used for agent reasoning and conversation. | GPT-4o |

### Automations

The **Automations** screen in **Settings** holds two account-wide controls for what Neotask does around a restart.

| Setting | Description | Default |
|---|---|---|
| **Continue work after restarts** | Resume interrupted scheduled runs, agent threads, and coding sessions after an update, an operating-system restart, a crash, or a power loss. | On |
| **Launch Neotask at login** | Start Neotask when you sign in to your computer. Supported on macOS, Windows, and Linux. | Off |

Both settings are saved to your account and synced to every device you sign in on. **Continue work after restarts** applies to every company, and a single company can override it in that company's own **Settings**.

For what resumes, what never resumes, and the Windows shutdown message, see [Continue Work After Restarts](continue-work-after-restarts.md).

---

## Keyboard Shortcuts

### Voice Activation

| Platform | Shortcut | Description |
|---|---|---|
| macOS | `Cmd + Shift + Space` | Activate voice input |
| Windows / Linux | `Ctrl + Shift + Space` | Activate voice input |

The voice activation shortcut is fully customizable in **Settings > Wake Mode > Keyboard Shortcut**.

### Standard Shortcuts

| Platform | Shortcut | Description |
|---|---|---|
| macOS | `Cmd + Q` | Quit Neotask |
| macOS | `Cmd + W` | Close current window |
| macOS | `Cmd + ,` | Open Settings |
| Windows / Linux | `Alt + F4` | Quit Neotask |
| Windows / Linux | `Ctrl + W` | Close current window |

---

## Auto-Updates

Neotask checks for updates automatically to ensure you always have the latest features and security patches.

### How Updates Work

1. **Check on launch**: Each time you open Neotask, it checks for available updates in the background.
2. **Background download**: If an update is available, it downloads automatically without interrupting your work.
3. **Install prompt**: Once the download completes, you are prompted to install the update and restart the application.
4. **Remind me later**: If you are not ready to update, select "Remind me later." You will be prompted again periodically, up to a maximum of **7 days**, after which the update becomes required.

Updates are incremental when possible to minimize download size and restart time.

---

### [Free Plan](https://neotask.ai/docs/free-plan)

## Track Free Usage

Open **Usage** to see remaining credits, model activity, and recent sessions. Open a session when you need to understand the usage behind a record.

![Usage overview with credits and recent model activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

The Free plan is a limited trial lane for using Neotask chat, connected apps, boards, and coding workflows before upgrading to a paid subscription.

Free users still cover their own hosted model/token usage. They can use starter credits, buy additional credits through Neotask, or use their own provider keys where supported. Free users do not receive overage. Local models on your computer are free.

---

## Message Quota

Free usage has two phases:

| Phase | Limit | Reset |
| --- | --- | --- |
| Intro allowance | First `10` messages/events | Lifetime, one time |
| Daily allowance | `5` messages/events per day after the intro allowance | User-local calendar day |

Each allowed hosted-model message still requires available usage coverage. If credits are depleted, or the user's provider key has no balance, the message is blocked even if daily message quota remains.

Messages to a local model on your computer are free and do not count against these limits. See [Models & Providers](/docs/neotask-desktop/models-providers).

---

## Included On Free

Free users can use:

- Main chat mode.
- Boards and live artifacts.
- The coding subpage for interactive coding tasks.
- App and integration connection flows.
- Connected apps inside allowed chat, boards, and coding usage.
- The system-managed daily memory learning job named `Learn from Daily Activities`.

The daily memory learning job is a platform-maintained system task. It is not a user-created automation and is the only automation exception on Free.

---

## Not Included On Free

Free users cannot:

- Run company tasks or company automations.
- Use an autonomous company in normal operating mode.
- Create, edit, run, or delete their own automations, cron jobs, reminders, or scheduled jobs.
- Add their own custom agents.
- Use overage after credits are depleted.

Company onboarding may be available so a user can preview or prepare a company, but the company dashboard remains locked until the account has at least an Individual subscription.

---

## Apps, Credits, And API Keys

Free users can connect apps and provider accounts, but connecting an app does not unlock paid automation.

For model usage:

- Hosted credits use the user's Neotask credit balance.
- BYOK routes base provider token costs to the user's provider account.
- Free daily message limits still apply to hosted models in either mode.
- Free users must have enough credits or provider balance for each allowed hosted-model message.
- Local models on your computer need no credits or provider balance and do not count against the Free message limits.
- Free users get no overage, so exhausted credits or insufficient provider balance hard-stop usage.

---

## When To Upgrade

Upgrade to at least Individual when you need company access, custom agents, or user-created automations. Business and Enterprise plans add company capacity, employee seats, setup-task allowances, and broader autonomous operations.

---

### [Getting Started](https://neotask.ai/docs/getting-started)

## Open Your First Workspace

1. Download and install Neotask.
2. Sign in and finish the setup prompts.
3. Open **Agents**.
4. Select the main agent and start in **Chat**.
5. Open **Overview** to see scheduled work, sessions, and recent activity.

![Main agent chat ready for a new request](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/chat-main-1280.webp)

This guide walks you through installing Neotask, activating your license, and deploying your first agent.

---

## What Is Neotask?

Neotask is a unified AI agent platform. It gives you a desktop application, web dashboard, and orchestration engine to deploy, manage, and scale AI agents across messaging channels, voice, automation workflows, and more.

You are not buying a fixed number of agent runs. You are getting full access to the platform. Your subscription covers the software, and your agents consume AI tokens as they work. See [Billing & Plans](billing.md) for a complete breakdown of how pricing works.

If you start on the Free plan, your access is narrower: you can use chat, boards, coding workflows, and connected apps within the Free message quota, but company operation, user-created automations, scheduled jobs, and custom agents require at least an Individual subscription. Free usage includes the first `10` messages/events, then `5` messages/events per user-local calendar day, and every allowed hosted-model message still needs credits or BYOK provider balance. Local models on your computer are free and do not count against these limits. See [Free Plan](free-plan.md).

---

## System Requirements

| Platform | Minimum Version | Format |
| --- | --- | --- |
| macOS | 11 (Big Sur) or later | DMG |
| Windows | 10 or later | EXE installer |
| Linux | 64-bit distribution | AppImage |

All platforms require an active internet connection for license activation and agent communication.

---

## Download & Install

Download the latest version from your Neotask account dashboard or the link provided in your purchase confirmation email.

### macOS

1. Open the downloaded `.dmg` file.
2. Drag **Neotask** into your **Applications** folder.
3. Launch Neotask from Applications or Spotlight.
4. If prompted by Gatekeeper, go to **System Settings > Privacy & Security** and click **Open Anyway**.

### Windows

1. Run the downloaded `.exe` installer.
2. Follow the installation wizard prompts.
3. Choose an install location or accept the default.
4. Launch Neotask from the Start Menu or desktop shortcut.

### Linux

1. Download the `.AppImage` file.
2. Make it executable:
   ```bash
   chmod +x Neotask-*.AppImage
   ```
3. Run the AppImage:
   ```bash
   ./Neotask-*.AppImage
   ```
4. Optionally, move it to `/usr/local/bin` or integrate it with your desktop environment.

---

## License Activation

On first launch, you will be prompted to enter your license key.

**License key format:**

```
NT-XXXXXXXX-XXXXXXXX-XXXXXXXX-XXXXXXXX
```

You can find your license key in your purchase confirmation email or your account dashboard at neotask.ai.

Enter the key, click **Activate**, and the app will verify your license online. Once activated, the app is tied to your machine. You can manage active devices from the Web Dashboard.

---

## Onboarding Wizard

After activation, the onboarding wizard guides you through initial setup in **11 steps**:

| Step | What You'll Do |
| --- | --- |
| 1. Select Agent Template | Choose a pre-built agent template to start from |
| 2. Configure Agent | Set your agent's name, personality, and behavior |
| 3. Set Budget | Define daily spending limits for token consumption |
| 4. Connect Channels | Link messaging platforms (Slack, Discord, Telegram, etc.) |
| 5. Connect Google Workspace | Authorize Gmail, Calendar, Drive, and Docs access |
| 6. Developer Features | Enable or disable terminal access, code execution, and Git integration |
| 7. Built-in Skills | Select from pre-packaged skills like web search, file management, and summarization |
| 8. Automations | Set up scheduled tasks, triggers, and automated workflows |
| 9. Voice Control | Configure wake word, language, and microphone input |
| 10. Notifications | Choose notification preferences for desktop, email, and mobile alerts |
| 11. Download & Launch | Finalize configuration and launch your first agent |

You can skip any step and return to it later from **Settings**.

Free users may complete company onboarding, but the company cannot run tasks or automations in Free mode. The company dashboard remains gated until the account upgrades to at least Individual.

---

## Agent Templates

Templates give you a pre-configured starting point. You can customize any template after creation.

| Template | Use Case |
| --- | --- |
| **Assistant** | General-purpose personal assistant for everyday tasks |
| **Developer** | Code generation, debugging, Git workflows, and terminal commands |
| **Researcher** | Web research, source synthesis, and report generation |
| **Social Media** | Content creation, scheduling, and engagement across platforms |
| **Data Analyst** | Data processing, visualization, and insight extraction |
| **Home Automation** | Smart home device control and routine management |
| **Voice Assistant** | Voice-first interaction optimized for hands-free use |
| **Tutor** | Interactive learning, explanations, and quiz generation |
| **Campaign Manager** | Marketing campaign planning, execution, and performance tracking |

---

## Quick Start

Once onboarding is complete, try these to get familiar with Neotask:

### Send Your First Message

1. Open the main chat panel.
2. Type a message in the input field, for example: *"Summarize my unread emails."*
3. Press **Enter** or click **Send**.
4. Your agent will process the request and respond in the chat.

### Try Voice Control

1. Click the **microphone icon** in the toolbar, or say your configured wake word.
2. Speak a command, for example: *"What's on my calendar today?"*
3. The agent will transcribe your speech, process the request, and respond with audio or text.

### Explore the Dashboard

1. Click **Dashboard** in the left sidebar.
2. Review agent activity, token usage, and connected integrations.
3. Use the dashboard to monitor performance and adjust agent settings.

![Neotask agent overview with scheduled work and recent activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

---

## Next Steps

- [Desktop App](desktop-app.md): Learn the interface in depth.
- [Agents](agents.md): Customize agent behavior and configuration.
- [Skills](skills.md): Add capabilities to your agents.
- [Voice Control](voice-control.md): Advanced voice commands and language settings.
- [Integrations](integrations.md): Connect additional messaging platforms and tools.
- [Billing & Plans](billing.md): Understand the platform pricing model and usage fees.

---

### [Integrations](https://neotask.ai/docs/integrations)

## Connect An App

1. Open **Apps**.
2. Search for the service you need.
3. Open it and choose the available authentication method.
4. Finish authorization and return to Neotask.
5. Confirm that the service appears under connected apps.

![Connected and available app catalog](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

## Overview

Neotask connects to **1200+ services**, from Google Workspace (via the **Integrations** tab) to 1200+ external apps via MCP (via the **Apps** tab). This page covers all available integrations and how to set them up.

There are two types of integrations:

- **Built-in integrations** = native API wrappers (Google Workspace, AI providers, messaging platforms). Configured in **Agents → Integrations** tab.
- **App integrations** = 1200+ external service connectors (Slack, Notion, GitHub, Stripe, and more). Configured in **Agents → Apps** tab.

## Google Workspace (25 Services)

Neotask integrates with 25 Google services across four categories. Connect via OAuth from the Integrations tab or during onboarding.

### Core Services

| Service | Capabilities |
|---------|-------------|
| **Gmail** | Read, compose, send, manage labels, search emails |
| **Google Calendar** | Create, update, delete events; check availability |
| **Google Drive** | Upload, download, organize files and folders |
| **Google Docs** | Create, read, edit documents |
| **Google Sheets** | Read, write, format spreadsheet data |
| **Google Slides** | Create and edit presentations |
| **Google Forms** | Create forms, read responses |
| **Google Keep** | Create and manage notes and lists |
| **Google Tasks** | Create, complete, organize task lists |

### Communication Services

| Service | Capabilities |
|---------|-------------|
| **Google People/Contacts** | Read, create, update contacts |
| **Google Chat** | Send messages in Spaces and DMs |
| **Google Meet** | Schedule and manage meetings |
| **YouTube** | Search, manage playlists, upload content |
| **Google Photos** | Access and organize photo library |

### Location Services

| Service | Capabilities |
|---------|-------------|
| **Places API** | Search for businesses and points of interest |
| **Routes/Directions** | Calculate routes and travel times |
| **Business Profile** | Manage Google Business listings |

### Specialized Services

| Service | Capabilities |
|---------|-------------|
| **Google Classroom** | Manage courses, assignments, students |
| **Google Play Developer** | App management and analytics |
| **Google AdSense** | Ad revenue reporting |
| **Google Ads** | Campaign management and reporting |

### Connecting Google Services

1. Go to **Agents → Integrations** tab
2. Click on a Google service card
3. Authorize via Google OAuth popup
4. Grant requested permissions
5. Service becomes active immediately

Confirm the account and each enabled Google service on the **Integrations** page.

![Google services connected to Neotask](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/google-services-1440.webp)

### Multi-Account Support

- Connect multiple Google accounts
- Switch between accounts per service
- Per-account token management
- Automatic token refresh

## AI Model Providers

Neotask supports multiple AI providers. Use our pooled keys (System mode) or bring your own (BYOK mode).

### Supported Providers

| Provider | Models | Key Type |
|----------|--------|----------|
| **Anthropic** | Claude 4 Sonnet, Claude 4 Opus, Claude 3.5 Haiku | API Key |
| **OpenAI** | GPT-5.1, GPT-4.1, GPT-4.1 Nano | API Key / OAuth (Codex) |
| **Google AI** | Gemini 3 Pro, Gemini 3 Flash, Gemini 2.5 Flash Lite | API Key |
| **DeepSeek** | DeepSeek V3.2 | API Key |
| **OpenRouter** | Multi-model routing (100+ models) | API Key |

### Setting Up a Provider

1. Go to **Settings** or **Onboarding → Provider Setup**
2. Select your provider
3. Enter your API key (encrypted with AES-256-GCM)
4. Key is verified and stored securely
5. Models from that provider become available

### Provider Fallback

If one provider is unavailable, the system automatically falls back:

- Primary: Claude Sonnet 4
- Fallback 1: GPT-5.1
- Fallback 2: Gemini 3 Pro

## Messaging Platforms

See the [Channels documentation](channels.md) for complete details on connecting messaging platforms. Supported platforms include:

- Slack, Discord, WhatsApp, Telegram, Signal, Microsoft Teams, Google Chat, SMS, Email/Gmail, Matrix, Line, Zalo, IRC, iMessage

Plus 6 additional via extensions: Twitch, Mattermost, Nextcloud Talk, Nostr, Feishu/Lark, BlueBubbles

## Smart Home

| Integration | Capabilities |
|-------------|-------------|
| **Philips Hue** | Control lights, scenes, rooms, schedules |
| **Sonos** | Playback control, volume, grouping |
| **Eight Sleep** | Temperature control, sleep tracking |
| **Camera/Security** | Snapshots, monitoring |

## Productivity Apps

| Integration | Capabilities |
|-------------|-------------|
| **Notion** | Database operations, page management |
| **Apple Notes** | Create and manage notes |
| **Apple Reminders** | Task management |
| **Bear Notes** | Note management |
| **Obsidian** | Vault access and editing |
| **Things 3** | Task management (macOS) |
| **Trello** | Board and card management |
| **1Password** | Credential retrieval (CLI) |

## Developer Tools

| Integration | Capabilities |
|-------------|-------------|
| **GitHub** | PRs, issues, workflows, code review |
| **Spotify** | Playback control, playlist management |
| **Weather** | Location-based weather data |
| **PDF Processing** | Extract and analyze PDF content |
| **Video Processing** | Frame extraction, analysis |
| **Image Generation** | DALL-E powered image creation |

## Webhooks

- Receive events via incoming webhooks
- Send results to external URLs
- Configure per-automation or per-channel
- JSON payload format
- Retry on failure

---

## App Integrations: 1200+ Services

In addition to the built-in integrations above, Neotask connects to **1200+ external services**
through the **Apps** tab, letting your agents interact with tools, APIs, and platforms through
natural language.

For the full directory of all 1200+ apps with links to each detail page, see **[Supported Apps](mcp-integrations.md)**.

### Browsing Integrations on the Website

On the [homepage](/) at neotask.ai, scroll down to the **"1200+ Integrations. One Platform."** section. You'll see:

- A dense grid of **service logos**, each clickable, linking to that integration's detail page at `neotask.ai/integrations/{id}`
- **16 category filter chips**: hover or click any chip to highlight matching integrations
- Each detail page includes: service logo, category badge, key capability bullets, "What You Can Do" section, "Try Asking" example prompts, "Pro Tips", pricing, and related integrations

### Connecting an App Integration

1. Open the Neotask desktop app
2. Click your agent name in the left sidebar
3. Click the **Apps** tab
4. Find the integration by browsing categories or using the search bar
5. Click the app card, then click **Connect**
6. Complete the auth flow:

| Auth Type | What Happens |
|-----------|-------------|
| **OAuth** | A browser window opens. Sign in to the service, authorize access. Tokens are stored and refreshed automatically. |
| **API Key** | An input field appears. Paste your API key. It's encrypted with AES-256-GCM and stored locally. |
| **None** | No auth needed; the integration works immediately after enabling. |

### Connection Status Reference

| Status | What It Means | What to Do |
|--------|--------------|------------|
| **Connected** | Auth is valid and the integration is ready to use | Nothing needed; you're good |
| **Pending** | Auth flow started but not yet completed | Complete the OAuth popup or enter your API key |
| **Expired** | Token has expired | Click **Reconnect** to re-authorize |
| **Error** | Auth failed | Check the error message; may need to re-enter credentials |
| **None** | Not yet connected | Click **Connect** to start |

---

### Complete App Integration Directory

Every integration listed below has a dedicated detail page at
`neotask.ai/integrations/{id}` with full capabilities, example prompts, and pricing.

#### Productivity (26)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Airtable](/integrations/airtable) | Spreadsheet-database hybrid: manage bases, tables, records, views | [→](/integrations/airtable) |
| [Anytype](/integrations/anytype) | Local-first knowledge management: pages, objects, relations | [→](/integrations/anytype) |
| [Archbee](/integrations/archbee) | Documentation platform: create and manage docs, spaces | [→](/integrations/archbee) |
| [Asana](/integrations/asana) | Project management: tasks, projects, teams, timelines | [→](/integrations/asana) |
| [Atlassian](/integrations/atlassian) | Jira + Confluence suite: issues, projects, pages, spaces | [→](/integrations/atlassian) |
| [Atono](/integrations/atono) | Workflow automation platform | [→](/integrations/atono) |
| [Backdocket](/integrations/backdocket) | Legal practice management: cases, documents, billing | [→](/integrations/backdocket) |
| [Basecamp](/integrations/basecamp) | Team collaboration: projects, todos, messages, schedules | [→](/integrations/basecamp) |
| [BoldSign](/integrations/boldsign) | E-signatures: send, track, manage signature requests | [→](/integrations/boldsign) |
| [Buildable](/integrations/buildable) | Developer workflow automation | [→](/integrations/buildable) |
| [Cal.com](/integrations/cal-com) | Open-source scheduling: manage calendars, bookings, availability | [→](/integrations/cal-com) |
| [ClickUp](/integrations/clickup) | Project management: tasks, docs, goals, dashboards | [→](/integrations/clickup) |
| [Coda](/integrations/coda) | All-in-one doc: docs, tables, automations | [→](/integrations/coda) |
| [Confluence](/integrations/confluence) | Team wiki: spaces, pages, comments, search | [→](/integrations/confluence) |
| [Evernote](/integrations/evernote) | Note-taking: notebooks, notes, tags, search | [→](/integrations/evernote) |
| [Jira](/integrations/jira) | Issue tracking: projects, boards, sprints, epics | [→](/integrations/jira) |
| [Linear](/integrations/linear) | Issue tracking: projects, cycles, teams, labels | [→](/integrations/linear) |
| [Make](/integrations/make) | Automation platform: scenarios, modules, connections | [→](/integrations/make) |
| [Microsoft To Do](/integrations/microsoft-todo) | Task management: lists, tasks, reminders | [→](/integrations/microsoft-todo) |
| [Monday.com](/integrations/monday) | Work management: boards, items, automations | [→](/integrations/monday) |
| [Notion](/integrations/notion) | All-in-one workspace: pages, databases, relations | [→](/integrations/notion) |
| [Obsidian](/integrations/obsidian) | Markdown knowledge base: vaults, notes, links | [→](/integrations/obsidian) |
| [Roam Research](/integrations/roam-research) | Networked note-taking: pages, blocks, references | [→](/integrations/roam-research) |
| [Smartsheet](/integrations/smartsheet) | Enterprise work management: sheets, reports, dashboards | [→](/integrations/smartsheet) |
| [Todoist](/integrations/todoist) | Task management: projects, tasks, labels, filters | [→](/integrations/todoist) |
| [Trello](/integrations/trello) | Kanban boards: boards, lists, cards, checklists | [→](/integrations/trello) |
| [Wrike](/integrations/wrike) | Enterprise project management: projects, tasks, timelines | [→](/integrations/wrike) |

#### Design (11)

| Integration | What It Does | Page |
|------------|-------------|------|
| [21st.dev](/integrations/21st-dev) | AI component marketplace | [→](/integrations/21st-dev) |
| [Adobe XD](/integrations/adobe-xd) | UI/UX design: artboards, components, prototypes | [→](/integrations/adobe-xd) |
| [Canva](/integrations/canva) | Graphic design: templates, images, branding | [→](/integrations/canva) |
| [FigJam](/integrations/figjam) | Collaborative whiteboard: brainstorming, diagramming | [→](/integrations/figjam) |
| [Figma](/integrations/figma) | Interface design: files, components, styles, variables | [→](/integrations/figma) |
| [Framer](/integrations/framer) | Web design and publishing: sites, pages, CMS | [→](/integrations/framer) |
| [InVision](/integrations/invision) | Design collaboration: prototypes, boards, specs | [→](/integrations/invision) |
| [Miro](/integrations/miro) | Visual collaboration: boards, widgets, frames | [→](/integrations/miro) |
| [Penpot](/integrations/penpot) | Open-source design: files, components, teams | [→](/integrations/penpot) |
| [Sketch](/integrations/sketch) | macOS design tool: artboards, symbols, libraries | [→](/integrations/sketch) |
| [Zeplin](/integrations/zeplin) | Design handoff: specs, style guides, assets | [→](/integrations/zeplin) |

#### Communication (17)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Discord](/integrations/discord) | Community messaging: servers, channels, messages, roles | [→](/integrations/discord) |
| [Front](/integrations/front) | Shared inbox: conversations, comments, assignments | [→](/integrations/front) |
| [Google Meet](/integrations/google-meet) | Video meetings: schedule, join, manage meetings | [→](/integrations/google-meet) |
| [Infobip](/integrations/infobip) | Omnichannel messaging: SMS, WhatsApp, email, voice | [→](/integrations/infobip) |
| [Intercom](/integrations/intercom) | Customer messaging: conversations, users, articles | [→](/integrations/intercom) |
| [Knock](/integrations/knock) | Notification infrastructure: workflows, channels, feeds | [→](/integrations/knock) |
| [Line](/integrations/line) | Messaging platform: messages, groups, media | [→](/integrations/line) |
| [Loom](/integrations/loom) | Video messaging: recordings, transcripts, sharing | [→](/integrations/loom) |
| [Mailgun](/integrations/mailgun) | Email API: send, receive, track, validate | [→](/integrations/mailgun) |
| [Mailjet](/integrations/mailjet) | Email delivery: campaigns, transactional, templates | [→](/integrations/mailjet) |
| [Microsoft Teams](/integrations/microsoft-teams) | Team collaboration: chats, channels, meetings | [→](/integrations/microsoft-teams) |
| [SendGrid](/integrations/sendgrid) | Email delivery: transactional, marketing, analytics | [→](/integrations/sendgrid) |
| [Slack](/integrations/slack) | Team messaging: channels, messages, threads, files | [→](/integrations/slack) |
| [Telegram](/integrations/telegram) | Messaging: chats, groups, bots, channels | [→](/integrations/telegram) |
| [Twilio](/integrations/twilio) | Communications API: SMS, voice, video, WhatsApp | [→](/integrations/twilio) |
| [WhatsApp](/integrations/whatsapp) | Messaging: conversations, media, templates | [→](/integrations/whatsapp) |
| [Zoom](/integrations/zoom) | Video conferencing: meetings, webinars, recordings | [→](/integrations/zoom) |

#### Analytics (12)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Amplitude](/integrations/amplitude) | Product analytics: events, funnels, cohorts, retention | [→](/integrations/amplitude) |
| [Audiense](/integrations/audiense) | Audience intelligence: segments, insights, reports | [→](/integrations/audiense) |
| [Datadog](/integrations/datadog) | Monitoring: metrics, traces, logs, dashboards | [→](/integrations/datadog) |
| [Google Analytics](/integrations/google-analytics) | Web analytics: traffic, conversions, audiences | [→](/integrations/google-analytics) |
| [Grafana](/integrations/grafana) | Observability: dashboards, alerts, data sources | [→](/integrations/grafana) |
| [Heap](/integrations/heap) | Digital analytics: auto-capture events, funnels | [→](/integrations/heap) |
| [Hotjar](/integrations/hotjar) | Behavior analytics: heatmaps, recordings, surveys | [→](/integrations/hotjar) |
| [Mixpanel](/integrations/mixpanel) | Product analytics: events, funnels, A/B tests | [→](/integrations/mixpanel) |
| [New Relic](/integrations/new-relic) | Observability: APM, infrastructure, logs | [→](/integrations/new-relic) |
| [Plausible](/integrations/plausible) | Privacy-friendly web analytics | [→](/integrations/plausible) |
| [PostHog](/integrations/posthog) | Product analytics: events, feature flags, session recordings | [→](/integrations/posthog) |
| [Segment](/integrations/segment) | Customer data platform: sources, destinations, audiences | [→](/integrations/segment) |

#### Sales & CRM (12)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Apollo](/integrations/apollo) | Sales intelligence: leads, contacts, sequences | [→](/integrations/apollo) |
| [Apollo.io](/integrations/apollo-io) | B2B data and engagement: prospecting, outreach | [→](/integrations/apollo-io) |
| [Attio](/integrations/attio) | Next-gen CRM: contacts, companies, pipelines | [→](/integrations/attio) |
| [Clay](/integrations/clay) | Data enrichment: lead scoring, waterfall enrichment | [→](/integrations/clay) |
| [Close](/integrations/close) | Sales CRM: leads, activities, pipelines, calling | [→](/integrations/close) |
| [Freshsales](/integrations/freshsales) | CRM: contacts, deals, accounts, territories | [→](/integrations/freshsales) |
| [Gong](/integrations/gong) | Revenue intelligence: call recordings, analytics | [→](/integrations/gong) |
| [HubSpot](/integrations/hubspot) | CRM platform: contacts, deals, tickets, marketing | [→](/integrations/hubspot) |
| [Outreach](/integrations/outreach) | Sales engagement: sequences, tasks, analytics | [→](/integrations/outreach) |
| [Pipedrive](/integrations/pipedrive) | Sales CRM: deals, contacts, activities, pipelines | [→](/integrations/pipedrive) |
| [Salesforce](/integrations/salesforce) | Enterprise CRM: accounts, opportunities, reports | [→](/integrations/salesforce) |
| [Zoho CRM](/integrations/zoho-crm) | CRM: leads, contacts, deals, analytics | [→](/integrations/zoho-crm) |

#### Development (24)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Apify](/integrations/apify) | Web scraping and automation: actors, runs, datasets | [→](/integrations/apify) |
| [AWS](/integrations/aws) | Cloud infrastructure: EC2, S3, Lambda, IAM | [→](/integrations/aws) |
| [Azure DevOps](/integrations/azure-devops) | DevOps: repos, pipelines, boards, artifacts | [→](/integrations/azure-devops) |
| [Bitbucket](/integrations/bitbucket) | Git hosting: repos, PRs, pipelines, issues | [→](/integrations/bitbucket) |
| [Bitrise](/integrations/bitrise) | Mobile CI/CD: builds, workflows, apps | [→](/integrations/bitrise) |
| [Buildkite](/integrations/buildkite) | CI/CD: pipelines, builds, agents | [→](/integrations/buildkite) |
| [Cloudflare](/integrations/cloudflare) | Edge platform: DNS, CDN, workers, security | [→](/integrations/cloudflare) |
| [Docker](/integrations/docker) | Containerization: images, containers, registries | [→](/integrations/docker) |
| [Firebase](/integrations/firebase) | App platform: auth, database, hosting, functions | [→](/integrations/firebase) |
| [GitHub](/integrations/github) | Code hosting: repos, PRs, issues, actions, releases | [→](/integrations/github) |
| [GitLab](/integrations/gitlab) | DevOps platform: repos, CI/CD, issues, registries | [→](/integrations/gitlab) |
| [Jenkins](/integrations/jenkins) | CI/CD server: jobs, builds, pipelines | [→](/integrations/jenkins) |
| [JetBrains](/integrations/jetbrains) | IDE platform: projects, plugins, configurations | [→](/integrations/jetbrains) |
| [JFrog](/integrations/jfrog) | Artifact management: repos, packages, security | [→](/integrations/jfrog) |
| [LaunchDarkly](/integrations/launchdarkly) | Feature flags: flags, environments, segments | [→](/integrations/launchdarkly) |
| [Liveblocks](/integrations/liveblocks) | Real-time collaboration: rooms, presence, storage | [→](/integrations/liveblocks) |
| [MinIO](/integrations/minio) | Object storage: buckets, objects, policies | [→](/integrations/minio) |
| [Neon](/integrations/neon) | Serverless Postgres: databases, branches, endpoints | [→](/integrations/neon) |
| [Netlify](/integrations/netlify) | Web hosting: sites, deploys, functions, forms | [→](/integrations/netlify) |
| [PlanetScale](/integrations/planetscale) | Serverless MySQL: databases, branches, deploy requests | [→](/integrations/planetscale) |
| [Railway](/integrations/railway) | Cloud deployment: services, environments, variables | [→](/integrations/railway) |
| [Sentry](/integrations/sentry) | Error tracking: issues, events, releases, alerts | [→](/integrations/sentry) |
| [Supabase](/integrations/supabase) | Backend platform: database, auth, storage, functions | [→](/integrations/supabase) |
| [Vercel](/integrations/vercel) | Frontend deployment: projects, deployments, domains | [→](/integrations/vercel) |

#### Databases (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [BigQuery](/integrations/bigquery) | Data warehouse: datasets, queries, tables | [→](/integrations/bigquery) |
| [ClickHouse](/integrations/clickhouse) | Column-oriented database: queries, tables, inserts | [→](/integrations/clickhouse) |
| [DynamoDB](/integrations/dynamodb) | NoSQL database: tables, items, queries | [→](/integrations/dynamodb) |
| [Elasticsearch](/integrations/elasticsearch) | Search engine: indices, documents, queries | [→](/integrations/elasticsearch) |
| [MongoDB](/integrations/mongodb) | Document database: collections, documents, aggregations | [→](/integrations/mongodb) |
| [PostgreSQL](/integrations/postgresql) | Relational database: tables, queries, schemas | [→](/integrations/postgresql) |
| [Redis](/integrations/redis) | In-memory database: keys, streams, pub/sub | [→](/integrations/redis) |
| [Snowflake](/integrations/snowflake) | Cloud data warehouse: warehouses, databases, queries | [→](/integrations/snowflake) |

#### Data (6)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Algolia](/integrations/algolia) | Search API: indices, records, analytics | [→](/integrations/algolia) |
| [Astra DB](/integrations/astra-db) | Serverless Cassandra: databases, collections, documents | [→](/integrations/astra-db) |
| [Atlan](/integrations/atlan) | Data catalog: assets, glossary, lineage | [→](/integrations/atlan) |
| [Baserow](/integrations/baserow) | Open-source database: tables, fields, rows | [→](/integrations/baserow) |
| [Bright Data](/integrations/brightdata) | Web data platform: collectors, datasets, proxies | [→](/integrations/brightdata) |
| [Meilisearch](/integrations/meilisearch) | Search engine: indexes, documents, facets | [→](/integrations/meilisearch) |

#### Finance (13)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Adfin](/integrations/adfin) | Financial operations: invoices, payments, reconciliation | [→](/integrations/adfin) |
| [Aiera](/integrations/aiera) | Financial intelligence: earnings, transcripts, analysis | [→](/integrations/aiera) |
| [Airwallex](/integrations/airwallex) | Global payments: accounts, transfers, FX | [→](/integrations/airwallex) |
| [Alpaca](/integrations/alpaca) | Trading API: orders, positions, market data | [→](/integrations/alpaca) |
| [Alpha Vantage](/integrations/alphavantage) | Market data: stocks, forex, crypto, indicators | [→](/integrations/alphavantage) |
| [Brex](/integrations/brex) | Corporate finance: cards, expenses, budgets | [→](/integrations/brex) |
| [Mercury](/integrations/mercury) | Banking: accounts, transactions, payments | [→](/integrations/mercury) |
| [Plaid](/integrations/plaid) | Banking API: accounts, transactions, identity | [→](/integrations/plaid) |
| [QuickBooks](/integrations/quickbooks) | Accounting: invoices, expenses, reports | [→](/integrations/quickbooks) |
| [Ramp](/integrations/ramp) | Corporate cards: transactions, receipts, budgets | [→](/integrations/ramp) |
| [Square](/integrations/square) | Commerce: payments, catalog, customers, locations | [→](/integrations/square) |
| [Stripe](/integrations/stripe) | Payments: charges, subscriptions, invoices, customers | [→](/integrations/stripe) |
| [Xero](/integrations/xero) | Accounting: invoices, contacts, bank reconciliation | [→](/integrations/xero) |

#### Marketing (11)

| Integration | What It Does | Page |
|------------|-------------|------|
| [ActiveCampaign](/integrations/activecampaign) | Email marketing and CRM: campaigns, automations, contacts | [→](/integrations/activecampaign) |
| [Ahrefs](/integrations/ahrefs) | SEO: site audit, keywords, backlinks, rank tracking | [→](/integrations/ahrefs) |
| [AirOps](/integrations/airops) | AI workflow automation for marketing | [→](/integrations/airops) |
| [Beehiiv](/integrations/beehiiv) | Newsletter platform: subscribers, campaigns, analytics | [→](/integrations/beehiiv) |
| [Brevo](/integrations/brevo) | Email marketing: campaigns, contacts, automations | [→](/integrations/brevo) |
| [BuiltWith](/integrations/builtwith) | Technology profiling: tech stacks, leads, trends | [→](/integrations/builtwith) |
| [ConvertKit](/integrations/convertkit) | Creator email marketing: subscribers, sequences, forms | [→](/integrations/convertkit) |
| [Customer.io](/integrations/customer-io) | Customer messaging: campaigns, segments, workflows | [→](/integrations/customer-io) |
| [Drip](/integrations/drip) | E-commerce email: automations, subscribers, campaigns | [→](/integrations/drip) |
| [Klaviyo](/integrations/klaviyo) | E-commerce marketing: flows, campaigns, segments | [→](/integrations/klaviyo) |
| [Mailchimp](/integrations/mailchimp) | Email marketing: audiences, campaigns, automations | [→](/integrations/mailchimp) |

#### AI & ML (13)

| Integration | What It Does | Page |
|------------|-------------|------|
| [AgentOps](/integrations/agentops) | AI agent observability: sessions, events, analytics | [→](/integrations/agentops) |
| [Anthropic](/integrations/anthropic) | Claude AI: completions, messages, embeddings | [→](/integrations/anthropic) |
| [Arize Phoenix](/integrations/arize-phoenix) | ML observability: traces, evaluations, datasets | [→](/integrations/arize-phoenix) |
| [Cohere](/integrations/cohere) | NLP API: embeddings, reranking, generation | [→](/integrations/cohere) |
| [ElevenLabs](/integrations/elevenlabs) | Voice AI: text-to-speech, voice cloning, dubbing | [→](/integrations/elevenlabs) |
| [Hugging Face](/integrations/hugging-face) | ML platform: models, datasets, spaces | [→](/integrations/hugging-face) |
| [Kagi Search](/integrations/kagi-search) | Premium search: web search, summarization | [→](/integrations/kagi-search) |
| [Label Studio](/integrations/label-studio) | Data labeling: projects, tasks, annotations | [→](/integrations/label-studio) |
| [OpenAI](/integrations/openai) | GPT AI: completions, embeddings, images, audio | [→](/integrations/openai) |
| [Pinecone](/integrations/pinecone) | Vector database: indexes, vectors, queries | [→](/integrations/pinecone) |
| [Replicate](/integrations/replicate) | ML inference: models, predictions, deployments | [→](/integrations/replicate) |
| [Stability AI](/integrations/stability-ai) | Image generation: text-to-image, upscaling, editing | [→](/integrations/stability-ai) |
| [Weaviate](/integrations/weaviate) | Vector database: schemas, objects, searches | [→](/integrations/weaviate) |

#### Security (9)

| Integration | What It Does | Page |
|------------|-------------|------|
| [1Password](/integrations/1password) | Password management: vaults, items, secrets | [→](/integrations/1password) |
| [Asgardeo](/integrations/asgardeo) | Identity management: users, apps, connections | [→](/integrations/asgardeo) |
| [Auth0](/integrations/auth0) | Identity platform: users, connections, rules | [→](/integrations/auth0) |
| [BoostSecurity](/integrations/boostsecurity) | DevSecOps: policies, findings, scanners | [→](/integrations/boostsecurity) |
| [CrowdStrike](/integrations/crowdstrike) | Endpoint security: detections, hosts, policies | [→](/integrations/crowdstrike) |
| [Dependabot](/integrations/dependabot) | Dependency security: alerts, updates, PRs | [→](/integrations/dependabot) |
| [Snyk](/integrations/snyk) | Security scanning: vulnerabilities, fixes, dependencies | [→](/integrations/snyk) |
| [SonarQube](/integrations/sonarqube) | Code quality: issues, rules, quality gates | [→](/integrations/sonarqube) |
| [Vault](/integrations/vault) | Secrets management: secrets, policies, auth | [→](/integrations/vault) |

#### Infrastructure (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Aiven](/integrations/aiven) | Managed data services: databases, queues, caches | [→](/integrations/aiven) |
| [Ansible](/integrations/ansible) | IT automation: playbooks, inventories, roles | [→](/integrations/ansible) |
| [Axiom](/integrations/axiom) | Log management: datasets, queries, dashboards | [→](/integrations/axiom) |
| [Azure](/integrations/azure) | Microsoft cloud: VMs, storage, functions, networking | [→](/integrations/azure) |
| [Chef](/integrations/chef) | Infrastructure automation: cookbooks, recipes, nodes | [→](/integrations/chef) |
| [Kubernetes](/integrations/kubernetes) | Container orchestration: pods, services, deployments | [→](/integrations/kubernetes) |
| [Prometheus](/integrations/prometheus) | Monitoring: metrics, alerts, queries | [→](/integrations/prometheus) |
| [Pulumi](/integrations/pulumi) | Infrastructure as code: stacks, resources, outputs | [→](/integrations/pulumi) |

#### Storage (7)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Backblaze B2](/integrations/backblaze-b2) | Cloud storage: buckets, files, uploads | [→](/integrations/backblaze-b2) |
| [Box](/integrations/box) | Cloud content: files, folders, collaborations | [→](/integrations/box) |
| [Dropbox](/integrations/dropbox) | File storage: files, folders, sharing, sync | [→](/integrations/dropbox) |
| [Google Drive](/integrations/google-drive) | Cloud storage: files, folders, permissions, search | [→](/integrations/google-drive) |
| [OneDrive](/integrations/onedrive) | Microsoft cloud storage: files, folders, sharing | [→](/integrations/onedrive) |
| [S3](/integrations/s3) | AWS object storage: buckets, objects, policies | [→](/integrations/s3) |
| [Wasabi](/integrations/wasabi) | Cloud storage: buckets, objects, hot storage | [→](/integrations/wasabi) |

#### Browser (4)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Browser Automation](/integrations/browser-automation) | Automated browsing: navigation, clicks, scraping | [→](/integrations/browser-automation) |
| [Browserbase](/integrations/browserbase) | Cloud browsers: sessions, screenshots, recording | [→](/integrations/browserbase) |
| [Playwright](/integrations/playwright) | Browser testing: pages, selectors, assertions | [→](/integrations/playwright) |
| [Puppeteer](/integrations/puppeteer) | Chrome automation: pages, screenshots, PDFs | [→](/integrations/puppeteer) |

#### Media (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [2Slides](/integrations/2slides) | Presentation creation: slides, templates, exports | [→](/integrations/2slides) |
| [AllVoiceLab](/integrations/allvoicelab) | Voice synthesis: voices, speech generation | [→](/integrations/allvoicelab) |
| [Cloudinary](/integrations/cloudinary) | Media management: images, video, transformations | [→](/integrations/cloudinary) |
| [Kaltura](/integrations/kaltura) | Video platform: uploads, playback, analytics | [→](/integrations/kaltura) |
| [Mux](/integrations/mux) | Video infrastructure: live streaming, VOD, analytics | [→](/integrations/mux) |
| [Spotify](/integrations/spotify) | Music streaming: playback, playlists, search | [→](/integrations/spotify) |
| [Vimeo](/integrations/vimeo) | Video hosting: uploads, players, analytics | [→](/integrations/vimeo) |
| [YouTube](/integrations/youtube) | Video platform: search, playlists, channels, uploads | [→](/integrations/youtube) |

#### Education (5)

| Integration | What It Does | Page |
|------------|-------------|------|
| [10x Genomics](/integrations/10x-genomics) | Genomics data analysis: datasets, experiments | [→](/integrations/10x-genomics) |
| [Coursera](/integrations/coursera) | Online learning: courses, certificates, progress | [→](/integrations/coursera) |
| [Duolingo](/integrations/duolingo) | Language learning: lessons, progress, streaks | [→](/integrations/duolingo) |
| [Khan Academy](/integrations/khan-academy) | Free education: courses, exercises, progress | [→](/integrations/khan-academy) |
| [Udemy](/integrations/udemy) | Online courses: enrollments, progress, certificates | [→](/integrations/udemy) |

#### Other (3)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Aqara](/integrations/aqara) | Smart home: sensors, switches, hubs | [→](/integrations/aqara) |
| [Baidu Map](/integrations/baidu-map) | Chinese mapping: geocoding, directions, places | [→](/integrations/baidu-map) |
| [Mapbox](/integrations/mapbox) | Maps and location: geocoding, directions, tiles | [→](/integrations/mapbox) |

---

## Troubleshooting Integrations

| Problem | Cause | Fix |
|---------|-------|-----|
| Google service shows "Disconnected" | OAuth token expired or revoked | Re-authorize in Integrations tab |
| App status shows "Expired" | MCP OAuth token expired | Click **Reconnect** in Apps tab |
| App status shows "Error" | API key invalid or service unreachable | Check error message; re-enter credentials |
| OAuth popup doesn't open | Browser popup blocker | Allow popups for neotask.ai |
| Integration connected but agent can't use it | Agent doesn't have the app enabled | Enable the app in the agent's Apps tab |
| "Not connected" after app restart | Token refresh failed | Click **Reconnect**; tokens persist across restarts normally |

For general troubleshooting, see [Troubleshooting](troubleshooting.md).

---

### [MCP Auth & OAuth Setup](https://neotask.ai/docs/mcp-auth-oauth)

## Start From The App Detail

1. Open **Apps** and select the service.
2. Read the authentication choices shown for that provider.
3. Use OAuth when it is offered.
4. Add a key or manual credential only when the provider requires it.
5. Return to the app detail and confirm the connected state.

![Searchable app picker for choosing a provider](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

This guide explains how app connections work in Neotask, which setup path applies to each provider, and how to complete high-friction OAuth and credential flows step by step.

---

## Start With The Setup Mode

Every app in Neotask falls into one of these setup modes:

1. Managed OAuth
2. Manual OAuth
3. API key
4. Manual credentials
5. Custom URL first
6. No auth or local runtime

Do not start with generic reconnect advice until you know which mode the provider actually uses.

---

## Managed OAuth

Use this when the provider supports a direct Neotask-led OAuth flow.

### Step-by-Step

1. Open the provider in the Apps tab.
2. Click Connect.
3. Complete the browser-based consent flow.
4. Return to Neotask.
5. Confirm the provider moves to `connected`.

Return to **Apps** and verify that the provider appears in the connected group before testing an agent request.

![Connected app catalog after authorization](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

If the flow fails repeatedly and the provider is known to be `managed_dcr_broken`, stop retrying the same path and move to the documented fallback.

### Important Surface Rule

For MCP providers, the auth flow lives in **Apps**, not the Google Workspace **Integrations** surface.

Examples that belong in **Apps** even when they use OAuth:

- Stripe
- Slack
- Salesforce
- HubSpot
- Airtable

---

## Manual OAuth

Use this when the provider needs an app registration or pre-registered OAuth app first.

### Step-by-Step

1. Open the provider card in Neotask.
2. Copy the exact callback URL shown there.
3. Open the provider developer portal.
4. Create or open the OAuth app.
5. Add the callback URL exactly.
6. Add the required scopes exactly.
7. Copy the `clientId`.
8. Copy the `clientSecret` if the provider uses one.
9. Paste the values into Neotask.
10. Start the connection flow and approve access.

### Important Callback Rule

The callback host often points at the Neotask MCP auth service. That is expected. Use the exact callback URL from Neotask rather than inventing your own redirect URL.

---

## API Key

Use this when the provider only needs an API key or token.

### Step-by-Step

1. Create or copy the API key from the provider.
2. Open the provider card in Neotask.
3. Paste the key into the required field.
4. Save the connection.
5. Re-test the provider.

---

## Manual Credentials

Use this when the provider needs structured credentials such as `baseUrl`, `instanceUrl`, account IDs, token IDs, or tenant-specific hosts.

### Step-by-Step

1. Gather every required field before saving.
2. Confirm the correct instance or tenant URL.
3. Paste each value exactly.
4. Save the connection.
5. Re-test the provider.

If a provider needs manual credentials, do not switch the caller into OAuth unless the provider card explicitly says to do so.

---

## Custom URL First

Some providers cannot be validated until the caller saves the correct instance URL first.

### Support Rule

If the provider asks for `baseUrl`, `instanceUrl`, `workspaceUrl`, `organizationUrl`, or another tenant-specific URL, confirm that value before discussing OAuth.

### High-Value Examples

1. Benchling
   Use the correct tenant-specific Benchling URL before saving the API key.
2. Visier
   Use the correct vanity or tenant URL before saving credentials.
3. NetSuite
   Use the correct account-specific host and account ID.

---

## No Auth Or Local Runtime

Some providers do not behave like cloud-hosted OAuth services.

### JetBrains

JetBrains depends on a local IDE and plugin path. If the supported IDE is not open, or the plugin/runtime is not active, tools can be unavailable even though the product surface looks fine.

### Support Rule

Do not force no-auth or local-runtime providers into OAuth troubleshooting.

---

## Provider Playbooks

These are the highest-friction OAuth flows that support should know cold.

### Slack

1. Create a Slack app from scratch in the Slack developer portal.
2. Add the exact Neotask callback URL to Slack.
3. Add the required user scopes shown by Neotask.
4. Copy the Client ID and Client Secret.
5. Paste them into Neotask and reconnect.

### Figma

1. Create the Figma app in the Figma developer portal.
2. Add the exact Neotask callback URL.
3. Request the `mcp:connect` scope.
4. Copy the Client ID and Client Secret.
5. Paste them into Neotask and reconnect.

Figma is sensitive to the exact app type and the restricted MCP scope.

### Airtable

1. Register the Airtable OAuth integration.
2. Add the exact Neotask callback URL.
3. Add the required Airtable data, schema, webhook, and user-email scopes shown by Neotask.
4. Copy the Client ID and Client Secret.
5. Paste them into Neotask and reconnect.

### Box

1. Create a custom Box app with OAuth 2.0 user authentication.
2. Add the exact Neotask callback URL.
3. Enable the required Box application scopes.
4. Copy the Client ID and Client Secret.
5. Paste them into Neotask and reconnect.

### Canva

1. Create the Canva integration.
2. Add the exact Neotask callback URL.
3. Copy the Client ID and Client Secret.
4. Paste them into Neotask and reconnect.

### monday.com

1. Create the monday.com app.
2. Add the exact Neotask callback URL.
3. Copy the Client ID and Client Secret.
4. Paste them into Neotask and reconnect.

### Asana

1. Create the Asana app in the developer console.
2. Add the exact Neotask callback URL.
3. Copy the Client ID and Client Secret.
4. Paste them into Neotask and reconnect.

Do not improvise extra scopes for Asana if the product truth says the provider does not use them.

### Salesforce

1. Open Salesforce App Manager and create or edit the Connected App.
2. Enable OAuth settings.
3. Paste the exact Neotask callback URL.
4. Add the required API and refresh scopes.
5. Copy the Consumer Key and Consumer Secret.
6. Paste them into Neotask and reconnect.

### Microsoft Family

Use this pattern for Microsoft Teams, Microsoft 365, Azure, Azure DevOps, and PowerBI:

1. Open Microsoft Entra.
2. Create or open the app registration.
3. Add the exact Neotask callback URL as a Web redirect URI.
4. Copy the Application (client) ID.
5. Create or copy the client secret if required.
6. Paste those values into Neotask.
7. Re-run consent in the correct Microsoft tenant.

---

## Connected But Tools Still Fail

If auth looks healthy but tools still fail, check this order:

1. Wrong scope.
2. Wrong workspace or provider account.
3. Missing scopes.
4. Wrong custom URL or instance URL.
5. Local runtime or plugin not active.
6. Company flow using a tenant connection, or tenant flow using a company connection.

For a dedicated runbook, use [Support MCP Scope And Runtime](./support-mcp-scope-and-runtime.md).

---

## When To Stop And Ask Support

Escalate when:

1. The callback URL in Neotask matches the provider portal, but the provider still rejects it.
2. The provider is marked `managed_dcr_broken` or `unreachable`.
3. The provider is connected in one scope and still fails in another.
4. The flow requires company-specific or admin-specific guidance that is not on the provider card.

---

## Related Docs

- [Support MCP Auth And OAuth](./support-mcp-auth.md)
- [Support MCP Scope And Runtime](./support-mcp-scope-and-runtime.md)
- [Support MCP Provider Exceptions](./support-mcp-provider-exceptions.md)
- [App Setup Troubleshooting](./app-setup-troubleshooting.md)

---

### [Supported Apps](https://neotask.ai/docs/mcp-integrations)

## Browse And Add An App

1. Open **Apps**.
2. Select **Add app**.
3. Search the catalog by service or capability.
4. Open the app and complete its authentication flow.

![Searchable app picker](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

Neotask connects to **1200+ external services** through the Apps tab, letting your agents
interact with tools, APIs, and platforms through natural language.
Each app has a dedicated page with full capability breakdowns, example prompts, and pricing.

For built-in integrations (Google Workspace, AI providers, messaging platforms), see [Integrations](integrations.md).

---

## Browsing Apps on the Website

### Landing Page Grid

On the [homepage](/) at neotask.ai, scroll down to the **"1200+ Integrations. One Platform."** section. You'll see:

- A dense grid of **service logos**, each clickable, linking to that app's detail page
- **16 category filter chips**: hover or click any chip to highlight matching apps
- Each chip shows its live count based on the current app catalog

### App Detail Pages

Click any logo on the landing page (or navigate directly to
`neotask.ai/integrations/{id}`) to open its detail page. Every detail page includes:

- **Service logo** at the top
- **Category badge** (e.g. "Finance")
- **Tagline**: one-line description
- **3 key capability bullets**: the top things this app can do
- **"What You Can Do" section**: detailed breakdown of all capabilities organized by sub-feature
- **"Try Asking" section**: 6-8 example natural language prompts you can use
- **"Pro Tips" section**: power-user advice for getting the most out of the app
- **Pricing comparison**: side-by-side cards for Single Agent ($49/mo) vs Enterprise ($200/mo)
- **Related apps**: 4 cards at the bottom linking to similar services in the same category

---

## Using Apps in the Desktop App

### Finding the Apps Tab

1. Open the Neotask desktop app
2. Click your agent name in the left sidebar
3. Click the **Apps** tab in the agent sub-navigation (alongside Chat, Skills, Integrations, etc.)

### What You See

The Apps page separates the services already connected to the agent from the rest of the catalog.

![App catalog with connected and available services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

- A searchable, filterable registry of all supported apps
- Each app card shows: name, description, category, status badge, and auth type
- **Pin/favorite** any app by clicking the pin icon; pinned apps appear at the top
- **Status badges** indicate readiness: dev, testing, ready, connected, error

### Connecting an Integration

1. Find the integration by browsing categories or using the search bar
2. Click the app card to expand it
3. Click **Connect**
4. Complete the auth flow:

| Auth Type | What Happens |
|-----------|-------------|
| **OAuth** | A browser window opens. Sign in to the service, authorize access. Tokens are stored and refreshed automatically. |
| **API Key** | An input field appears. Paste your API key. It's encrypted with AES-256-GCM and stored locally. |
| **None** | No auth needed; the integration works immediately after enabling. |

5. The status badge updates to **Connected** when successful

### Connection Status Reference

| Status | What It Means | What to Do |
|--------|--------------|------------|
| **Connected** | Auth is valid and the integration is ready to use | Nothing needed; you're good |
| **Pending** | Auth flow started but not yet completed | Complete the OAuth popup or enter your API key |
| **Expired** | Token has expired | Click **Reconnect** to re-authorize |
| **Error** | Auth failed | Check the error message; may need to re-enter credentials |
| **None** | Not yet connected | Click **Connect** to start |

---

## Complete App Directory

Every app listed below has a dedicated detail page at
`neotask.ai/integrations/{id}` with full capabilities, example prompts, and pricing.

### Productivity (26)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Airtable](/integrations/airtable) | Spreadsheet-database hybrid: manage bases, tables, records, views | [→](/integrations/airtable) |
| [Anytype](/integrations/anytype) | Local-first knowledge management: pages, objects, relations | [→](/integrations/anytype) |
| [Archbee](/integrations/archbee) | Documentation platform: create and manage docs, spaces | [→](/integrations/archbee) |
| [Asana](/integrations/asana) | Project management: tasks, projects, teams, timelines | [→](/integrations/asana) |
| [Atlassian](/integrations/atlassian) | Jira + Confluence suite: issues, projects, pages, spaces | [→](/integrations/atlassian) |
| [Atono](/integrations/atono) | Workflow automation platform | [→](/integrations/atono) |
| [Backdocket](/integrations/backdocket) | Legal practice management: cases, documents, billing | [→](/integrations/backdocket) |
| [Basecamp](/integrations/basecamp) | Team collaboration: projects, todos, messages, schedules | [→](/integrations/basecamp) |
| [BoldSign](/integrations/boldsign) | E-signatures: send, track, manage signature requests | [→](/integrations/boldsign) |
| [Buildable](/integrations/buildable) | Developer workflow automation | [→](/integrations/buildable) |
| [Cal.com](/integrations/cal-com) | Open-source scheduling: manage calendars, bookings, availability | [→](/integrations/cal-com) |
| [ClickUp](/integrations/clickup) | Project management: tasks, docs, goals, dashboards | [→](/integrations/clickup) |
| [Coda](/integrations/coda) | All-in-one doc: docs, tables, automations | [→](/integrations/coda) |
| [Confluence](/integrations/confluence) | Team wiki: spaces, pages, comments, search | [→](/integrations/confluence) |
| [Evernote](/integrations/evernote) | Note-taking: notebooks, notes, tags, search | [→](/integrations/evernote) |
| [Jira](/integrations/jira) | Issue tracking: projects, boards, sprints, epics | [→](/integrations/jira) |
| [Linear](/integrations/linear) | Issue tracking: projects, cycles, teams, labels | [→](/integrations/linear) |
| [Make](/integrations/make) | Automation platform: scenarios, modules, connections | [→](/integrations/make) |
| [Microsoft To Do](/integrations/microsoft-todo) | Task management: lists, tasks, reminders | [→](/integrations/microsoft-todo) |
| [Monday.com](/integrations/monday) | Work management: boards, items, automations | [→](/integrations/monday) |
| [Notion](/integrations/notion) | All-in-one workspace: pages, databases, relations | [→](/integrations/notion) |
| [Obsidian](/integrations/obsidian) | Markdown knowledge base: vaults, notes, links | [→](/integrations/obsidian) |
| [Roam Research](/integrations/roam-research) | Networked note-taking: pages, blocks, references | [→](/integrations/roam-research) |
| [Smartsheet](/integrations/smartsheet) | Enterprise work management: sheets, reports, dashboards | [→](/integrations/smartsheet) |
| [Todoist](/integrations/todoist) | Task management: projects, tasks, labels, filters | [→](/integrations/todoist) |
| [Trello](/integrations/trello) | Kanban boards: boards, lists, cards, checklists | [→](/integrations/trello) |
| [Wrike](/integrations/wrike) | Enterprise project management: projects, tasks, timelines | [→](/integrations/wrike) |

### Design (11)

| Integration | What It Does | Page |
|------------|-------------|------|
| [21st.dev](/integrations/21st-dev) | AI component marketplace | [→](/integrations/21st-dev) |
| [Adobe XD](/integrations/adobe-xd) | UI/UX design: artboards, components, prototypes | [→](/integrations/adobe-xd) |
| [Canva](/integrations/canva) | Graphic design: templates, images, branding | [→](/integrations/canva) |
| [FigJam](/integrations/figjam) | Collaborative whiteboard: brainstorming, diagramming | [→](/integrations/figjam) |
| [Figma](/integrations/figma) | Interface design: files, components, styles, variables | [→](/integrations/figma) |
| [Framer](/integrations/framer) | Web design and publishing: sites, pages, CMS | [→](/integrations/framer) |
| [InVision](/integrations/invision) | Design collaboration: prototypes, boards, specs | [→](/integrations/invision) |
| [Miro](/integrations/miro) | Visual collaboration: boards, widgets, frames | [→](/integrations/miro) |
| [Penpot](/integrations/penpot) | Open-source design: files, components, teams | [→](/integrations/penpot) |
| [Sketch](/integrations/sketch) | macOS design tool: artboards, symbols, libraries | [→](/integrations/sketch) |
| [Zeplin](/integrations/zeplin) | Design handoff: specs, style guides, assets | [→](/integrations/zeplin) |

### Communication (17)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Discord](/integrations/discord) | Community messaging: servers, channels, messages, roles | [→](/integrations/discord) |
| [Front](/integrations/front) | Shared inbox: conversations, comments, assignments | [→](/integrations/front) |
| [Google Meet](/integrations/google-meet) | Video meetings: schedule, join, manage meetings | [→](/integrations/google-meet) |
| [Infobip](/integrations/infobip) | Omnichannel messaging: SMS, WhatsApp, email, voice | [→](/integrations/infobip) |
| [Intercom](/integrations/intercom) | Customer messaging: conversations, users, articles | [→](/integrations/intercom) |
| [Knock](/integrations/knock) | Notification infrastructure: workflows, channels, feeds | [→](/integrations/knock) |
| [Line](/integrations/line) | Messaging platform: messages, groups, media | [→](/integrations/line) |
| [Loom](/integrations/loom) | Video messaging: recordings, transcripts, sharing | [→](/integrations/loom) |
| [Mailgun](/integrations/mailgun) | Email API: send, receive, track, validate | [→](/integrations/mailgun) |
| [Mailjet](/integrations/mailjet) | Email delivery: campaigns, transactional, templates | [→](/integrations/mailjet) |
| [Microsoft Teams](/integrations/microsoft-teams) | Team collaboration: chats, channels, meetings | [→](/integrations/microsoft-teams) |
| [SendGrid](/integrations/sendgrid) | Email delivery: transactional, marketing, analytics | [→](/integrations/sendgrid) |
| [Slack](/integrations/slack) | Team messaging: channels, messages, threads, files | [→](/integrations/slack) |
| [Telegram](/integrations/telegram) | Messaging: chats, groups, bots, channels | [→](/integrations/telegram) |
| [Twilio](/integrations/twilio) | Communications API: SMS, voice, video, WhatsApp | [→](/integrations/twilio) |
| [WhatsApp](/integrations/whatsapp) | Messaging: conversations, media, templates | [→](/integrations/whatsapp) |
| [Zoom](/integrations/zoom) | Video conferencing: meetings, webinars, recordings | [→](/integrations/zoom) |

### Analytics (12)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Amplitude](/integrations/amplitude) | Product analytics: events, funnels, cohorts, retention | [→](/integrations/amplitude) |
| [Audiense](/integrations/audiense) | Audience intelligence: segments, insights, reports | [→](/integrations/audiense) |
| [Datadog](/integrations/datadog) | Monitoring: metrics, traces, logs, dashboards | [→](/integrations/datadog) |
| [Google Analytics](/integrations/google-analytics) | Web analytics: traffic, conversions, audiences | [→](/integrations/google-analytics) |
| [Grafana](/integrations/grafana) | Observability: dashboards, alerts, data sources | [→](/integrations/grafana) |
| [Heap](/integrations/heap) | Digital analytics: auto-capture events, funnels | [→](/integrations/heap) |
| [Hotjar](/integrations/hotjar) | Behavior analytics: heatmaps, recordings, surveys | [→](/integrations/hotjar) |
| [Mixpanel](/integrations/mixpanel) | Product analytics: events, funnels, A/B tests | [→](/integrations/mixpanel) |
| [New Relic](/integrations/new-relic) | Observability: APM, infrastructure, logs | [→](/integrations/new-relic) |
| [Plausible](/integrations/plausible) | Privacy-friendly web analytics | [→](/integrations/plausible) |
| [PostHog](/integrations/posthog) | Product analytics: events, feature flags, session recordings | [→](/integrations/posthog) |
| [Segment](/integrations/segment) | Customer data platform: sources, destinations, audiences | [→](/integrations/segment) |

### Sales & CRM (12)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Apollo](/integrations/apollo) | Sales intelligence: leads, contacts, sequences | [→](/integrations/apollo) |
| [Apollo.io](/integrations/apollo-io) | B2B data and engagement: prospecting, outreach | [→](/integrations/apollo-io) |
| [Attio](/integrations/attio) | Next-gen CRM: contacts, companies, pipelines | [→](/integrations/attio) |
| [Clay](/integrations/clay) | Data enrichment: lead scoring, waterfall enrichment | [→](/integrations/clay) |
| [Close](/integrations/close) | Sales CRM: leads, activities, pipelines, calling | [→](/integrations/close) |
| [Freshsales](/integrations/freshsales) | CRM: contacts, deals, accounts, territories | [→](/integrations/freshsales) |
| [Gong](/integrations/gong) | Revenue intelligence: call recordings, analytics | [→](/integrations/gong) |
| [HubSpot](/integrations/hubspot) | CRM platform: contacts, deals, tickets, marketing | [→](/integrations/hubspot) |
| [Outreach](/integrations/outreach) | Sales engagement: sequences, tasks, analytics | [→](/integrations/outreach) |
| [Pipedrive](/integrations/pipedrive) | Sales CRM: deals, contacts, activities, pipelines | [→](/integrations/pipedrive) |
| [Salesforce](/integrations/salesforce) | Enterprise CRM: accounts, opportunities, reports | [→](/integrations/salesforce) |
| [Zoho CRM](/integrations/zoho-crm) | CRM: leads, contacts, deals, analytics | [→](/integrations/zoho-crm) |

### Development (24)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Apify](/integrations/apify) | Web scraping and automation: actors, runs, datasets | [→](/integrations/apify) |
| [AWS](/integrations/aws) | Cloud infrastructure: EC2, S3, Lambda, IAM | [→](/integrations/aws) |
| [Azure DevOps](/integrations/azure-devops) | DevOps: repos, pipelines, boards, artifacts | [→](/integrations/azure-devops) |
| [Bitbucket](/integrations/bitbucket) | Git hosting: repos, PRs, pipelines, issues | [→](/integrations/bitbucket) |
| [Bitrise](/integrations/bitrise) | Mobile CI/CD: builds, workflows, apps | [→](/integrations/bitrise) |
| [Buildkite](/integrations/buildkite) | CI/CD: pipelines, builds, agents | [→](/integrations/buildkite) |
| [Cloudflare](/integrations/cloudflare) | Edge platform: DNS, CDN, workers, security | [→](/integrations/cloudflare) |
| [Docker](/integrations/docker) | Containerization: images, containers, registries | [→](/integrations/docker) |
| [Firebase](/integrations/firebase) | App platform: auth, database, hosting, functions | [→](/integrations/firebase) |
| [GitHub](/integrations/github) | Code hosting: repos, PRs, issues, actions, releases | [→](/integrations/github) |
| [GitLab](/integrations/gitlab) | DevOps platform: repos, CI/CD, issues, registries | [→](/integrations/gitlab) |
| [Jenkins](/integrations/jenkins) | CI/CD server: jobs, builds, pipelines | [→](/integrations/jenkins) |
| [JetBrains](/integrations/jetbrains) | IDE platform: projects, plugins, configurations | [→](/integrations/jetbrains) |
| [JFrog](/integrations/jfrog) | Artifact management: repos, packages, security | [→](/integrations/jfrog) |
| [LaunchDarkly](/integrations/launchdarkly) | Feature flags: flags, environments, segments | [→](/integrations/launchdarkly) |
| [Liveblocks](/integrations/liveblocks) | Real-time collaboration: rooms, presence, storage | [→](/integrations/liveblocks) |
| [MinIO](/integrations/minio) | Object storage: buckets, objects, policies | [→](/integrations/minio) |
| [Neon](/integrations/neon) | Serverless Postgres: databases, branches, endpoints | [→](/integrations/neon) |
| [Netlify](/integrations/netlify) | Web hosting: sites, deploys, functions, forms | [→](/integrations/netlify) |
| [PlanetScale](/integrations/planetscale) | Serverless MySQL: databases, branches, deploy requests | [→](/integrations/planetscale) |
| [Railway](/integrations/railway) | Cloud deployment: services, environments, variables | [→](/integrations/railway) |
| [Sentry](/integrations/sentry) | Error tracking: issues, events, releases, alerts | [→](/integrations/sentry) |
| [Supabase](/integrations/supabase) | Backend platform: database, auth, storage, functions | [→](/integrations/supabase) |
| [Vercel](/integrations/vercel) | Frontend deployment: projects, deployments, domains | [→](/integrations/vercel) |

### Databases (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [BigQuery](/integrations/bigquery) | Data warehouse: datasets, queries, tables | [→](/integrations/bigquery) |
| [ClickHouse](/integrations/clickhouse) | Column-oriented database: queries, tables, inserts | [→](/integrations/clickhouse) |
| [DynamoDB](/integrations/dynamodb) | NoSQL database: tables, items, queries | [→](/integrations/dynamodb) |
| [Elasticsearch](/integrations/elasticsearch) | Search engine: indices, documents, queries | [→](/integrations/elasticsearch) |
| [MongoDB](/integrations/mongodb) | Document database: collections, documents, aggregations | [→](/integrations/mongodb) |
| [PostgreSQL](/integrations/postgresql) | Relational database: tables, queries, schemas | [→](/integrations/postgresql) |
| [Redis](/integrations/redis) | In-memory database: keys, streams, pub/sub | [→](/integrations/redis) |
| [Snowflake](/integrations/snowflake) | Cloud data warehouse: warehouses, databases, queries | [→](/integrations/snowflake) |

### Data (6)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Algolia](/integrations/algolia) | Search API: indices, records, analytics | [→](/integrations/algolia) |
| [Astra DB](/integrations/astra-db) | Serverless Cassandra: databases, collections, documents | [→](/integrations/astra-db) |
| [Atlan](/integrations/atlan) | Data catalog: assets, glossary, lineage | [→](/integrations/atlan) |
| [Baserow](/integrations/baserow) | Open-source database: tables, fields, rows | [→](/integrations/baserow) |
| [Bright Data](/integrations/brightdata) | Web data platform: collectors, datasets, proxies | [→](/integrations/brightdata) |
| [Meilisearch](/integrations/meilisearch) | Search engine: indexes, documents, facets | [→](/integrations/meilisearch) |

### Finance (13)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Adfin](/integrations/adfin) | Financial operations: invoices, payments, reconciliation | [→](/integrations/adfin) |
| [Aiera](/integrations/aiera) | Financial intelligence: earnings, transcripts, analysis | [→](/integrations/aiera) |
| [Airwallex](/integrations/airwallex) | Global payments: accounts, transfers, FX | [→](/integrations/airwallex) |
| [Alpaca](/integrations/alpaca) | Trading API: orders, positions, market data | [→](/integrations/alpaca) |
| [Alpha Vantage](/integrations/alphavantage) | Market data: stocks, forex, crypto, indicators | [→](/integrations/alphavantage) |
| [Brex](/integrations/brex) | Corporate finance: cards, expenses, budgets | [→](/integrations/brex) |
| [Mercury](/integrations/mercury) | Banking: accounts, transactions, payments | [→](/integrations/mercury) |
| [Plaid](/integrations/plaid) | Banking API: accounts, transactions, identity | [→](/integrations/plaid) |
| [QuickBooks](/integrations/quickbooks) | Accounting: invoices, expenses, reports | [→](/integrations/quickbooks) |
| [Ramp](/integrations/ramp) | Corporate cards: transactions, receipts, budgets | [→](/integrations/ramp) |
| [Square](/integrations/square) | Commerce: payments, catalog, customers, locations | [→](/integrations/square) |
| [Stripe](/integrations/stripe) | Payments: charges, subscriptions, invoices, customers | [→](/integrations/stripe) |
| [Xero](/integrations/xero) | Accounting: invoices, contacts, bank reconciliation | [→](/integrations/xero) |

### Marketing (11)

| Integration | What It Does | Page |
|------------|-------------|------|
| [ActiveCampaign](/integrations/activecampaign) | Email marketing and CRM: campaigns, automations, contacts | [→](/integrations/activecampaign) |
| [Ahrefs](/integrations/ahrefs) | SEO: site audit, keywords, backlinks, rank tracking | [→](/integrations/ahrefs) |
| [AirOps](/integrations/airops) | AI workflow automation for marketing | [→](/integrations/airops) |
| [Beehiiv](/integrations/beehiiv) | Newsletter platform: subscribers, campaigns, analytics | [→](/integrations/beehiiv) |
| [Brevo](/integrations/brevo) | Email marketing: campaigns, contacts, automations | [→](/integrations/brevo) |
| [BuiltWith](/integrations/builtwith) | Technology profiling: tech stacks, leads, trends | [→](/integrations/builtwith) |
| [ConvertKit](/integrations/convertkit) | Creator email marketing: subscribers, sequences, forms | [→](/integrations/convertkit) |
| [Customer.io](/integrations/customer-io) | Customer messaging: campaigns, segments, workflows | [→](/integrations/customer-io) |
| [Drip](/integrations/drip) | E-commerce email: automations, subscribers, campaigns | [→](/integrations/drip) |
| [Klaviyo](/integrations/klaviyo) | E-commerce marketing: flows, campaigns, segments | [→](/integrations/klaviyo) |
| [Mailchimp](/integrations/mailchimp) | Email marketing: audiences, campaigns, automations | [→](/integrations/mailchimp) |

### AI & ML (13)

| Integration | What It Does | Page |
|------------|-------------|------|
| [AgentOps](/integrations/agentops) | AI agent observability: sessions, events, analytics | [→](/integrations/agentops) |
| [Anthropic](/integrations/anthropic) | Claude AI: completions, messages, embeddings | [→](/integrations/anthropic) |
| [Arize Phoenix](/integrations/arize-phoenix) | ML observability: traces, evaluations, datasets | [→](/integrations/arize-phoenix) |
| [Cohere](/integrations/cohere) | NLP API: embeddings, reranking, generation | [→](/integrations/cohere) |
| [ElevenLabs](/integrations/elevenlabs) | Voice AI: text-to-speech, voice cloning, dubbing | [→](/integrations/elevenlabs) |
| [Hugging Face](/integrations/hugging-face) | ML platform: models, datasets, spaces | [→](/integrations/hugging-face) |
| [Kagi Search](/integrations/kagi-search) | Premium search: web search, summarization | [→](/integrations/kagi-search) |
| [Label Studio](/integrations/label-studio) | Data labeling: projects, tasks, annotations | [→](/integrations/label-studio) |
| [OpenAI](/integrations/openai) | GPT AI: completions, embeddings, images, audio | [→](/integrations/openai) |
| [Pinecone](/integrations/pinecone) | Vector database: indexes, vectors, queries | [→](/integrations/pinecone) |
| [Replicate](/integrations/replicate) | ML inference: models, predictions, deployments | [→](/integrations/replicate) |
| [Stability AI](/integrations/stability-ai) | Image generation: text-to-image, upscaling, editing | [→](/integrations/stability-ai) |
| [Weaviate](/integrations/weaviate) | Vector database: schemas, objects, searches | [→](/integrations/weaviate) |

### Security (9)

| Integration | What It Does | Page |
|------------|-------------|------|
| [1Password](/integrations/1password) | Password management: vaults, items, secrets | [→](/integrations/1password) |
| [Asgardeo](/integrations/asgardeo) | Identity management: users, apps, connections | [→](/integrations/asgardeo) |
| [Auth0](/integrations/auth0) | Identity platform: users, connections, rules | [→](/integrations/auth0) |
| [BoostSecurity](/integrations/boostsecurity) | DevSecOps: policies, findings, scanners | [→](/integrations/boostsecurity) |
| [CrowdStrike](/integrations/crowdstrike) | Endpoint security: detections, hosts, policies | [→](/integrations/crowdstrike) |
| [Dependabot](/integrations/dependabot) | Dependency security: alerts, updates, PRs | [→](/integrations/dependabot) |
| [Snyk](/integrations/snyk) | Security scanning: vulnerabilities, fixes, dependencies | [→](/integrations/snyk) |
| [SonarQube](/integrations/sonarqube) | Code quality: issues, rules, quality gates | [→](/integrations/sonarqube) |
| [Vault](/integrations/vault) | Secrets management: secrets, policies, auth | [→](/integrations/vault) |

### Infrastructure (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Aiven](/integrations/aiven) | Managed data services: databases, queues, caches | [→](/integrations/aiven) |
| [Ansible](/integrations/ansible) | IT automation: playbooks, inventories, roles | [→](/integrations/ansible) |
| [Axiom](/integrations/axiom) | Log management: datasets, queries, dashboards | [→](/integrations/axiom) |
| [Azure](/integrations/azure) | Microsoft cloud: VMs, storage, functions, networking | [→](/integrations/azure) |
| [Chef](/integrations/chef) | Infrastructure automation: cookbooks, recipes, nodes | [→](/integrations/chef) |
| [Kubernetes](/integrations/kubernetes) | Container orchestration: pods, services, deployments | [→](/integrations/kubernetes) |
| [Prometheus](/integrations/prometheus) | Monitoring: metrics, alerts, queries | [→](/integrations/prometheus) |
| [Pulumi](/integrations/pulumi) | Infrastructure as code: stacks, resources, outputs | [→](/integrations/pulumi) |

### Storage (7)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Backblaze B2](/integrations/backblaze-b2) | Cloud storage: buckets, files, uploads | [→](/integrations/backblaze-b2) |
| [Box](/integrations/box) | Cloud content: files, folders, collaborations | [→](/integrations/box) |
| [Dropbox](/integrations/dropbox) | File storage: files, folders, sharing, sync | [→](/integrations/dropbox) |
| [Google Drive](/integrations/google-drive) | Cloud storage: files, folders, permissions, search | [→](/integrations/google-drive) |
| [OneDrive](/integrations/onedrive) | Microsoft cloud storage: files, folders, sharing | [→](/integrations/onedrive) |
| [S3](/integrations/s3) | AWS object storage: buckets, objects, policies | [→](/integrations/s3) |
| [Wasabi](/integrations/wasabi) | Cloud storage: buckets, objects, hot storage | [→](/integrations/wasabi) |

### Browser (4)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Browser Automation](/integrations/browser-automation) | Automated browsing: navigation, clicks, scraping | [→](/integrations/browser-automation) |
| [Browserbase](/integrations/browserbase) | Cloud browsers: sessions, screenshots, recording | [→](/integrations/browserbase) |
| [Playwright](/integrations/playwright) | Browser testing: pages, selectors, assertions | [→](/integrations/playwright) |
| [Puppeteer](/integrations/puppeteer) | Chrome automation: pages, screenshots, PDFs | [→](/integrations/puppeteer) |

### Media (8)

| Integration | What It Does | Page |
|------------|-------------|------|
| [2Slides](/integrations/2slides) | Presentation creation: slides, templates, exports | [→](/integrations/2slides) |
| [AllVoiceLab](/integrations/allvoicelab) | Voice synthesis: voices, speech generation | [→](/integrations/allvoicelab) |
| [Cloudinary](/integrations/cloudinary) | Media management: images, video, transformations | [→](/integrations/cloudinary) |
| [Kaltura](/integrations/kaltura) | Video platform: uploads, playback, analytics | [→](/integrations/kaltura) |
| [Mux](/integrations/mux) | Video infrastructure: live streaming, VOD, analytics | [→](/integrations/mux) |
| [Spotify](/integrations/spotify) | Music streaming: playback, playlists, search | [→](/integrations/spotify) |
| [Vimeo](/integrations/vimeo) | Video hosting: uploads, players, analytics | [→](/integrations/vimeo) |
| [YouTube](/integrations/youtube) | Video platform: search, playlists, channels, uploads | [→](/integrations/youtube) |

### Education (5)

| Integration | What It Does | Page |
|------------|-------------|------|
| [10x Genomics](/integrations/10x-genomics) | Genomics data analysis: datasets, experiments | [→](/integrations/10x-genomics) |
| [Coursera](/integrations/coursera) | Online learning: courses, certificates, progress | [→](/integrations/coursera) |
| [Duolingo](/integrations/duolingo) | Language learning: lessons, progress, streaks | [→](/integrations/duolingo) |
| [Khan Academy](/integrations/khan-academy) | Free education: courses, exercises, progress | [→](/integrations/khan-academy) |
| [Udemy](/integrations/udemy) | Online courses: enrollments, progress, certificates | [→](/integrations/udemy) |

### Other (3)

| Integration | What It Does | Page |
|------------|-------------|------|
| [Aqara](/integrations/aqara) | Smart home: sensors, switches, hubs | [→](/integrations/aqara) |
| [Baidu Map](/integrations/baidu-map) | Chinese mapping: geocoding, directions, places | [→](/integrations/baidu-map) |
| [Mapbox](/integrations/mapbox) | Maps and location: geocoding, directions, tiles | [→](/integrations/mapbox) |

---

## Troubleshooting App Connections

| Problem | Cause | Fix |
|---------|-------|-----|
| Status shows "Expired" | OAuth token expired | Click **Reconnect** in Apps tab to re-authorize |
| Status shows "Error" | API key invalid or service unreachable | Check error message; re-enter credentials |
| OAuth popup doesn't open | Browser popup blocker | Allow popups for neotask.ai |
| Integration connected but agent can't use it | Agent doesn't have the app enabled | Enable the app in the agent's Apps tab |
| "Not connected" after desktop app restart | Token refresh failed | Click **Reconnect**; tokens persist across restarts normally |

For general troubleshooting, see [Troubleshooting](troubleshooting.md).

---

### [Neotask Support Agent Capability Reference](https://neotask.ai/docs/neotask-support-agent-capability-reference)

## Agent Screens

Use the main chat for the active support conversation. Use Overview for recent sessions, scheduled work, and activity.

![Main agent chat](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/chat-main-1280.webp)

This page is a supplemental deep dive. If you want the full one-page overview of what the support agent can do, start with [Neotask Support Call Agent](./neotask-support-call-agent.md).

This is the detailed capability reference for Neotask support call agent.

If you want the short marketing version, use [Neotask Support Call Agent](./neotask-support-call-agent.md).

If you want the exhaustive version of what the live support system can actually inspect, explain, and do, use this page.

---

## Support Phone Number

Call the support line here:

- `+1 (331) 296-2762`

---

## Support Access Levels

The support call agent can help at different depths depending on the support context available during the call.

### 1. General support

The agent can help with:

- public setup guidance
- public troubleshooting guidance
- signup friction
- pre-activation checkout questions
- public app-connection docs
- where to find the right docs or next step

### 2. Recognized member support

This is the mode where the support agent can use account-linked support context.

In the current support system, this is driven by account-linked recognition and authenticated member support context.

At this level, the support agent can inspect:

- tenant account state
- subscription and credit state
- recent sessions and jobs
- recent automation failures
- recent gateway and app runtime errors
- recent failure-linked log excerpts
- relevant support case history
- relevant workspace/company memory context

### 3. Company-scoped support context

Some support questions are tenant-wide. Others are really about one company or sector.

When the issue is company-specific, the support agent can scope into that company and inspect:

- company profile and health hints
- company schedules
- company approvals
- company automation failures
- company memory files and setup notes
- company-specific app runtime errors

If there are multiple companies available, support can scope into the right one.

---

## Live Specialist Lanes

The support stack currently routes into these specialist areas.

### Guest support

Best for:

- signup help
- pre-activation checkout issues
- public support guidance
- “where do I start?”

What it relies on:

- curated support documentation
- MCP setup documentation
- support case follow-up when it is needed

### Setup and troubleshooting support

Best for:

- app setup problems
- first launch problems
- local runtime confusion
- desktop app onboarding friction

What it can inspect:

- live account overview
- recent sessions
- recent jobs
- recent failures
- recent technical diagnostics
- recent failure-linked logs
- relevant memory and workspace notes

### MCP and OAuth support

Best for:

- app connection status
- OAuth reconnect issues
- callback URL questions
- API key setup questions
- “connected but still fails” questions

What it can inspect:

- curated provider guidance
- exact app-setup docs
- live account overview
- recent activity
- technical diagnostics
- recent failure-linked logs
- relevant memory context

### Billing and account access support

Best for:

- current plan
- subscription state
- credit balances
- top-up balances
- seat and company allowance questions
- recent usage questions
- account-state follow-up

What it can inspect:

- live account overview
- live billing summary
- recent activity
- diagnostics when the caller thinks something is blocked by a technical problem
- memory context when historical setup or support notes matter

### Teams, channels, automation, and enterprise support

Best for:

- recent automation failures
- approvals blocking work
- channels not delivering
- recent sessions or workflow activity
- company-specific workflow troubleshooting
- enterprise/admin operational questions

What it can inspect:

- live account overview
- recent activity
- technical diagnostics
- recent failure-linked logs
- memory context
- curated support docs for supported public workflows

---

## Live Read Capabilities

These are the main data classes the current support system can inspect through its support-specific data access layer.

### Live account overview

The agent overview is the useful screen when support needs recent sessions, scheduled work, or agent state. Open it before reading account-wide activity to the caller.

![Agent overview with sessions, scheduled work, and activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

The support agent can retrieve a structured live account overview that includes:

- tenant name
- plan
- subscription status
- credit balance
- top-up balance
- overage/blocking state
- employee seats purchased and granted
- company or sector allowance
- additional paid company or sector slots
- dedicated onboarding state
- member counts by role
- company list
- selected company detail
- latest support case summary
- latest support call summary
- recent usage rollup
- recent account credit movement

This is the most useful “what does my account actually look like right now?” support view.

### Live billing summary

The support agent can produce a live billing summary with:

- current plan
- subscription status
- main credit balance
- top-up balance
- current company/sector count
- company/sector allowance
- paid add-on slot count
- employee seats purchased and granted
- seven-day usage cost and event count
- seven-day error-event count
- subscription period end
- dedicated onboarding status when applicable
- latest recent credit movement

This is the main live answer for:

- “What plan am I on?”
- “How much balance do I have left?”
- “How much did I use recently?”
- “Do I have any add-on company slots?”

### Recent activity and workflow state

The support agent can inspect recent activity including:

- recent agent sessions
- recent account sessions
- recent jobs
- recent workflow activity
- recent schedules
- recent approvals
- recent usage events
- recent automation events
- recent gateway errors
- recent app runtime errors

This is the main support path for:

- “What happened recently?”
- “Did anything fail?”
- “Did the system run my workflow?”
- “Were there approvals waiting?”

### Technical diagnostics

The support agent can inspect technical evidence including:

- bounded warning/error log excerpts
- recent gateway runtime error excerpts
- recent app runtime error excerpts
- recent workflow error excerpts
- bounded transcript excerpts for incidents
- recent usage events with error codes
- blocked schedules with auth/runtime indicators
- recent failed or blocked automation events

This is the main tool path for:

- “Why did this fail?”
- “Can you check the logs?”
- “Do you see runtime errors?”
- “Was this blocked by auth or by the runtime?”
- “Can you inspect the recent logs and tell me the most likely cause?”

### Memory and workspace context

The support agent can inspect:

- relevant company memory excerpts
- relevant setup notes
- relevant support notes
- bounded previews from specific memory files when requested
- a list of available company memory files

This is the main tool path for:

- “Look at the notes for this company.”
- “Check the prior setup memory.”
- “Preview the file that explains this workflow.”

### Support knowledge and app-setup reference

The support agent can also use support documentation and app-setup guidance for:

- curated support docs
- exact provider setup guidance
- callback URL information
- auth field requirements
- provider-specific scope notes
- provider exception runbooks

This is what it should use when the answer is documentation-first rather than account-state-first.

---

## Evidence Classes It Can Use

The support agent can work from the following evidence classes when they are relevant:

- plan and subscription state
- credit and top-up state
- recent usage summaries
- recent billing movement summaries
- company and sector counts
- employee-seat counts
- recent agent sessions
- recent account sessions
- recent jobs
- recent workflow activity
- recent schedules
- recent approvals
- recent automation events
- recent gateway errors
- recent app runtime errors
- recent log excerpts connected to failed events
- relevant support cases
- relevant support call history
- relevant support notes
- relevant workspace/company memory excerpts
- bounded transcript excerpts
- bounded file or artifact previews

---

## Actions The Support Agent Can Take Today

These are the live support-side actions that exist now.

### Create a support case

The agent can create a structured support case when:

- the issue needs follow-up
- the issue needs escalation
- the call is ending before full resolution
- a durable record should exist

### Update a support case

The agent can update an existing support case with:

- status
- priority
- route or department
- notes
- summary
- outcome

### Send a support doc link

The agent can send a doc link by SMS when:

- a specific runbook or support doc is the right next step
- the user wants a written reference after the call

### Send a general follow-up SMS

The agent can send an SMS follow-up message when:

- the user needs next steps
- the user needs a written reminder
- a lightweight follow-up helps without escalation

### Send human handoff options

The agent can send the user:

- the support Intercom path
- the support scheduling path when configured

This is the right move when the issue needs human follow-up.

---

## Things It Can Explain Well

The support agent is especially useful for explaining:

- what failed recently
- whether the likely cause was auth, runtime, routing, approval, or billing state
- whether a schedule is blocked
- whether a company or workflow has recent support-relevant notes
- whether the issue already shows up in recent errors
- whether the account state matches what the user thinks should be true

Good example questions:

- “Check my recent automation failures and tell me the top likely cause.”
- “Check whether auth or runtime is blocking this schedule.”
- “Look at my recent app runtime errors.”
- “Tell me whether approvals are waiting.”
- “Look at the recent notes for this company.”
- “Check my recent sessions and tell me what happened.”

---

## When It Will Escalate

Some issues still need human follow-up.

When that happens, the support agent can:

- summarize what it already confirmed
- create or update the support case
- preserve the account context and issue history
- send the next step by SMS when that is useful

---

## Example Request Catalog

### Billing and plan

- “What plan am I on?”
- “What is my top-up balance right now?”
- “How much did I use in the last week?”
- “Do I have any extra company slots?”
- “How many seats are on this account?”

### Recent system state

- “What happened recently on my account?”
- “Check my recent sessions.”
- “Check my recent jobs.”
- “Check my recent failures and tell me the likely cause.”

### Dynamic troubleshooting

- “Check the recent logs for this issue.”
- “Tell me whether this looks like auth, runtime, routing, approval, or schedule trouble.”
- “Look at the last failed workflow and explain what actually happened.”
- “Check whether the system recorded any app or gateway errors around that time.”
- “Look at recent tool execution problems.”

### Technical issues

- “Look at recent runtime errors.”
- “Check whether the gateway has been failing.”
- “Show me the likely cause of the last blocked automation.”
- “Do you see auth-related schedule blocking?”

### Workflow and company issues

- “Help me with the operations company.”
- “Check whether this company has pending approvals.”
- “Look at the notes for the marketing company.”
- “Preview the memory file for this workflow.”

### Support continuity

- “I already called about this. Can you check the prior support case?”
- “Look at the prior support notes before we continue.”
- “Send me the doc link for this issue.”
- “Create a support case and escalate this.”

---

## Best Way To Use This Agent

If you want the best help, tell the support agent:

- what failed
- whether it is happening now or happened recently
- which company, workflow, app, or channel is involved
- whether you want diagnosis, explanation, or a follow-up action

The more specific you are about what to inspect, the better it performs.

---

## Related Docs

- [Neotask Support Call Agent](./neotask-support-call-agent.md)
- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)
- [Support Routing Overview](./neotask-support-routing-overview.md)
- [Support Common Resolution Paths](./neotask-support-common-resolution-paths.md)
- [Support Teams, Channels, And Automation](./neotask-support-teams-channels-and-automation.md)

---

### [Neotask Authenticated Support Access](https://neotask.ai/docs/neotask-support-authenticated-account-support)

## Account Screens Support May Ask You To Open

Use **Settings → License** for account and plan status. Use **Usage** for model and cost records. Open a session for its full conversation and tool history.

![License plan and account status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/license-plan-1280.webp)

This page is a supplemental deep dive. If you want the full one-page overview of what the support agent can do, start with [Neotask Support Call Agent](./neotask-support-call-agent.md).

This page explains what changes when Neotask support call agent is working with authenticated member support context instead of general support only.

If you want the short version:

- guest support can help with public guidance and basic triage
- authenticated member support can work from your real account, workspace, and support context

---

## Support Number

To reach the support call agent, dial:

- `+1 (331) 296-2762`

If your phone number is already linked to your account, support can often move into member context much faster.

---

## What Authenticated Support Means

Authenticated support means the support call is no longer limited to general support guidance.

Instead, the support agent can work from the account and workspace context tied to the support session, including the specific tenant and, when needed, the specific company or sector you are calling about.

That is what turns the support line from a generic phone assistant into a real account-aware support specialist.

---

## What Authenticated Support Can Do

Once support is in authenticated member context, it can help with:

- your actual plan and subscription state
- your real balance, credits, and top-up state
- your recent usage and cost direction
- your employee-seat and company or sector allowance state
- your recent sessions and recent job activity
- your recent approvals or requires-action state
- your recent automation failures
- your recent gateway runtime issues
- your recent desktop-app runtime issues that were reported to the platform
- relevant support notes, memories, and prior context
- relevant transcript excerpts or bounded file previews when the issue clearly depends on them
- the likely cause of recent failed workflows, jobs, sessions, or company automations

This is the level where support can often explain what happened without making you manually reconstruct everything.

---

## What It Can Pull Into A Support Session

Depending on the issue, authenticated support can pull in:

- live billing and subscription context
- recent sessions and workflow activity
- recent approvals and automation failures
- recent gateway and app runtime errors
- recent log excerpts connected to the failure
- relevant support notes and support history
- relevant workspace notes and company notes
- relevant transcript excerpts when the issue depends on what happened during a session
- relevant file or artifact previews when the output itself is the problem

This is what allows the support line to troubleshoot dynamically instead of only repeating documentation.

---

## Example Questions That Authenticated Support Should Be Able To Answer

### Billing and usage

Open **Usage** when the question concerns model activity, agent attribution, or cost. Select the relevant date range before reading the record back to the caller.

![Usage overview with model activity and cost](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

- “Tell me my current plan, credit state, and recent usage.”
- “Check whether billing is the reason my account feels blocked.”
- “Look at my subscription state and tell me if anything is wrong.”

### Recent failures

- “Check what failed recently.”
- “Look at my recent automation failures and explain the likely cause.”
- “Do you see any recent gateway or runtime errors?”
- “Check the recent logs and tell me whether this was an auth problem or a runtime problem.”

### Session and workflow support

- “Look at my recent session and tell me why it stopped.”
- “Check whether approvals are blocking this workflow.”
- “Look at the recent jobs for this company and tell me what failed.”
- “Check whether this route or schedule was blocked before it ever ran.”

### Company-specific support

- “Help me with the marketing company first.”
- “Now switch to the operations company and check that one too.”
- “Compare what failed between the two companies.”

### Context-heavy troubleshooting

- “Look at the recent support notes before I explain this again.”
- “Check the relevant memory or workspace notes for this company.”
- “Preview the relevant output file because that is the thing that broke.”

---

## What To Ask For If You Want The Best Results

If you want the best member-support help, say exactly what you want it to inspect.

Examples:

- “Check my recent billing and tell me if anything is blocking me.”
- “Look at my last failed automation and tell me the likely cause.”
- “Check my app runtime errors from today.”
- “Look at the recent support notes for this company.”
- “Preview the relevant file output because it is malformed.”

That phrasing tells the support agent which context to load instead of making it guess.

---

## When It Will Create A Support Case

Some issues still need follow-up.

When that happens, the support agent can:

- summarize what it already found
- create a support case
- preserve the context for the next step

---

## Related Docs

- [Neotask Support Agent Capability Reference](./neotask-support-agent-capability-reference.md)
- [Neotask Support Call Agent](./neotask-support-call-agent.md)
- [Support Routing Overview](./neotask-support-routing-overview.md)
- [Support Login And Account Access](./neotask-support-login-and-account-access.md)
- [Support Common Resolution Paths](./neotask-support-common-resolution-paths.md)

---

### [Neotask Support Call Agent](https://neotask.ai/docs/neotask-support-call-agent)

## Voice Session Screen

The voice session keeps the current briefing, tool results, and governed next actions visible while you speak with the agent.

![Voice briefing with governed next actions](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/voice-briefing-1280.webp)

This is the main page for what the Neotask support call agent can do.

If a user wants to know whether the support line is actually worth using, this should be the only page they need to read.

---

## Call The Support Line

Dial the public support line here:

- `+1 (331) 296-2762`

---

## One-Page Summary

Neotask support call agent is designed to feel like a real support specialist, not a generic phone tree.

Its job is to:

- get you unblocked quickly
- explain what is actually happening in your account or workspace
- inspect live support context instead of forcing you to guess
- route you into the right support lane without making you repeat yourself over and over
- preserve context when escalation is needed

At a high level, it can:

- answer public support and setup questions
- help authenticated account holders at a much deeper level
- inspect live billing, usage, account, session, workflow, and diagnostics context
- inspect relevant support notes, workspace notes, memories, and bounded excerpts when needed
- inspect recent logs, runtime errors, failed jobs, approvals, schedules, and session context to troubleshoot live issues
- explain what failed, what is blocked, and what the likely cause is
- send follow-up links or SMS messages
- create or update support cases
- escalate to a human with preserved context when the issue is too sensitive or too deep

It is not just a FAQ bot.

For account-level help, the support line uses strict authentication before it works from live account, billing, workspace, or company data. General questions can be answered immediately, but private account support only opens after the call is tied to the right account context.

---

## Why Use It

The support call agent is worth using when you want:

- fast live troubleshooting instead of waiting on email
- help understanding what failed and why
- help checking billing, usage, credits, or subscription status
- help with app setup, gateway problems, channels, routing, or automations
- help with app connections, OAuth, API keys, and provider-specific setup
- help finding the right next step when the platform is behaving differently than you expected

When your call is recognized as a member support call, the support agent can work from live account context instead of only public documentation.

That means it can often tell you:

- what your plan and billing state actually are
- whether recent automations failed
- whether recent runtime or gateway errors were recorded
- whether approvals, schedules, or routing issues are blocking work
- whether support notes or prior context already explain the issue
- whether the issue looks like auth, runtime, routing, approval, schedule, or company-scoping trouble

---

## Why It Feels Like Real Support

The value of the support line is not just that it can answer docs questions.

The real value is that it can actively investigate what happened.

That means it can often:

- inspect what happened in your recent sessions
- check whether a job ran, failed, stalled, or was blocked
- inspect recent gateway or app runtime errors
- check whether an automation was blocked by auth, approvals, routing, or scheduling
- look at recent company or workspace support context
- pull in relevant notes, memory, transcript excerpts, or file previews when the issue depends on them
- tell you the likely cause instead of making you manually guess

That is what makes it feel closer to a real product specialist than a basic support bot.

---

## Support Access Levels

The support line can help at different depths depending on whether the call is just a general support question or tied to a recognized account and company.

### General support

This level focuses on:

- public setup guidance
- public troubleshooting guidance
- signup friction
- pre-activation checkout questions
- public app-connection docs
- where to find the right docs or next step

### Account-linked support

This is where the support agent can use real account-linked context.

At this level it can inspect:

- tenant account state
- subscription and credit state
- recent sessions and jobs
- recent automation failures
- recent gateway and app runtime errors
- relevant support case history
- relevant workspace or company memory context

### Company or sector support

Some support questions are account-wide. Others are really about one company or sector.

When the issue is company-specific, the support agent can scope into that company and inspect:

- company profile and health hints
- company schedules
- company approvals
- company automation failures
- company memory files and setup notes
- company-specific app runtime errors

If there are multiple companies or sectors available, support can scope into the right one instead of guessing.

---

## What It Can Help With

### Setup and onboarding

- desktop app setup
- first launch problems
- gateway startup and local runtime issues
- microphone, voice, and call setup
- onboarding confusion and “what do I do next” questions

### Billing and account help

- active plan and subscription state
- credits and top-up balance questions
- recent usage and cost direction
- billing-access confusion
- invoice, checkout, and account-state follow-up

### App connections and authentication

- MCP app connection status
- OAuth reconnect and provider setup guidance
- API key and manual credential guidance
- callback URL and field-entry confusion
- “it says connected but still fails” troubleshooting

### Autonomous companies, channels, and automation

- channel routing problems
- delivery failures
- approval blocks
- scheduled automations not firing
- employee and company routing confusion
- company-specific troubleshooting when the call is properly scoped
- autonomous company dashboard navigation
- task-edit guidance inside one company
- SEO and company document troubleshooting
- company video-generation follow-up when the issue is tied to SEO pages or generated assets

### Technical support

- recent automation failures
- recent gateway runtime errors
- recent app runtime errors that were reported to the platform
- recent session activity
- recent job activity
- recent support-relevant context for the affected workspace or company
- live troubleshooting based on what the platform actually recorded

---

## What Authenticated Support Means

Authenticated support means the call is no longer limited to general guidance.

Instead, the support agent can work from the account and workspace context tied to the support session, including the specific tenant and, when needed, the specific company or sector you are calling about.

That is what turns the support line from a generic phone assistant into a real account-aware support specialist.

Once support is in authenticated member context, it can help with:

- your actual plan and subscription state
- your real balance, credits, and top-up state
- your recent usage and cost direction
- your employee-seat and company or sector allowance state
- your recent sessions and recent job activity
- your recent approvals or requires-action state
- your recent automation failures
- your recent gateway runtime issues
- your recent desktop-app runtime issues that were reported to the platform
- relevant support notes, memories, and prior context
- relevant transcript excerpts or bounded file previews when the issue clearly depends on them
- the likely reason a recent workflow, session, or automation failed

This is the level where support can often explain what happened without making you manually reconstruct everything.

---

## What Makes It Better Than A Basic Support Bot

Neotask support call agent is not just reading a static FAQ.

When the call is in the right support scope, it can combine:

- curated support documentation
- live account summaries
- recent activity summaries
- recent technical diagnostics
- recent session, job, and workflow evidence
- bounded transcript or note excerpts when they are relevant
- company and workspace support context when the issue depends on it

That lets it answer questions like:

- “What failed recently?”
- “Can you check whether my automation died because auth expired?”
- “Do you see any app errors from today?”
- “Can you tell me whether I’m actually on Business or Enterprise?”
- “Can you see whether there are approvals waiting?”
- “Can you look at what happened in my recent session?”
- “Can you check the recent logs and tell me why this workflow stopped?”
- “Can you tell me whether this was blocked by routing, auth, or runtime?”
- “How do I tell if the SEO page video was created for this company?”

---

## What It Can See During Support

These are the main kinds of account and workspace context the support system can use when relevant:

- live account overview
- live billing summary
- recent sessions
- recent jobs
- recent workflow activity
- recent schedules
- recent approvals
- recent automation failures
- recent gateway errors
- recent app runtime errors
- recent log excerpts tied to failures
- recent support cases
- relevant support notes
- relevant workspace or company memory excerpts
- bounded transcript excerpts
- bounded file or artifact previews
- curated support docs and exact provider setup guidance

---

## Actions It Can Take Today

These are the live support-side actions it can already take.

### Create a support case

The agent can create a structured support case when:

- the issue needs follow-up
- the issue needs escalation
- the call is ending before full resolution
- a durable record should exist

### Update a support case

The agent can update an existing support case with:

- status
- priority
- route or department
- notes
- summary
- outcome

### Send a support doc link

The agent can send a doc link by SMS when:

- a specific runbook or support doc is the right next step
- the user wants a written reference after the call

### Send a general follow-up SMS

The agent can send an SMS follow-up message when:

- the user needs next steps
- the user needs a written reminder
- a lightweight follow-up helps without escalation

### Send human handoff options

The agent can send the user:

- the support Intercom path
- the support scheduling path when configured

This is the right move when the issue needs human follow-up.

---

## When It May Slow Down

Some support lookups are fast because they come from preloaded call context.

Other lookups take longer because the agent is checking live systems, recent diagnostics, or deeper account context.

When that happens, the support agent should tell you it is checking instead of going silent. For example:

- “Checking that for you now.”
- “Let me pull the recent activity first.”
- “I’m looking at the recent support context now.”

That is normal for deeper troubleshooting.

---

## Example Things To Ask

### Fast account questions

- “What plan am I on right now?”
- “How much top-up balance do I have left?”
- “Can you check my recent usage?”
- “Is my subscription active?”
- “How many seats are on this account?”
- “Do I have any extra company or sector slots?”

### Troubleshooting questions

- “Can you check whether anything failed recently?”
- “Look at my recent automation failures.”
- “Can you see any runtime errors from the app?”
- “Check whether my gateway has been erroring.”
- “Can you tell me why messages are not delivering?”

### App connection questions

- “My Google connection says connected but my automation still fails.”
- “Can you help me reconnect Salesforce?”
- “What auth method does this app actually need?”
- “Can you tell me what setup fields I’m missing?”

### Workspace and company questions

- “Can you check whether there are approvals waiting?”
- “Can you look at my recent jobs?”
- “Can you tell me what company this route is targeting?”
- “Can you help me figure out why this company workflow is not running?”
- “How do I open the SEO settings for this company?”
- “How do I tell whether this company page video was created?”

### Deeper support questions

- “Can you look at the recent support notes for this issue?”
- “Can you check the recent session context before I explain everything again?”
- “Can you preview the relevant file or artifact if that is what is breaking?”
- “Look at my recent billing and tell me if anything is blocking me.”
- “Check my app runtime errors from today.”
- “Preview the relevant file output because it is malformed.”

---

## Best Way To Use It

You will get the best results if you tell the support agent:

- what failed
- whether it is happening now or happened recently
- which app, company, workflow, or channel is affected
- whether the issue is billing, setup, app auth, routing, or automation
- whether you want an explanation, diagnosis, or follow-up action

Good examples:

- “Check my recent automation failures and tell me the top likely cause.”
- “Look at my billing and tell me whether anything is blocking my account.”
- “Help me reconnect HubSpot and tell me exactly what fields I need.”
- “Look at my recent session and tell me why the workflow stopped.”

The more specific you are about what to inspect, the better it performs.

---

## Optional Deep Dives

You should not need these pages just to understand what the support agent can do. This page is the main overview.

Use these only if you want more detail on one slice:

- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)
- [Neotask Support Agent Capability Reference](./neotask-support-agent-capability-reference.md)
- [Neotask Autonomous Companies Support](./neotask-support-autonomous-companies.md)
- [Support Routing Overview](./neotask-support-routing-overview.md)
- [Support Teams, Channels, And Automation](./neotask-support-teams-channels-and-automation.md)
- [Support Billing And Credits](./neotask-support-billing-and-credits.md)
- [Support MCP And OAuth](./neotask-support-mcp-and-oauth.md)

---

### [Neotask Support Routing Overview](https://neotask.ai/docs/neotask-support-routing-overview)

## Route By Product Surface

Troubleshoot owns runtime and connection problems. Apps owns authentication. Usage and Billing own cost questions. Company Approvals owns work waiting on review.

![Troubleshoot settings with the system check and repair controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

Neotask support line is designed to help callers get to the right support path quickly.

## Support Routing Model

There are two top-level access modes:

- `guest support`
- `member support`

Guest support is for callers who are not yet recognized as a saved member or employee phone number. Member support is for callers whose number matches a saved tenant voice route or active company employee record.

## General Support Scope

App authentication questions begin in the app catalog. The connected state shown on the card determines whether support should reconnect the app or investigate a later runtime failure.

![App catalog and connected services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

General support can help with:

- sign-up friction
- checkout and payment problems before activation
- where to log in or start
- public troubleshooting docs
- basic app installation guidance

## Member Support Departments

For usage or cost questions, open **Usage** and select the caller's date range before reviewing the provider and model records.

![Usage overview with provider and model activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

Recognized member callers can be routed into these departments:

- `App setup and troubleshooting`
- `MCP auth and OAuth setup`
- `Billing and account access`
- `Teams, channels, automation, and enterprise support`

## Recognition Rules

Support recognizes member callers from these saved product records:

- active company employee phone numbers
- saved tenant `voice_call` route phone numbers
- saved company `voice_call` route phone numbers

If a phone number matches more than one record, support may continue with general support first and then move into the right account path once the workspace is confirmed.

## Route Outcomes

Calls should end in one of these support outcomes:

- resolved live on the call
- resolved with follow-up SMS
- requires manual follow-up
- scheduled support session
- escalated to human support
- caller disconnected before resolution

## Related Docs

- [Neotask Support Call Agent](./neotask-support-call-agent.md)
- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)

---

### [Neotask SOP Retrieval Troubleshooting](https://neotask.ai/docs/neotask-support-sop-retrieval-troubleshooting)

## Verify The Procedure And Its Source

1. Open **SOPs** and select the procedure.
2. Confirm the active version and publication state.
3. Open the related run and inspect the cited source.
4. Review a pending revision before publishing it.

![SOP procedure library](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-procedure-library-1280.webp)

Use this page when the caller says:

- “My agent isn’t using my SOP.”
- “I uploaded an SOP but the company ignored it.”
- “An SOP I just published isn’t showing up in the work.”
- “How do I approve the SOP my agent proposed?”

---

## How SOPs Are Meant To Work

Two things to keep in the back of your mind before troubleshooting:

- A company follows its **published** SOPs. A draft, including an SOP an agent proposed, is not active until a person publishes it.
- When an agent starts work, the company automatically attaches the SOPs that are most relevant to that task. The caller does not have to point an agent at a specific SOP; the right ones are found by meaning.

So “the agent isn’t using my SOP” is almost always one of: the SOP is not published, it is in a different company, or it simply was not the most relevant guidance for that particular task.

---

## First Checks (in order)

### 1. Is the SOP actually published?

The most common cause. A proposed or uploaded SOP that is still a **pending draft** is not yet active guidance.

Have the caller open the company and go to the **SOPs** tab:

- if the SOP is a pending draft (for example one an agent proposed), they need to review and **publish** it
- if it was uploaded but never published, the same applies

Once published, it becomes active operating guidance.

### 2. Is it in the right company?

SOPs are scoped to one company. A company only ever uses its own SOPs.

Confirm the SOP lives in the **same company** the caller is running the work in. An SOP published in Company A will never appear for Company B.

### 3. Was it just published?

Newly published SOPs become available to retrieval right away, matching updates on publish, so a fresh SOP can be picked up on the very next task. There is no long wait.

If the caller published an SOP and then re-ran a task, the new SOP should be eligible immediately. If a task was already in flight when they published, have them start a fresh task or action so the new guidance is considered.

### 4. Was this task simply not relevant to that SOP?

Open the source review for the related run. It shows which procedure and supporting material the agent actually received.

![SOP source review](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-source-review-1280.webp)

The company attaches the SOPs that best fit the task by meaning. If an SOP did not show up for a given task, it may be because that task was not closely related to it, not because anything is broken.

Have the caller try a task that clearly relates to the SOP’s subject. If relevant SOPs attach there, retrieval is working as designed.

---

## “An Agent Proposed An SOP, How Do I Approve It?”

When the proposal changes an existing procedure, use the diff to review the exact additions and removals before publishing.

![Pending SOP revision diff](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r1/sops-pending-diff-1280.webp)

Route the caller to the **SOPs** tab in the company.

1. open the company → **SOPs**
2. find the pending draft the agent proposed
3. review it, if it updates an existing SOP, they can use the diff to see exactly what changed
4. choose to **approve / publish**, **edit then publish**, or **discard**

Only after they publish does it become guidance the company follows.

---

## Good To Know: Graceful Fallback

Retrieval is built so SOPs never simply “stop working.” If the meaning-based matching is momentarily unavailable, the company falls back to the older keyword-based selection and still attaches SOPs to the agent. So a caller will not hit a state where no SOPs are applied at all, at worst, matching is briefly less precise, then recovers on its own. This is not something the caller needs to act on.

---

## When To Escalate

Escalate only after the first checks pass, that is, when the caller confirms:

- the SOP is **published** (not a draft), and
- it is in the **same company** they are running work in, and
- they ran a task that is clearly relevant to the SOP’s subject, and
- relevant SOPs are still not being applied across multiple fresh tasks

At that point it is no longer a setup or expectations issue, and the case should be escalated with the company name, the SOP in question, and an example task where the relevant SOP was not applied.

---

## Related Docs

- [Neotask Autonomous Companies Support](./neotask-support-autonomous-companies.md)
- [Neotask Company Dashboard Tab Map](./neotask-support-company-dashboard-tab-map.md)
- [Auto, SOP Authoring & Review](./auto-companies-sops-authoring.md)
- [Auto, Smart SOP Retrieval](./auto-companies-sops-semantic-retrieval.md)

---

### [Security](https://neotask.ai/docs/security)

## Review Safety Controls In The App

Open **Settings → Privacy and Data** for data-sharing controls. Open **Settings → Permissions** for Safe Mode. Open **Approvals** for actions waiting on a person.

![Session data sharing control, off by default](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-consent-1280.webp)

## Overview

Neotask is built with security as a core principle. Every layer, from license activation to data storage to network communication, is designed to protect your data and prevent unauthorized access.

## License Security

- **Device Binding**: Each license is cryptographically bound to a single device via SHA-256 fingerprint.
- **HMAC-SHA256 Request Signing**: All API requests are signed with a per-device secret, nonce, and timestamp.
- **Timing-Safe Comparison**: All secret comparisons use constant-time algorithms to prevent timing attacks.
- **Token Lifecycle**: Access and refresh token expiry is controlled server-side via signed JWT claims, enabling instant revocation.
- **Automatic Revalidation**: Your license is checked every 6 hours.
- **Offline Grace Period**: 72 hours of offline access before revalidation is required.
- **Remote Revocation**: Licenses can be revoked server-side instantly.

## Two-Factor Authentication (TOTP)

Neotask supports optional (but recommended) two-factor authentication for dashboard access.

- Compatible with Google Authenticator, Authy, 1Password, and any TOTP-compatible app.
- Backup codes are provided during setup. These codes are SHA-256 hashed and are one-time use only.
- TOTP can be enabled or disabled at any time from the **Security** section in your dashboard settings.

## Encryption

### Data at Rest

- **AES-256-GCM**: All tokens, secrets, and API keys are encrypted at rest.
- **Machine-Derived Keys**: Encryption keys are derived via scrypt from your device identity.
- **No Plaintext Storage**: Tokens are never stored in plaintext.

### Data in Transit

- **TLS 1.3**: All API communication travels over HTTPS.
- **WebSocket Secure (WSS)**: Real-time gateway communication is encrypted.
- **HMAC Signing**: Every request includes a nonce and timestamp signature.

### Environment Secrets

- **Two-Layer Encryption**: Build-time environment encryption uses separate keys.
- **No Hardcoded Secrets**: Zero API keys or tokens exist in source code.
- **Log Redaction**: Sensitive data is automatically stripped from logs.

## Desktop App Security

### Electron Hardening

Neotask applies strict Electron security settings to protect the desktop application:

- `sandbox: true`: The renderer process runs in a sandboxed environment.
- `contextIsolation: true`: No direct access to the main process from the UI.
- `nodeIntegration: false`: No Node.js APIs are exposed to the UI.
- `webSecurity: true`: Same-origin policy is enforced.

### Content Security Policy (CSP)

- `default-src 'self'`: Only resources from the app itself are loaded.
- `script-src 'self'`: No external scripts are permitted.
- Popup windows are blocked.
- Navigation is blocked to prevent redirect attacks.
- Page reload is blocked to prevent state manipulation.

### App Integrity

- **ASAR Integrity**: SHA-256 hash verification of the packaged app.
- **Version Attestation**: A server-signed manifest with kill switch capability.
- **Code Obfuscation**: JavaScript obfuscation is applied in production builds.
- **Hard Fail Mode**: The app blocks entirely on any integrity failure.

## Network Security

### Gateway Isolation

- The gateway runs on **loopback only** (127.0.0.1), meaning there is zero external network exposure.
- No incoming connections from outside your machine are accepted.
- Session grants have a 10-minute lifetime, are device-bound, and are HMAC-signed.

### 3-Strike Lockout

- After 3 consecutive gateway operation failures, all operations are blocked.
- A manual reset is required to restore access.
- This mechanism prevents brute-force attempts.

## API Security

### Authentication Methods

| Method | Used For | Security Level |
|--------|----------|----------------|
| **JWT Bearer Token** | Web dashboard, API calls | Standard (90-day expiry) |
| **License HMAC** | Desktop app operations | High (per-device secret) |
| **Session Grants** | Gateway operations | Very High (10-min, HMAC-signed) |
| **TOTP** | Dashboard login | Additional factor |

### Rate Limiting

| Endpoint | Limit |
|----------|-------|
| Contact form | 5 requests per 15 minutes |
| Login attempts | 10 requests per 15 minutes |
| Analytics/tracking | 30 requests per 60 seconds |

### Input Validation

- All IPC parameters are validated before processing.
- All API inputs are sanitized.
- Protections against SQL injection, XSS, and command injection are in place.

## Provider Key Security (BYOK)

When using Bring Your Own Key (BYOK) mode, Neotask applies additional safeguards to your API keys:

- API keys are encrypted with AES-256-GCM before storage.
- Keys are never logged or exposed in error messages.
- Keys are displayed in masked form in the dashboard (only the last 4 characters are visible).
- Secure deletion is performed when a key is removed.

### Provider Destination Controls

In hosted-credit mode, Neotask sends model requests only to the official HTTPS
hosts assigned to the selected provider. A provider name cannot be paired with
an arbitrary target host, and upstream redirects are blocked before they can
carry a Neotask-managed or BYOK credential to another destination.

Custom provider base URLs are available only on direct BYOK gateway paths,
where the user supplies the destination and credential. They are not accepted
by the hosted-credit passthrough.

## Safe Mode

The confirmation explains the wider access that becomes available before Safe Mode is disabled.

![Safe Mode confirmation explaining the effect of disabling protection](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/safe-mode-confirmation-832.webp)

Neotask includes a Safe Mode feature that provides an execution sandbox for agents:

- Per-agent execution sandbox isolates each agent's operations.
- Sensitive operations require explicit user approval before proceeding.
- A master toggle allows you to enable or disable Safe Mode globally.
- Auto-re-enable on schedule ensures Safe Mode is reactivated after temporary changes.
- Real-time policy synchronization keeps settings consistent across all connections.

## Audit and Compliance

Approval records show the requested action and the controls available to the reviewer.

![Approval detail with the requested action and review controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-detail-1280.webp)

- All configuration changes are logged with timestamps.
- Config hashing detects unauthorized modifications.
- Usage telemetry (opt-in) enables anomaly detection.
- CORS is restricted to authorized domains only.
- Helmet security headers are applied on all web responses.

## Best Practices

1. **Enable TOTP** on your dashboard for two-factor authentication.
2. **Download backup codes** and store them securely.
3. **Set daily budgets** to prevent unexpected charges.
4. **Use BYOK mode** if you want full control over your API keys.
5. **Keep the app updated**, as auto-updates include security patches.
6. **Review agent permissions** periodically in Safe Mode settings.

---

### [Skills](https://neotask.ai/docs/skills)

## Add Or Review A Skill

1. Open **Skills**.
2. Search the catalog or open **My skills**.
3. Review required apps, tools, and configuration.
4. Open pending review when Neotask proposes a learned skill.
5. Approve it only after the source and behavior are correct.

![Skills catalog](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/skills-main-1280.webp)

## What Are Skills?

Skills are modular capabilities that extend what your Neotask agents can do. Think of them as plugins -- each skill adds a specific ability, from sending emails to controlling smart home devices to running CI/CD pipelines.

### Skill Types

| Type | Description |
|------|-------------|
| **Built-in** | Always available to every agent. These cover core operations like file management, shell execution, and web requests. No installation required. |
| **Marketplace** | Installable from the Neotask skill catalog. Browse, install, and configure with a few clicks. |
| **Custom** | User-authored skills that you write yourself. Define your own logic, API integrations, and workflows. |

### Authentication Modes

Each skill declares how it authenticates with external services:

| Auth Mode | Description |
|-----------|-------------|
| **OAuth** | Browser-based authorization flow. You sign in to the third-party service and grant Neotask access. Tokens are managed automatically. |
| **API Key** | Enter an API key or secret token. The key is stored encrypted and injected at runtime. |
| **Local App** | The skill connects to an application running on your local machine (e.g., Signal Desktop, iMessage). |
| **None** | No authentication required. The skill operates without external credentials. |

## Skill Categories

Skills are organized into six categories:

### Productivity

Tools for managing files, notes, and tasks.

- File management and cloud storage integration
- Note-taking and knowledge base operations
- Task tracking and project management
- Calendar and scheduling
- Spreadsheet and document generation

### Messaging

Integrations for email, chat, and SMS platforms.

- Email composition and inbox management
- Chat platform messaging (Slack, Discord, Teams, etc.)
- SMS and MMS sending
- Notification routing and delivery
- Message templating and bulk sends

### Media

Processing and generation for images, video, and audio.

- Image resizing, cropping, and format conversion
- Image generation and editing with AI
- Video transcoding and thumbnail extraction
- Audio transcription and text-to-speech
- Screenshot and screen recording capture

### Smart Home

Device control and home automation.

- Light, thermostat, and switch control
- Scene and routine activation
- Sensor monitoring and alerts
- Camera snapshot and stream access
- Device status reporting

### Developer

Tools for coding, building, and deploying software.

- Code generation, review, and refactoring
- Git operations and repository management
- CI/CD pipeline triggering and monitoring
- Container and deployment management
- Log analysis and error tracking

### Data

Analytics, scraping, and data processing.

- Web scraping and content extraction
- Data transformation and cleaning
- Database querying and reporting
- Chart and visualization generation
- CSV, JSON, and XML processing

## Installing Skills

### Browsing the Catalog

1. Navigate to **Skills** in the dashboard.
2. Browse by category or use the search bar to find a specific skill.
3. Each skill listing shows its name, description, category, auth mode, and required dependencies.

### One-Click Install

1. Click **Install** on the skill you want.
2. Neotask performs a dependency check to ensure all prerequisites are met:
   - Required system binaries (e.g., `ffmpeg`, `chromium`)
   - Required environment variables
   - Required companion skills
3. If all dependencies are satisfied, the skill is installed immediately.
4. If dependencies are missing, you will be prompted to resolve them before installation can proceed.

### Post-Install Configuration

After installation, you may need to configure the skill:

- **API keys** -- Enter the required key or token in the skill's settings panel.
- **Environment variables** -- Set any environment variables the skill needs (e.g., database connection strings, service URLs).
- **OAuth connections** -- Complete the OAuth authorization flow to grant the skill access to a third-party service.

### Enabling Per Agent

Skills are not automatically available to all agents. After installing a skill:

1. Open the agent's **Settings** tab.
2. Under **Skills**, toggle the skill on or off for that agent.
3. Only agents with the skill enabled can use it.

This gives you fine-grained control over what each agent can and cannot do.

## Skill Configuration

When Neotask proposes a learned skill, review its instructions and permissions before approving it.

![Learned skill proposal awaiting review](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/skill-proposal-1280.webp)

### API Key Storage

All API keys and secrets are stored encrypted at rest. Keys are decrypted only at runtime when the skill needs them, and they are never exposed in logs, session history, or API responses.

To manage API keys:

1. Navigate to **Skills** and select the installed skill.
2. Open the **Configuration** tab.
3. Enter, update, or remove API keys as needed.

### Environment Variables

Some skills require environment variables to function. These are set per skill and injected into the skill's execution environment.

- Variables are scoped to the individual skill and do not leak to other skills or agents.
- Common examples include database URLs, service endpoints, and feature flags.

### OAuth Connections

For skills that use OAuth authentication:

1. Click **Connect** in the skill's configuration panel.
2. You will be redirected to the third-party service to authorize access.
3. After authorization, the access token is stored and managed by Neotask.
4. Tokens are automatically refreshed when they expire.
5. You can disconnect at any time by clicking **Disconnect**.

### Dependency Verification

Each skill declares its dependencies, which may include:

- **Required binaries** -- System-level programs the skill needs (e.g., `git`, `node`, `ffmpeg`).
- **Required environment variables** -- Variables that must be set for the skill to operate.
- **Required skills** -- Other skills that must be installed first.

Neotask verifies these dependencies at install time and warns you if any are missing. You can also re-check dependencies at any time from the skill's settings.

## Built-in Skills

Neotask ships with over 55 built-in skills that are always available without installation. Below is an overview organized by category.

### Productivity

| Skill | Description |
|-------|-------------|
| File Manager | Read, write, move, copy, and delete files within agent workspaces |
| Directory Browser | List and navigate directory structures |
| Archive Manager | Create and extract ZIP, TAR, and GZIP archives |
| Clipboard | Copy and paste text between operations |
| Search | Full-text search across workspace files and content |
| Note Taker | Create, organize, and retrieve structured notes |
| Task Tracker | Manage to-do lists with status, priority, and due dates |
| Calendar | Schedule events and set reminders |

### Messaging

| Skill | Description |
|-------|-------------|
| Email Sender | Compose and send emails via SMTP or connected providers |
| Email Reader | Read and parse incoming emails from connected inboxes |
| Notification Router | Route alerts and notifications to the appropriate channel |
| Template Engine | Create and apply message templates with variable substitution |
| Bulk Messenger | Send messages to multiple recipients across channels |

### Media

| Skill | Description |
|-------|-------------|
| Image Processor | Resize, crop, rotate, and convert image formats |
| Screenshot Capture | Take screenshots of web pages or applications |
| Audio Transcriber | Convert speech to text using AI transcription |
| Text-to-Speech | Generate spoken audio from text |
| Video Thumbnailer | Extract thumbnail frames from video files |
| QR Code Generator | Create QR codes from text, URLs, or data |

### Smart Home

| Skill | Description |
|-------|-------------|
| Light Controller | Turn lights on/off, adjust brightness and color |
| Thermostat Manager | Read and set temperature targets |
| Switch Controller | Toggle smart switches and outlets |
| Scene Activator | Trigger pre-defined smart home scenes |
| Sensor Monitor | Read data from temperature, humidity, motion, and door sensors |
| Camera Viewer | Capture snapshots from connected security cameras |

### Developer

| Skill | Description |
|-------|-------------|
| Git Operations | Clone, commit, push, pull, branch, and merge repositories |
| Code Formatter | Auto-format code files using language-appropriate formatters |
| Linter | Run static analysis and linting across codebases |
| Test Runner | Execute test suites and report results |
| Dependency Manager | Install, update, and audit project dependencies |
| Docker Manager | Build, run, stop, and manage Docker containers |
| Deploy Trigger | Trigger deployment pipelines on connected CI/CD platforms |
| Log Analyzer | Parse and search application logs for errors and patterns |
| Port Scanner | Check open ports and service availability |
| SSL Checker | Verify SSL certificate status and expiration dates |

### Data

| Skill | Description |
|-------|-------------|
| Web Scraper | Extract structured data from web pages |
| CSV Processor | Read, write, filter, and transform CSV files |
| JSON Transformer | Parse, query, and restructure JSON data |
| XML Parser | Read and process XML documents |
| Database Query | Execute SQL queries against connected databases |
| Chart Generator | Create bar, line, pie, and scatter charts from data |
| PDF Reader | Extract text and tables from PDF documents |
| PDF Generator | Create PDF documents from text, HTML, or templates |
| Data Validator | Validate data against schemas and rules |
| Hash Generator | Compute MD5, SHA-256, and other hash digests |

---

## Skill Showcase Pages

Every skill has a **dedicated showcase page** on the website at `neotask.ai/skills/{id}`.
These pages provide detailed information about what each skill can do, example prompts, pro tips,
and pricing.

### Accessing Skill Pages

- **From the landing page:** Scroll to the skills carousel section and click any skill card
- **Direct URL:** Navigate to `neotask.ai/skills/{id}`, for example:
  - Google Workspace: [/skills/gog](/skills/gog)
  - Slack: [/skills/slack](/skills/slack)
  - GitHub: [/skills/github](/skills/github)

### What Each Showcase Page Contains

- **Skill identity:** name, category badge, and one-line tagline
- **3 key capability bullets:** top features at a glance
- **"What You Can Do" section:** detailed breakdown of all capabilities organized by sub-feature
- **"Try Asking" section:** 6-8 example natural language prompts
- **"Pro Tips" section:** power-user advice
- **Pricing comparison:** Single Agent ($49/mo) vs Enterprise ($200/mo)
- **Related skills:** 4 cards linking to similar skills in the same category

### Skills on the Landing Page

The landing page at neotask.ai features a **skills carousel** that scrolls through
highlighted skills. Each card in the carousel shows the skill name, tagline, and top 3 capabilities.
Clicking any card navigates to that skill's full showcase page.

---

### Complete Skill Directory

#### Productivity & Workspace (10)

| Skill | Tagline | Page |
|-------|---------|------|
| [Google Workspace](/skills/gog) | All 21 Google services in one unified skill | [View](/skills/gog) |
| [Notion](/skills/notion) | Search, create, and query your Notion workspace | [View](/skills/notion) |
| [Trello](/skills/trello) | Manage boards, lists, and cards conversationally | [View](/skills/trello) |
| [Obsidian](/skills/obsidian) | Search, create, and organize your Obsidian vault | [View](/skills/obsidian) |
| [Bear Notes](/skills/bear-notes) | Create, search, and tag notes in Bear app | [View](/skills/bear-notes) |
| [Apple Notes](/skills/apple-notes) | Read, write, and search Apple Notes hands-free | [View](/skills/apple-notes) |
| [Apple Reminders](/skills/apple-reminders) | Create and manage reminders synced to all devices | [View](/skills/apple-reminders) |
| [Things 3](/skills/things-mac) | Full task management for Things 3 on Mac | [View](/skills/things-mac) |
| [Canvas](/skills/canvas) | Display interactive HTML content on any device | [View](/skills/canvas) |
| [Summarizer](/skills/summarize) | Summarize articles, PDFs, and YouTube videos instantly | [View](/skills/summarize) |

#### Communication (7)

| Skill | Tagline | Page |
|-------|---------|------|
| [Slack](/skills/slack) | Send messages and manage channels hands-free | [View](/skills/slack) |
| [Discord](/skills/discord) | Send messages, polls, and reactions in Discord | [View](/skills/discord) |
| [iMessage](/skills/imsg) | Send and read iMessages directly from conversation | [View](/skills/imsg) |
| [BlueBubbles](/skills/bluebubbles) | Full iMessage control with effects and reactions | [View](/skills/bluebubbles) |
| [WhatsApp](/skills/wacli) | Send WhatsApp messages and files to anyone | [View](/skills/wacli) |
| [Email (IMAP)](/skills/himalaya) | Read, compose, and manage email from any provider | [View](/skills/himalaya) |
| [Voice Calls](/skills/voice-call) | Make and manage voice calls programmatically | [View](/skills/voice-call) |

#### Media & Creative (12)

| Skill | Tagline | Page |
|-------|---------|------|
| [Video Creator](/skills/remotion-videos) | Batch-render videos with parallel processing and upload | [View](/skills/remotion-videos) |
| [Video Frames](/skills/video-frames) | Extract frames from any video at precise timestamps | [View](/skills/video-frames) |
| [AI Image Gen](/skills/openai-image-gen) | Generate and batch-create images with multiple AI models | [View](/skills/openai-image-gen) |
| [Music Recognition](/skills/songsee) | Visualize audio with spectrograms and analysis tools | [View](/skills/songsee) |
| [GIF Search](/skills/gifgrep) | Search, preview, and download GIFs from Tenor and Giphy | [View](/skills/gifgrep) |
| [Screenshot](/skills/peekaboo) | Control your Mac screen, windows, and UI elements | [View](/skills/peekaboo) |
| [Camera](/skills/camsnap) | Capture snapshots and video from network cameras | [View](/skills/camsnap) |
| [Spotify](/skills/spotify-player) | Control Spotify playback and search from conversation | [View](/skills/spotify-player) |
| [Sonos](/skills/sonoscli) | Control Sonos speakers and multi-room audio | [View](/skills/sonoscli) |
| [Text-to-Speech](/skills/sherpa-onnx-tts) | Generate speech completely offline with zero API costs | [View](/skills/sherpa-onnx-tts) |
| [Banana AI](/skills/nano-banana-pro) | Generate and edit images with Gemini 3 Pro | [View](/skills/nano-banana-pro) |

#### Development & DevOps (10)

| Skill | Tagline | Page |
|-------|---------|------|
| [GitHub](/skills/github) | Manage repos, PRs, and code from conversation | [View](/skills/github) |
| [GitHub Issues](/skills/gh-issues) | Auto-fix GitHub issues with parallel AI agents | [View](/skills/gh-issues) |
| [Coding Agent](/skills/coding-agent) | Delegate coding tasks to Codex, Claude, or Pi | [View](/skills/coding-agent) |
| [Skill Creator](/skills/skill-creator) | Build custom skills without writing any code | [View](/skills/skill-creator) |
| [MCP Porter](/skills/mcporter) | Connect and call any MCP server from one interface | [View](/skills/mcporter) |
| [Terminal Multiplexer](/skills/tmux) | Monitor and control tmux sessions remotely | [View](/skills/tmux) |
| [Session Logs](/skills/session-logs) | Search past conversations and track token costs | [View](/skills/session-logs) |
| [Health Check](/skills/healthcheck) | Audit and harden your system security posture | [View](/skills/healthcheck) |
| [Blog Watcher](/skills/blogwatcher) | Track RSS feeds and get notified of new posts | [View](/skills/blogwatcher) |

#### AI Models (5)

| Skill | Tagline | Page |
|-------|---------|------|
| [Gemini](/skills/gemini) | Quick one-shot queries to Google Gemini models | [View](/skills/gemini) |
| [Oracle](/skills/oracle) | Bundle files and prompts to query any AI model | [View](/skills/oracle) |
| [Speech-to-Text](/skills/openai-whisper) | Transcribe audio locally with zero API costs | [View](/skills/openai-whisper) |
| [Whisper API](/skills/openai-whisper-api) | Fast cloud transcription via OpenAI Whisper API | [View](/skills/openai-whisper-api) |
| [Model Usage](/skills/model-usage) | Track AI model costs and token usage over time | [View](/skills/model-usage) |

#### Utilities & Smart Home (9)

| Skill | Tagline | Page |
|-------|---------|------|
| [1Password](/skills/1password) | Access secrets and passwords without leaving your workflow | [View](/skills/1password) |
| [Weather](/skills/weather) | Get current weather and forecasts for any location | [View](/skills/weather) |
| [Google Places](/skills/goplaces) | Search nearby places with ratings and reviews | [View](/skills/goplaces) |
| [Food Ordering](/skills/food-order) | Browse order history and reorder favorite meals | [View](/skills/food-order) |
| [Order CLI](/skills/ordercli) | Manage food delivery orders from the command line | [View](/skills/ordercli) |
| [PDF Tools](/skills/nano-pdf) | Edit PDF content using plain English instructions | [View](/skills/nano-pdf) |
| [System Agent](/skills/sag) | High-quality text-to-speech with expressive voices | [View](/skills/sag) |
| [Bluetooth](/skills/blucli) | Control Bluesound and NAD speakers over your network | [View](/skills/blucli) |
| [8sleep Control](/skills/eightctl) | Control your Eight Sleep pod temperature and alarms | [View](/skills/eightctl) |
| [Philips Hue](/skills/openhue) | Control lights, colors, and scenes in every room | [View](/skills/openhue) |

---

### Google Workspace Sub-Skill Pages

The Google Workspace skill (gog) also has **21 individual sub-skill pages**, one for each Google
service. These provide focused documentation for each service within the suite:

| Sub-Skill | Service | Page |
|-----------|---------|------|
| [Gmail](/skills/gog-gmail) | Email management | [View](/skills/gog-gmail) |
| [Calendar](/skills/gog-calendar) | Event scheduling | [View](/skills/gog-calendar) |
| [Drive](/skills/gog-drive) | File storage | [View](/skills/gog-drive) |
| [Docs](/skills/gog-docs) | Document editing | [View](/skills/gog-docs) |
| [Sheets](/skills/gog-sheets) | Spreadsheets | [View](/skills/gog-sheets) |
| [Slides](/skills/gog-slides) | Presentations | [View](/skills/gog-slides) |
| [Forms](/skills/gog-forms) | Form creation | [View](/skills/gog-forms) |
| [Keep](/skills/gog-keep) | Quick notes | [View](/skills/gog-keep) |
| [Tasks](/skills/gog-tasks) | Task management | [View](/skills/gog-tasks) |
| [Contacts](/skills/gog-contacts) | Contact management | [View](/skills/gog-contacts) |
| [Chat](/skills/gog-chat) | Team messaging | [View](/skills/gog-chat) |
| [Meet](/skills/gog-meet) | Video meetings | [View](/skills/gog-meet) |
| [YouTube](/skills/gog-youtube) | Video platform | [View](/skills/gog-youtube) |
| [Photos](/skills/gog-photos) | Photo storage | [View](/skills/gog-photos) |
| [Places](/skills/gog-places) | Location search | [View](/skills/gog-places) |
| [Routes](/skills/gog-routes) | Directions | [View](/skills/gog-routes) |
| [Business](/skills/gog-business) | Business profiles | [View](/skills/gog-business) |
| [Classroom](/skills/gog-classroom) | Education | [View](/skills/gog-classroom) |
| [Play](/skills/gog-play) | App store | [View](/skills/gog-play) |
| [AdSense](/skills/gog-adsense) | Ad revenue | [View](/skills/gog-adsense) |
| [Ads](/skills/gog-ads) | Advertising | [View](/skills/gog-ads) |

---

### Multi-Language Support

Skill showcase pages are currently available in **English only**. Multi-language support for skill
pages is planned for a future update.

---

### [Neotask Support](https://neotask.ai/docs/support)

## Include The Screen And Status

Open the affected product page before contacting support. Include the displayed status, the action you attempted, and the time it occurred.

![Troubleshoot settings with the system check and repair controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

This page explains how support is organized and where to start.

---

## Support Lanes

Neotask support is designed around the problem you are calling about, not around sales.

### Guest support

Use guest support if you are not fully signed up yet or if you cannot access your account at all. This lane is for:

- signup and checkout issues
- payment or billing trouble before your workspace is active
- account creation problems
- questions about whether Neotask is the right fit

### Member support

Member support is for active users, employees, and tenant-linked callers. This lane focuses on real product help:

- app setup and onboarding
- MCP app connection issues
- OAuth setup and reconnect flows
- login and workspace access
- billing and credits for active members
- teams, channels, and automation troubleshooting

### Specialized member lanes

When available, support routes deeper into the right specialist lane:

- App setup and troubleshooting
- MCP auth and OAuth setup
- Billing and account access
- Teams, channels, and automation
- Company and enterprise admin support

---

## Start Here

If you want one page that explains what the support phone system can do, start here:

- [Neotask Support Call Agent](./neotask-support-call-agent.md)

If you want more detail after that, use these deeper pages:

- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)
- [Neotask Support Agent Capability Reference](./neotask-support-agent-capability-reference.md)

---

## How Caller Recognition Works

Member-only support is intended for recognized tenant or employee callers.

In practice, the safest support flow is:

1. If your phone number matches a saved member or employee support route, support can continue in the member-only lanes.
2. If your number is not recognized, support should keep you in guest support or collect enough information for a manual follow-up.
3. If the issue is billing, sign-up, or access-before-activation, guest support can still help even when your number is not yet recognized.

If your number changed, support may need to verify your workspace another way before discussing member-only details.

### What the router should do with that recognition

1. Unrecognized callers stay in the guest lane.
2. Recognized member or employee callers can enter the deeper support lanes.
3. Ambiguous matches should stay in general support until the account context is confirmed.
4. If the caller needs a deeper answer but the number is not recognized yet, support should create a case and send a follow-up instead of guessing.

---

## What Support Can Help With

### App setup

- installation questions
- first launch and gateway startup issues
- onboarding confusion
- microphone and voice-control setup

### App connections

- managed OAuth connections
- manual OAuth app setup
- API key and credential entry
- custom MCP URL setup
- expired or broken auth sessions

### Workspace and access

- account sign-in problems
- membership confusion
- workspace access after invite or ownership changes
- employee and company support routing

### Billing

- failed checkout
- credit balance questions
- overage and auto top-up questions
- plan or invoice follow-up

### Operations

- teams and agent routing
- channels not delivering
- scheduled automations not firing
- company task setup blockers caused by missing app auth

---

## What Helps Support Resolve an Issue Fast

For an agent issue, open **Overview** and include the visible agent status and recent session. That gives support a precise starting point without requiring a full account export.

![Agent overview with recent sessions and activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

Before contacting support, gather as much of this as you can:

- the email address on the workspace
- the phone number tied to the workspace, if any
- the exact app or integration that is failing
- the status shown in the Apps tab
- the error message, if one is shown
- whether the problem is in a personal workspace, tenant, or company setup
- whether the failure happens during setup, reconnect, or live task execution

For OAuth issues, it is especially helpful to know:

- whether the provider opened a browser-based consent flow
- whether you were asked for a client ID and client secret
- whether the provider requires a custom instance URL
- whether the auth used to work and then later expired

For workspace and automation issues, support should also gather:

- which workspace or company is affected
- whether the failure is in a personal workspace, team workflow, or company automation
- whether the issue is about routing, approvals, scheduled jobs, or channel delivery
- whether the problem happens every time or only on a specific workflow

---

## Best Entry Points

- Start with [Account Access](./account-access.md) for login, membership, or billing-access issues.
- Start with [MCP Auth & OAuth Setup](./mcp-auth-oauth.md) for app connections.
- Start with [App Setup Troubleshooting](./app-setup-troubleshooting.md) for installation, gateway, channels, and automation issues.

## Support Call Agent Docs

- [Neotask Support Call Agent](./neotask-support-call-agent.md)
- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)
- [Neotask Support Agent Capability Reference](./neotask-support-agent-capability-reference.md)

---

### [Support Billing, Login, And Account Access](https://neotask.ai/docs/support-account-and-billing)

This page covers the support lane for account access, billing, and member-access questions.

---

## Login And Account Access

Use this lane when the caller is blocked from signing in or cannot reach the right workspace.

### Basic Login Recovery

1. Confirm the caller is using the correct email or sign-in method.
2. Confirm whether they are trying to access an existing member account or complete a first sign-in.
3. If the caller has multiple workspaces, confirm which one they expect to enter.
4. If the caller can sign in but cannot access the correct workspace, treat that as an account-context issue rather than a password issue.

### If Sign-In Works But Product Access Does Not

Check:

- whether the caller is on the correct account
- whether the tenant membership is active
- whether the caller expects company access that belongs to a different tenant

If the caller is recognized as a member but the workspace context still looks wrong, support should treat that as an account-context mismatch, not a generic login reset.

For plan or entitlement questions, ask the customer to open **Settings → General → License** and read the plan, agent count, status, and expiration date.

![License settings with plan and account status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/license-plan-settings-1280.webp)

---

## Billing Support

Use this lane for:

- payment failures
- subscription questions
- credits and usage questions
- scheduled task-fee questions
- blocked automations caused by task-fee balance or billing-readiness state
- BYOK questions
- renewal or access drift after payment

### Exact Billing Navigation

For callers who need to know exactly where billing lives in the app:

1. Click the avatar in the upper-right corner.
2. Click **Settings**.
3. In the left settings menu, open the **AI** section.
4. Click **Credit top-up**.

The current balance and purchase controls appear at the top of this screen.

![Credit top-up settings with balance and saved payment method](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/credit-top-up-settings-1280.webp)

### Update Payment Method

1. Open **Settings**.
2. Go to **AI** → **Credit top-up**.
3. In **Saved Payment Method**, click **Update Payment Method**.

### Add Credits Or Enable Auto Top-Up

1. Open **Settings**.
2. Go to **AI** → **Credit top-up**.
3. Choose an amount and confirm the purchase for a one-time top-up.
4. Use **Auto top-up when balance is low** for automatic replenishment.
5. For autonomous companies or billable automations, confirm the **Daily budget cap** is also set.

![Auto top-up and daily budget controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/auto-top-up-settings-1280.webp)

### Payment Failure Checklist

1. Confirm whether the failure is on signup or an existing subscription.
2. Confirm whether the card was declined or the payment page failed before submission.
3. If payment succeeded but access did not update, wait briefly for billing propagation before escalating.
4. Keep guest checkout issues separate from active-member billing disputes.

### Usage And Credits Checklist

1. Confirm whether the caller is using system keys or BYOK.
2. Confirm whether the complaint is about plan access or variable usage.
3. If the caller believes a provider was connected but billing still looks wrong, separate the billing issue from the OAuth issue.
4. If the caller says automation stopped, check whether the issue is:
   - scheduled task-fee balance exhausted
   - no-task-overage launch blocking
   - auto top-up disabled
   - daily budget missing or exhausted
5. If the caller disputes a task-fee debit, direct them to the task-fee statement and payment-allocation PDFs in billing history.

Support should also separate:

- plan access problems
- metered usage or credits questions
- scheduled task-fee questions
- BYOK configuration questions
- invoice or renewal questions

### Free Plan Checklist

If the caller is on Free, confirm:

- whether they are still in the first `10` intro messages/events or the later `5` messages/events per user-local day
- whether hosted credits or BYOK provider balance are available for the attempted message
- whether they are trying to use an allowed surface: chat, boards, coding, or app connections
- whether they are trying to use a locked surface: company operation, company tasks, user-created automations, scheduled jobs, custom agents, or overage

If a Free caller says a company was created during onboarding but will not run, explain that onboarding/preview can exist on Free, but actual company operation requires at least an Individual subscription.

---

## Guest Sign-Up Support

Guest support may help with:

- trouble paying during signup
- confusion about which plan to choose
- how to become a supported member user
- where to log in after creating an account

Guest support must not disclose member-specific billing details without recognized account context.

---

## When To Escalate

Escalate when:

- payment appears successful but account state is still wrong after reasonable delay
- a recognized member reports access drift that does not match the visible billing state
- the caller needs a tenant-specific account review

Before escalating, gather:

- whether the caller is guest or recognized member support
- whether the issue is pre-activation checkout or an existing subscription
- the workspace email and phone number on file
- what changed first: payment state, workspace access, or provider access

---

## Related Docs

- [Neotask Billing Click Paths](./neotask-support-billing-click-paths.md)
- [Neotask Billing And Credits](./neotask-support-billing-and-credits.md)
- [Billing & Plans](./billing.md)
- [Usage Rates](./usage-rates.md)
- [Security](./security.md)
- [Support Overview](./support-overview.md)

---

### [Support App Setup And Troubleshooting](https://neotask.ai/docs/support-app-setup)

This page is the first-stop runbook for desktop app setup, startup problems, and common user-visible failures.

---

## First Checks

If the app is not behaving correctly, run this sequence first:

1. Confirm you are signed in to the correct account.
2. Restart the desktop app.
3. Re-open the Apps or Integrations page and check the current status.
4. If the issue is provider-specific, reconnect only that provider instead of resetting everything.
5. If the problem started after an update, note the exact version before changing settings.

If the caller is already lost, support should reduce the issue to one of these first:

- app will not open
- app opens but the provider will not connect
- provider shows connected but tools still fail
- channel or automation behavior is wrong
- voice or microphone setup is failing

---

## Desktop App Setup

### Fresh Setup Checklist

1. Install the latest desktop app build.
2. Sign in with your account.
3. Complete onboarding for model/provider access if prompted.
4. Open the **Apps** tab for MCP apps.
5. Open the **Integrations** tab for built-in services such as Google or messaging channels.
6. Connect only the providers you actively need first.

The Apps screen shows connected services and the catalog of services that are available to add.

![Connected and available apps in the agent workspace](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

Use this split:

- **Apps** for MCP providers such as Stripe, Slack, Salesforce, HubSpot, Airtable, and Box
- **Integrations** for built-in Google Workspace services such as Gmail, Docs, Sheets, Drive, and Calendar

If the issue belongs to one company instead of the caller's general workspace:

1. Click **Auto**.
2. Open the company.
3. Use the company **Integrations** tab for company Google Workspace connections.
4. Use the company **Apps** tab for company MCP app connections.

When the requested service is missing, open the searchable app picker and add it from there.

![Searchable app picker for adding a service](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

### If Setup Stops Midway

- If the Apps tab shows `pending`, the auth flow was started but not completed.
- If the Apps tab shows `error`, the provider rejected auth or the saved credentials are no longer usable.
- If the app opens but tools are missing, the provider may not actually be connected even if an earlier flow looked successful.

---

## Symptom-First Troubleshooting

### The app opens, but the provider never finishes connecting

Check:

- whether a browser auth window was blocked
- whether the callback URL matches the one required for that provider
- whether the provider needs manual OAuth instead of managed OAuth
- whether you entered the correct `clientId` and `clientSecret`

Usually fix by:

1. Disconnect the provider.
2. Re-open the provider setup panel.
3. Re-check the callback URL and required fields.
4. Start the connection again.

If the provider clearly needs manual OAuth or manual credentials, route to MCP auth support instead of repeating generic reconnect steps.

### The provider used to work, but now shows an error

Common causes:

- expired or invalid token
- failed refresh
- wrong or rotated client secret
- provider-side app registration drift

Use the auth status guide in [Support MCP Auth And OAuth](./support-mcp-auth.md) to interpret the exact state before reconnecting.

### The provider shows connected, but tools do not behave correctly

Check:

- whether the connection was made in the right scope
- whether the provider supports the exact features you expect
- whether the provider needs additional scopes
- whether the app or workspace account you connected is the right one

Then ask:

1. Is the provider connected in the workspace the caller is actually using?
2. Does the provider need broader scopes or a different account?
3. Are the missing tools tied to a local runtime, plugin, or companion app state?

If support cannot prove the connection state from the docs and current product truth, escalate instead of inventing a hidden setting.

---

## Startup Issues

### The app will not start

1. Fully quit the desktop app.
2. Start it again.
3. If the app reaches the main window, open **Settings → Infrastructure → Connection**.
4. Check the **Local gateway** stage and use **Start** or **Restart** under **Gateway controls**.
5. Run **Settings → Infrastructure → Health**.
6. Review **Settings → Advanced → System report**.
7. Open **Settings → Advanced → Troubleshoot** and click **Check system** when repairs are needed.
8. Click **Apply repairs** when the check offers it.
9. Wait for the follow-up check and copy the technical output if it still fails.

![Gateway connection path and Gateway controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-connection-1280.webp)

The Health screen shows whether the connected Gateway is responsive and whether work is backing up.

![Runtime Health with responsiveness, workload, channels, and stability checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-health-full-1280.webp)

Use Troubleshoot only after checking the live state above. It is the screen that runs the Doctor and applies repairs.

![Troubleshoot settings used for the system check and automated repair](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

Support should also ask:

- whether the issue started after an update
- whether the user is blocked before login or after login
- whether the failure is desktop-wide or only affects one provider flow

### Voice or microphone support is failing

Check:

- operating-system microphone permissions
- the selected input device
- whether another app already owns the microphone

### Channels are connected, but behavior is wrong

Use the public channel runbooks:

- [Channels](./channels.md)
- [Troubleshooting](./troubleshooting.md)

Support should classify channel failures as one of:

- auth or connection failure
- routing or destination-targeting failure
- approvals or automation handoff failure
- delayed delivery or platform-side limitation

---

## When To Escalate

Escalate to a human or deeper support lane when:

- the caller is blocked after a reconnect attempt
- the provider needs manual OAuth setup guidance
- the auth state conflicts with what the user sees
- the issue appears tenant-specific or company-specific

If the issue is clearly about OAuth, callback URLs, auth classifications, or reconnect logic, route directly to the MCP/OAuth support lane.

---

### [Support Channel Routing And Approvals](https://neotask.ai/docs/support-channel-routing-and-approvals)

## Verify Connection Before Routing

1. Open **Channels**.
2. Run the connection tests.
3. Open access settings and confirm who can message the agent.
4. Open **Approvals** when delivery is waiting on a person.

![Passed channel connection tests](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/channel-tests-1280.webp)

Use this runbook when support needs to troubleshoot channel delivery, approval routing, requires-action follow-ups, or voice escalation behavior.

---

## Use This Page For

Use this page when the caller says:

1. The message went to the wrong place.
2. The approval reached the wrong person.
3. The escalation never fired.
4. A voice call or follow-up SMS was expected but never arrived.
5. The workflow ran, but the final delivery lane was wrong.

---

## The First Split

Before troubleshooting, classify the failure:

1. Connection failure
   The channel or provider is not connected.
2. Targeting failure
   The route fired, but the destination was the wrong room, thread, user, or phone.
3. Approval-routing failure
   The workflow paused for approval, but the wrong approver or delivery lane was used.
4. Escalation failure
   A follow-up, reminder, or voice escalation should have fired, but did not.

Do not treat these as the same problem class.

---

## Channel Delivery Runbook

For Slack, confirm the allowed channels and direct-message access after the connection test passes. A valid connection can still be restricted from the requested destination.

![Slack access and approval settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/slack-access-1280.webp)

Use this when the caller says a message did not arrive or arrived in the wrong destination.

1. Confirm the exact channel.
   Slack, Discord, Microsoft Teams, SMS, or another supported channel.
2. Confirm the exact account or binding.
   Multiple account rows can exist for the same channel family.
3. Confirm the expected target.
   Team, room, channel, thread, DM, or recipient.
4. Confirm the scope.
   Tenant routing, company routing, or agent-page routing.
5. Re-run the exact delivery after the expected target is confirmed.

### Common Failure Pattern

The transport is healthy, but the saved target belongs to the wrong team, workspace, or company route.

---

## Approval And Requires-Action Runbook

Open the waiting action and confirm its app, scope, owner, and requested operation before approving, denying, or changing policy.

![Approval detail for a waiting action](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/approval-detail-1280.webp)

Use this when the caller says an approval was stuck, went to the wrong person, or never surfaced in the expected place.

1. Confirm whether the workflow is waiting for approval or already failed for another reason.
2. Confirm who was supposed to receive the approval.
3. Confirm whether the route should use:
   a channel delivery
   an employee route
   or a voice escalation
4. Confirm the exact company or team context where the approval was created.
5. Re-test after the approver and route context are confirmed.

### Common Failure Pattern

The workflow is healthy, but the saved approval route points at the wrong recipient or the wrong company context.

### Exact Click Path For Approval Policy Changes

When the caller is not asking about a pending approval item, but about the company approval policy itself:

1. Click **Auto**.
2. Open the company.
3. Click **Approvals**.
4. Scroll to **Approval Settings**.
5. Change the overall mode or the **Per-App Rules** overrides.
6. Click **Save Approval Settings**.

---

## Escalation And Follow-Up Runbook

Use this when the caller expected a follow-up message, escalation, or support callback path.

1. Confirm the expected escalation type.
   Channel follow-up, SMS follow-up, or voice call.
2. Confirm the destination.
   Phone number, channel target, thread, or assigned employee route.
3. Confirm whether the escalation belongs to a normal tenant flow or a company flow.
4. Re-run the exact escalation path only after the destination is confirmed.
5. If the escalation depends on undocumented company behavior, escalate instead of improvising.

### Voice Escalation Checks

1. Confirm the expected phone number.
2. Confirm whether the issue is no call, wrong call context, or wrong specialist route.
3. Confirm whether the caller expected a general support route or a member-only specialist lane.

---

## Company And Employee Route Checks

Use these checks when the workflow depends on company-backed routing:

1. Confirm the company is the one that owns the route.
2. Confirm whether the caller expected employee routing or general team routing.
3. Confirm whether the route depends on company-only auth or company-only channel bindings.
4. If the company path is under-documented, escalate with the exact company, route, and failing behavior.

---

## When To Escalate

Escalate when:

1. The route target is correct, but delivery still fails.
2. The same action behaves differently across tenant and company paths.
3. The issue depends on company-specific behavior that is not fully documented yet.
4. The workflow needs engineering-only runtime details.

---

## Related Docs

- [Support Workspaces, Teams, And Automation](./support-workspaces-and-automation.md)
- [Neotask Company Approval Settings](./neotask-support-company-approval-settings.md)
- [Channels](./channels.md)
- [Automation](./automation.md)
- [Support App Setup And Troubleshooting](./support-app-setup.md)

---

### [Support MCP Auth And OAuth](https://neotask.ai/docs/support-mcp-auth)

## Start From The Provider Screen

1. Open **Apps** and select the provider.
2. Read the authentication modes shown for that provider.
3. Use OAuth when it is available.
4. Use a key or manual credential when the provider requires it.
5. Confirm the connected state before investigating tool scope.

![Searchable provider and app picker](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

This is the main support reference for MCP app authentication, OAuth setup, reconnect paths, scope placement, and auth-status interpretation.

---

## Support-Safe Rules

Support should separate three kinds of instructions:

### What the caller does in Neotask

After authorization, return to **Apps** and confirm that the service appears in the connected group. This verifies the saved app state before testing a tool.

![Connected and available apps](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

Examples:

1. Open the provider card.
2. Paste `clientId` or `clientSecret`.
3. Save a custom URL.
4. Reconnect the provider in the correct scope.

### What the caller or admin does in the provider portal

Examples:

1. Create an OAuth app.
2. Add the exact callback URL.
3. Approve scopes.
4. Create an API key or token-based auth credential.

### What only Neotask staff should touch

Examples:

1. Proxy environment variables.
2. Internal auth-truth overrides.
3. Engineering remediation notes.
4. Hidden provider fallbacks that are not part of the support-safe user flow.

Do not tell callers to edit staff-only proxy config.

---

## MCP Auth Modes

Neotask supports these main auth paths:

### Managed OAuth

The provider should complete OAuth directly from Neotask.

Typical support flow:

1. Open the provider card.
2. Click Connect.
3. Complete the browser consent flow.
4. Return to Neotask and confirm the state moves to `connected`.

### Managed DCR

The provider supports dynamic client registration, so the caller should not have to create their own app in the provider portal.

Support action:

1. Treat this as a managed flow first.
2. If it repeatedly fails, check whether the provider has a known fallback or broken-managed note.

### Manual OAuth

The caller must create or configure an OAuth app before connecting.

Typical support flow:

1. Open the provider card.
2. Copy the exact callback URL shown by Neotask.
3. Create or open the provider app in the developer portal.
4. Paste the callback URL there exactly.
5. Copy the `clientId`.
6. Copy the `clientSecret` if the provider issues one.
7. Paste those values into Neotask.
8. Start the OAuth flow and approve access.

### API Key

The provider only needs an API key or token.

Typical support flow:

1. Create or copy the API key from the provider.
2. Paste it into Neotask.
3. Save the connection.
4. Re-test the provider.

### Manual Credentials

The provider needs structured credential fields instead of a browser OAuth flow.

Typical support flow:

1. Gather every required field.
2. Confirm the right instance or account URL.
3. Enter each field exactly.
4. Save and test again.

### No Auth Or Local Runtime

Some providers do not need saved auth, or only work when a local runtime is active.

Support action:

1. Do not force these into OAuth troubleshooting.
2. Verify the runtime or plugin prerequisite first.

---

## Support Source Of Truth Order

When docs disagree, use this order:

1. The provider card and field requirements in the product.
2. Manual auth truth overrides and reviewed classifications.
3. Generated provider-auth truth and current support MCP catalog.
4. OAuth setup guides for exact callback URLs, scopes, and required fields.

If those sources still conflict, escalate instead of guessing.

---

## Scope Placement Runbook

Before reconnecting any provider, confirm where the caller is failing.

## Apps Versus Integrations

Support must distinguish these two surfaces before giving click-by-click guidance:

- **Apps**
  - MCP providers such as Stripe, Slack, Salesforce, HubSpot, Airtable, Box, Figma, and other provider cards
  - these can use OAuth, manual OAuth, API keys, or manual credentials
- **Integrations**
  - built-in Google Workspace services such as Gmail, Docs, Sheets, Drive, and Calendar

Do not route a caller to **Integrations** just because the provider uses OAuth.

Many MCP apps use OAuth inside **Apps**.

### Tenant

Use tenant scope for normal workspace usage and personal workspace tools.

### Company

Use company scope for company automations, company tasks, company channels, and company-run autonomy.

### Main Agent

Use main-agent scope for global agent-page teams and agent-specific routing.

### The Exact Support Sequence

1. Ask where the failure is happening.
   Normal workspace, company workflow, or agent/team surface.
2. Ask where the caller originally connected the provider.
   Workspace Apps tab, company app modal, or agent page.
3. Reconnect only in the failing scope.
4. Re-run the exact failing action.

If the provider is connected in the wrong scope, a green status somewhere else does not fix the real failure.

If the failure is in a company workflow and the caller is talking about Google Workspace services for that company, prefer:

1. **Auto**
2. open the company
3. **Integrations**

Do not default to the generic agent-page **Apps** tab for that case.

If the failure is in a company workflow and the caller is talking about an MCP app such as Stripe, Slack, Salesforce, or another provider card, prefer:

1. **Auto**
2. open the company
3. **Apps**
4. open the app card
5. reconnect there

For a deeper scope/run-time guide, use [Support MCP Scope And Runtime](./support-mcp-scope-and-runtime.md).

---

## Auth Classification Cheatsheet

These classifications matter most in support:

### `managed_dcr`

Start with the one-click flow. Only move to fallback if product truth says the managed path is known-bad.

### `managed_dcr_broken`

Do not loop the same broken one-click flow forever. Move to the documented fallback or escalate.

### `manual_oauth`

Confirm callback URL, client credentials, scopes, and tenant/account choice.

### `custom_url_first`

Confirm the instance or tenant URL before retrying auth.

### `manual_credentials`

Stay in the manual-credentials path. Do not switch the caller into OAuth unless product truth requires it.

### `api_key`

Check the exact API key field and whether the caller pasted the correct key for the correct account.

### `no_auth`

Check runtime prerequisites instead of discussing OAuth.

### `unreachable`

Do not improvise a setup path. Escalate.

---

## Auth Status Meanings

### `none`

The provider is not connected yet.

### `pending`

Setup started but did not finish.

Common causes:

1. The browser flow was never completed.
2. The callback did not return cleanly.
3. The provider is still waiting for final validation.

### `connected`

The saved auth is currently valid in that scope.

This does not automatically prove the provider is connected in the correct scope, with the correct account, or with the correct custom URL.

### `error`

The saved auth or credential state is unusable.

Common causes:

1. Wrong credentials.
2. Wrong callback URL or scopes.
3. Token refresh failure.
4. Wrong instance URL or tenant URL.

### `expired`

The saved token is no longer usable and needs reconnect or refresh.

Treat `expired` and `error` as reconnect-required unless provider-specific truth says otherwise.

---

## Manual OAuth Checklist

When support helps with manual OAuth, verify all of these:

1. The exact provider name.
2. The exact callback URL shown by Neotask.
3. Whether the provider requires both `clientId` and `clientSecret`.
4. The exact scopes shown in product truth.
5. The correct tenant, workspace, or provider account.
6. The correct scope in Neotask: tenant, company, or main agent.

### Why The Callback Host Is A Neotask Domain

Many OAuth providers redirect back to the Neotask MCP auth service. That is expected. Support should confirm the callback URL exactly as shown by Neotask and should not swap in an unrelated domain.

---

## High-Friction Provider Exceptions

Use these rules before giving generic reconnect advice:

### Microsoft family

Microsoft Teams, Microsoft 365, Azure, Azure DevOps, and PowerBI should be treated as Microsoft Entra manual OAuth setups unless current truth says otherwise.

### Vercel

Vercel is a managed-flow exception. If the managed path has already failed, move to fallback or escalate instead of repeating it.

### NetSuite

NetSuite is manual credentials with account-specific values, not a normal hosted OAuth flow.

### JetBrains

JetBrains is a local-runtime/plugin case. Do not force it into standard cloud OAuth troubleshooting.

### Salesforce

Salesforce requires careful Connected App setup and can still need escalation even after the documented steps are correct.

### Indeed

Indeed should be treated as a manual or pre-registered OAuth path instead of an open managed DCR path.

For the detailed versions of those flows, use [Support MCP Provider Exceptions](./support-mcp-provider-exceptions.md).

---

## Connected But Tools Still Fail

If the caller says the provider is connected but tools still fail, check this order:

1. Wrong scope.
2. Wrong workspace, tenant, or provider account.
3. Missing scopes.
4. Wrong custom URL, base URL, or instance URL.
5. Local runtime or plugin not active.
6. Auth-state versus runtime-readiness mismatch.

Do not assume a reconnect alone fixes these.

---

## What Support Should Gather Before Escalating

Collect these every time:

1. Provider name.
2. Current auth status.
3. Auth mode.
4. Failing surface.
5. Connection scope used.
6. Callback URL or custom URL if relevant.
7. Exact credential field names already filled in.
8. Whether the issue is first-time setup or reconnect after a previous success.

---

## Related Docs

- [MCP Auth & OAuth Setup](./mcp-auth-oauth.md)
- [Support MCP Scope And Runtime](./support-mcp-scope-and-runtime.md)
- [Support MCP Provider Exceptions](./support-mcp-provider-exceptions.md)
- [Support App Setup And Troubleshooting](./support-app-setup.md)
- [Support MCP Provider Reference](./support-mcp-provider-reference.md)

---

### [Neotask Support Overview](https://neotask.ai/docs/support-overview)

## Start At The Product Screen

Open the screen where the problem appears before contacting support. Gateway, Apps, Usage, and Approvals each expose the status and history support needs.

![Troubleshoot settings with the system check and repair controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

This guide explains what the support phone system is designed to help with and how support access works.

---

## What Support Covers

Neotask support is built to help with:

- desktop app setup
- app startup problems
- MCP app connections
- OAuth setup and reconnects
- API key and manual credential setup
- billing and subscription questions
- login and account-access issues
- teams, channels, automation, and enterprise admin questions
- live troubleshooting based on recent failures, sessions, runtime errors, and workflow context

The support system is not a sales line. It is optimized for troubleshooting and getting members unblocked.

---

## Member Support vs Guest Support

The support number has two support modes:

### Guest Support

Guest support is for callers who are not yet recognized as members.

Guest support can help with:

- how to sign up
- where to log in
- public setup guidance
- payment or sign-up friction before the account is active
- where to find official docs

### Member Support

Member support is for recognized callers whose phone number already matches a saved support route or employee record.

Member support can help with:

- tenant-specific troubleshooting
- deeper app and OAuth setup
- billing and account context
- workspace, automation, and company-admin questions
- recent failures, logs, session history, and runtime context
- escalations and support follow-up

That means the support line can often investigate the issue directly instead of only telling the user what the docs say.

---

## Main Support Categories

For app authentication, open **Apps** and check whether the requested service is connected before troubleshooting its tools.

![Connected and available apps](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

Recognized member callers are routed into these main categories:

1. **Setup and Troubleshooting**
2. **MCP and OAuth Integrations**
3. **Billing, Login, and Account Access**
4. **Workspace, Teams, Automation, and Enterprise Admin**

Guest callers stay in the limited guest-support lane unless a saved member route is recognized.

---

## Recommended Routing Sequence

When support answers a call, the fastest sequence is:

1. Decide whether the caller should stay in general support or move into recognized member support.
2. Identify the primary issue class: setup, MCP/OAuth, billing/access, or workspace/automation.
3. Keep the caller in one lane unless the problem clearly changes categories.
4. If the issue needs follow-up, create a support case before the call ends.

Use these routing cues:

- "I cannot finish setup" usually starts in setup support.
- "OAuth", "callback URL", "client ID", "client secret", or a provider name usually routes to MCP auth support.
- "I paid but cannot access it" usually starts in billing/account support.
- "Messages are going to the wrong place" or "automation is not firing" usually starts in workspace and automation support.

---

## What To Have Ready Before You Call

Support is fastest when you have:

- the phone number saved on your account or employee profile
- the name of the app or provider you are trying to connect
- the exact status shown in the Apps tab
- the error text, if there is one
- the callback URL or environment variables you already configured

For OAuth issues, support may ask:

- which provider you are connecting
- whether it uses managed OAuth, manual OAuth, API keys, or manual credentials
- whether the app shows `pending`, `connected`, `error`, or another state

---

## Best Support Entry Points

Use the support line when:

- you are blocked from setup
- reconnecting an app did not fix the problem
- you need step-by-step OAuth setup guidance
- you need help understanding auth status or reconnect paths
- you want the system to inspect what failed recently and explain the likely cause

Use the docs first when:

- you only need the public setup steps
- you are browsing supported apps
- you need the callback URL or environment-variable naming pattern

---

## Related Docs

- [Neotask Support Call Agent](./neotask-support-call-agent.md)
- [Neotask Authenticated Support Access](./neotask-support-authenticated-account-support.md)
- [Neotask Support Agent Capability Reference](./neotask-support-agent-capability-reference.md)
- [Support App Setup And Troubleshooting](./support-app-setup.md)
- [Support MCP Auth And OAuth](./support-mcp-auth.md)
- [Support Billing, Login, And Account Access](./support-account-and-billing.md)
- [Support Workspaces, Teams, And Automation](./support-workspaces-and-automation.md)

---

### [Support Workspaces, Teams, And Automation](https://neotask.ai/docs/support-workspaces-and-automation)

## Match The Issue To The Surface

Use **Teams** for multi-agent work, **Channels** for message delivery, and **Automations** for scheduled or event-driven runs.

![Team overview with specialists and connected apps](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/teams/2026-08-13-r2/teams-launch-overview-1280.webp)

This page covers workspace configuration, teams, channels, automation, approvals, and enterprise-admin support.

---

## Use This Lane For

- team membership and workspace questions
- channels and delivery routing
- company and automation questions
- approvals and follow-up behavior
- enterprise-admin support

---

## Workspace And Team Checks

If the caller says something is “set up but not reaching the right people,” check:

1. whether they are in the correct tenant
2. whether the correct team or company is selected
3. whether the expected channel or route is actually configured
4. whether the caller expects employee routing or general team routing

Support should also ask whether the issue is:

- wrong target
- no delivery
- delayed delivery
- approval stuck
- automation not firing at all

### Step-by-Step Workspace Runbook

1. Ask which surface is failing.
   Normal workspace, agent team, channel delivery, or company automation.
2. Ask which workspace or company the caller expected to use.
3. Confirm whether the caller is looking at the same workspace, company, or team where the route was configured.
4. Re-run the exact failing action only after the expected context is clear.

If the caller is in the wrong workspace or company, reconnecting or rerunning the task does not fix the real problem.

---

## Channel Support

When channel behavior is wrong, separate:

- transport connection problems
- routing/allowlist problems
- mention, thread, or destination-targeting problems

Start with:

1. which channel is affected
2. whether it is connected
3. whether the wrong room, thread, or recipient is being targeted

Then narrow it further:

1. Is the connection broken, or is the message going to the wrong destination?
2. Is the issue limited to one workspace, one team, or one company workflow?
3. Did the workflow ever work, or is this first-time setup?

### Step-by-Step Channel Triage

1. Confirm the affected channel and account row.
2. Confirm whether the failure is no delivery, wrong destination, or delayed delivery.
3. Confirm the saved target.
   Team, room, thread, recipient, or route-specific destination.
4. Confirm whether the caller expected tenant routing, company routing, or employee routing.
5. Re-test the exact route after the target and scope are confirmed.

Use the public docs for deeper channel-specific guidance:

- [Channels](./channels.md)
- [Troubleshooting](./troubleshooting.md)
- [Support Channel Routing And Approvals](./support-channel-routing-and-approvals.md)

---

## Automation And Company Support

The Automations screen shows the saved schedule, enabled state, and recent runs. Use those fields to separate configuration problems from failures during execution.

![Automations with schedules and recent runs](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/automations-main-1280.webp)

Use this lane for:

- tasks not running
- approvals not reaching the right person
- follow-up or escalation routing issues
- company-level setup confusion

Free-plan callers can connect apps and use chat, boards, and the coding subpage, but they cannot create or run their own automations, cron jobs, reminders, scheduled jobs, company tasks, or custom agents. The only Free automation exception is the system-managed daily `Learn from Daily Activities` memory learning job. If a Free caller expects a company task or automation to run, route the issue as an upgrade/plan-gate explanation before troubleshooting runtime behavior.

Support should confirm:

1. whether the issue is on a normal tenant workflow or an auto-company workflow
2. whether the caller expects a scheduled task, approval, escalation, or delivery
3. whether the issue is setup-related or runtime-related

Helpful symptom buckets:

- scheduled job never started
- approval reached the wrong person
- escalation did not fire
- automation ran but lacked the right app auth
- company workflow behaved differently than a normal tenant workflow

### Step-by-Step Automation Runbook

1. Confirm whether the failure is scheduling, approval routing, escalation, or final delivery.
2. Confirm whether the failing path is a normal workspace automation or a company automation.
3. Confirm whether the blocker is auth, routing, or runtime behavior.
4. Re-run the exact failing automation only after the affected path is identified.
5. If the issue depends on undocumented company-specific behavior, escalate instead of guessing.

### Step-by-Step Approvals And Escalations

1. Ask who was supposed to receive the approval or escalation.
2. Ask where it actually went, or whether it never arrived.
3. Confirm whether the route should use a channel, an employee route, or a voice escalation.
4. Confirm the exact company or team context.
5. Re-test only after the route target is confirmed.

### Exact Click Path For Company Approval Policy

When the caller is asking how to change approval behavior inside one company:

1. Click **Auto**.
2. Open the company.
3. Click **Approvals**.
4. Scroll to **Approval Settings**.
5. Use the mode cards for company-wide behavior.
6. Use **Per-App Rules** for one app or integration.
7. Click **Save Approval Settings** after making changes.

### Exact Click Path For One-App Approval Rules

1. Open the company.
2. Click **Approvals**.
3. In **Approval Settings**, switch to **Semi-Supervised**.
4. In **Per-App Rules**, search for the app name.
5. Expand the app row.
6. Set the provider override or the per-action dropdowns.
7. Click **Save Approval Settings**.

---

## Enterprise And Admin Support

This lane is appropriate when the caller needs help with:

- multiple teams or companies
- employee-backed workflows
- support escalation and handoff
- broader admin visibility questions

If the current public docs do not fully cover a company-admin feature, treat that as a known documentation boundary and escalate without inventing unsupported instructions.

Support should summarize the exact boundary before escalating:

- which workspace or company is affected
- which route, approval, or automation behavior is wrong
- whether the issue is reproducible
- whether the blocker is configuration, auth, or runtime behavior

---

## Related Docs

- [Teams](./teams.md)
- [Automation](./automation.md)
- [Auto Companies](./auto-companies.md)
- [Channels](./channels.md)
- [Support Channel Routing And Approvals](./support-channel-routing-and-approvals.md)
- [Neotask Company Approval Settings](./neotask-support-company-approval-settings.md)

---

### [Teams & Multi-Agent Orchestration](https://neotask.ai/docs/teams)

## What Are Teams?

Teams are Neotask's multi-agent orchestration system that decomposes complex goals into roles and tasks. When you create a team, an AI planning phase analyzes your objective and builds the optimal team structure automatically. A lead agent then coordinates all subagents and monitors progress through to completion.

Teams are available on the **Business and Enterprise plans**, with agent count uncapped on every plan. This is the core of what makes the higher tiers powerful: massive parallel orchestration coordinated by the platform.

## How It Works

1. **Describe your objective.** Provide a high-level goal (e.g., "Build a landing page for my startup").
2. **AI analysis.** The system analyzes complexity, identifies the required roles, and determines task breakdown.
3. **Agent assignment.** Agents are created and assigned per role (up to 5 roles per team).
4. **Task decomposition.** Your objective is broken down into discrete tasks (up to 8 tasks per project).
5. **Orchestration.** A lead orchestrator agent coordinates everything across all subagents.
6. **Real-time updates.** Status updates stream in as each task completes.

## Orchestration Modes

| Mode | Description |
| --- | --- |
| **Parallel** | All agents work simultaneously on independent tasks. Best for objectives where tasks have no dependencies on each other. |
| **Sequential** | Tasks complete in strict order. The output of each task feeds into the next. Best for objectives where each step builds on the previous one. |
| **Hybrid** | A mix of parallel and sequential execution: tasks fan out in parallel, then merge results at synchronization points. Best for complex objectives with a combination of independent and dependent work. |

## Creating a Team

To create a team in Neotask:

1. **Name and description.** Give your team a descriptive name and summary of its purpose.
2. **Select agents.** Choose which agents to include in the team.
3. **Choose orchestration mode.** Pick Parallel, Sequential, or Hybrid based on your workflow needs.
4. **Define the objective.** Write the prompt or objective that the team will work toward.
5. **Start execution.** Launch the team and monitor progress in real time.

The team table keeps the roster, connected apps, model, spend, and current state together while the run is active.

![Agent Teams table showing a launch team and its connected apps](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/teams/2026-08-13-r2/teams-launch-overview-1280.webp)

## Task Types & Model Routing

Neotask now uses frontier-only task routing. Team runs no longer route standard coding, documentation, formatting, or simple tasks to older light/medium models such as Claude Haiku, DeepSeek, GPT-4.1 Nano, or Gemini Flash Lite. Every routed task stays on a frontier model at Claude Sonnet 4.6 level or higher, GPT-5 level or higher, or the equivalent top-tier family configured for the workspace.

| Task Type | Model | Tier |
| --- | --- | --- |
| Architecture and critical decisions | Claude Opus 4.7 or GPT-5.5 Pro | Frontier |
| Complex coding and multi-file implementation | GPT-5.5, GPT-5.4, or GPT-5.3 Codex | Frontier |
| Standard coding | Claude Sonnet 4.6 or GPT-5.4 | Frontier |
| Research and analysis | Claude Opus 4.7, Claude Sonnet 4.6, or GPT-5.5 | Frontier |
| Documentation and technical writing | Claude Sonnet 4.6 or GPT-5.4 | Frontier |
| Data extraction and structured reasoning | GPT-5.4 or Claude Sonnet 4.6 | Frontier |
| Simple tasks and formatting | GPT-5+ or Claude Sonnet 4.6, using the lightest approved frontier path | Frontier |

This routing happens automatically based on the nature of each decomposed task. You do not need to manually select models, and Neotask does not degrade team work onto non-frontier fallback models.

## Budget-Aware Routing

When budget constraints are active, Neotask intelligently chooses the least expensive approved frontier model that can still handle the task:

- **Highest-risk tasks** stay on the strongest available frontier model, such as Claude Opus 4.7 or GPT-5.5 Pro.
- **Standard operational tasks** can shift to lower-cost frontier models, such as Claude Sonnet 4.6, GPT-5.4, or GPT-5.3 Codex.
- **Simple tasks** use the lightest approved frontier path, but still do not route to legacy light models.

The system also applies provider fallback ordering when a specific model is unavailable or over budget:

**Fallback chain:** Claude Opus 4.7 → Claude Sonnet 4.6 → GPT-5.5/GPT-5.4 → GPT-5.3 Codex

This ensures your team runs can continue under tight budget constraints while preserving the frontier-only quality bar.

## Session Isolation

Each agent in a team operates within its own isolated session:

- **No cross-contamination.** Agents cannot access or interfere with each other's session data.
- **Goal context sharing.** The lead agent shares the relevant goal context with each subagent, but nothing more.
- **Completion reporting.** Real-time completion reports flow back from each subagent to the lead agent, keeping orchestration on track.

This isolation ensures that agents work independently and reliably without side effects from other agents in the same team.

## Team Runs

Neotask tracks every team execution as a run. From the team runs view, you can:

- **View run history.** See past runs with their status, duration, and results.
- **Start and stop runs.** Launch new executions or halt runs in progress.
- **Monitor individual agent progress.** Track what each agent is working on and how far along it is.
- **Task cascade.** Completion of one task automatically triggers any dependent tasks, keeping the workflow moving without manual intervention.

Open the active run to read the team conversation, app activity, and work returned by each agent.

![Active agent team run showing agents working together](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/teams/2026-08-13-r2/teams-launch-active-1280.webp)

## Scaling with Enterprise

Enterprise plans are designed for teams that need to run complex, multi-agent workflows at scale:

- **Uncapped concurrent agents.** Deploy as many agents as your workflows need, working in parallel across different teams and objectives.
- **Usage-based pricing.** You pay the platform subscription for the orchestration infrastructure, and usage fees are billed based on actual token consumption across all agents. This means you only pay for the compute your agents actually use.
- **Per-agent budgets.** Set individual spend limits on every agent to maintain cost control even at scale.
- **Smart routing.** Automatic model selection ensures each task uses the most cost-effective model capable of handling it.

---

### [Troubleshooting](https://neotask.ai/docs/troubleshooting)

This guide covers common issues you may encounter while using Neotask and how to resolve them.

## Check Gateway Health Before Applying Repairs

Neotask has separate screens for connection status, runtime health, the combined system report, and automated repairs. Use them in this order.

### 1. Check The Connection Path

1. Open **Settings → Infrastructure → Connection**.
2. Confirm the path from **Desktop app** to **Local gateway**, **Neotask cloud**, and **Connected services**.
3. Read the first stage that is not ready.
4. Use **Start** or **Restart** under **Gateway controls** when the local Gateway is stopped or stale.
5. Refresh the page and confirm that every stage is ready.

![Gateway connection path and Gateway controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-connection-1280.webp)

### 2. Run The Runtime Health Check

1. Open **Settings → Infrastructure → Health**.
2. Click **Run check**.
3. Review event-loop delay, active work, queued work, channels observed, and recent stability events.
4. A high queue, delayed runtime, or dropped diagnostic events can explain slow or stalled work even when the Gateway is connected.

![Runtime Health with responsiveness, workload, channels, and stability checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-health-full-1280.webp)

### 3. Review The Combined System Report

1. Open **Settings → Advanced → System report**.
2. Review **Local gateway**, **Session access**, **Runtime response**, **Connected services**, **Current work**, and **Scheduled checks**.
3. Start with the first item that needs attention.
4. Copy the sanitized report when support needs the full result.

![System report with live Gateway and service checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-system-report-1280.webp)

### 4. Run The Doctor And Apply Repairs

1. Open **Settings → Advanced → Troubleshoot**.
2. Click **Check system**. This check does not change your installation.
3. Select **Show technical output** to read the detailed findings.
4. Click **Apply repairs** when the check makes it available.
5. Keep Settings open while Neotask applies repairs and runs the follow-up check.

![Troubleshoot settings with Check system and Apply repairs](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

Retry the original action after the follow-up check completes. If it still fails, copy the technical output and include it when you contact support.

### Additional Checks

- Open **Settings → Infrastructure → External Dependencies** when browser, document, media, local-search, or coding tools are missing. Install only the component the failed work requires.
- Open **Settings → Advanced → Logs** to search the redacted local Gateway stream by message, subsystem, or level. Copy the visible entries when support asks for runtime events.

![External Dependencies with installed optional components](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/external-dependencies-1280.webp)

The Logs screen is local and redacted. Use it when the health screens identify a runtime or service problem that needs event detail.

![Redacted local Gateway Logs](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-logs-1280.webp)

---

## License & Activation

### "Invalid license key"

- Verify the key format: `NT-XXXXXXXX-XXXXXXXX-XXXXXXXX-XXXXXXXX`
- Check for extra spaces or missing characters
- Ensure you have an active internet connection
- Try re-entering the key carefully

### "License revoked"

- Your license has been revoked by the server
- Check your billing status at [neotask.ai](https://neotask.ai)
- Contact support if you believe this is an error

### "License expired"

- Your subscription has lapsed
- Renew your plan through the billing portal
- The app enters read-only mode when expired

### "Offline grace period expired"

- You've been offline for more than 72 hours
- Connect to the internet for automatic revalidation
- The app will resume normal operation once connected

### "Version blocked"

- Your app version has been blocked for security reasons
- Update to the latest version immediately
- Auto-update should prompt you; if not, download from the website

---

## Gateway Issues

### "Gateway failed to start"

1. Fully quit Neotask and open it again.
2. Open **Settings → Infrastructure → Connection** and check the **Local gateway** stage.
3. Use **Start** or **Restart** under **Gateway controls**.
4. Open **Settings → Advanced → System report** and review the live system checks.
5. If the Gateway still fails, run **Check system** from **Settings → Advanced → Troubleshoot**.
6. Review the first failed item in the technical output.
7. Click **Apply repairs** when it is available.
8. If the follow-up check still fails, copy the technical output and send it to support.

### "3-strike lockout"

- After 3 consecutive gateway failures, operations are blocked
- Restart the app to reset the lockout counter
- If it persists, check logs for the underlying error

### "Gateway disconnected"

- The local runtime lost connection
- The app now **detects a stalled or unresponsive gateway automatically** (a connection liveness check) and shows a clear "reconnecting" state, then re-establishes the connection on its own, it no longer hangs indefinitely on a frozen connection
- If a request was in flight when the connection stalled, it surfaces a clear "gateway unavailable, retrying" message instead of spinning forever (in-progress AI responses and pending approvals are never interrupted by this)
- Refresh **Settings → Infrastructure → Connection** and confirm which stage disconnected
- Run **Settings → Infrastructure → Health** to check runtime delay, queued work, channels, and stability events
- Review **Settings → Advanced → System report** for session access, connected services, current work, and scheduled checks
- Use **Settings → Advanced → Troubleshoot** when the earlier checks identify a problem that needs repair
- Apply the offered repairs and wait for the follow-up check
- Copy the technical output for support if the disconnect continues

---

## Voice Issues

### "No audio input detected"

- Grant microphone permission:
  - **macOS**: System Preferences → Privacy & Security → Microphone → Enable for Neotask
  - **Windows**: Settings → Privacy → Microphone → Allow apps to access
- Check that your microphone is not muted
- Verify the correct input device is selected in system settings

### "Wake word not triggering"

1. Ensure voice activation is enabled in Settings
2. Check that wake mode is set to "Porcupine" (not "Shortcut" or "Off")
3. Say the wake phrase clearly: "Hey Neotask"
4. Try the keyboard shortcut instead (Cmd+Shift+Space / Ctrl+Shift+Space)
5. Check if another app is using the microphone

### "TTS audio choppy or delayed"

- Check your internet connection (TTS requires ElevenLabs API)
- ElevenLabs may be experiencing rate limiting
- Try a different voice from Settings
- Reduce background network activity

### "Voice shortcut not working"

- Another app may have claimed the same shortcut
- Change the shortcut in Settings → Activation → Keyboard Shortcut
- On macOS, check System Preferences → Keyboard → Shortcuts for conflicts

---

## Connection & Network

### "Cannot connect to server"

- Verify your internet connection
- Check if [neotask.ai](https://neotask.ai) is accessible in your browser
- If behind a corporate firewall, ensure HTTPS traffic is allowed
- Try disabling VPN temporarily

### "API request failed"

- Check your internet connection
- Verify your license is active
- If using BYOK, check that your API key is valid and has credits
- Try refreshing the dashboard

---

## Agents & Sessions

### "Cannot create agent: limit reached"

- Agent count is not capped by plan; custom agent creation requires a paid plan
- If you see this on the Free plan, upgrade to a paid plan to create custom agents
- Delete unused agents to keep your workspace tidy

### "Agent not responding"

- Check gateway health (must be running)
- Verify the agent's model is accessible (API key valid, provider available)
- Check daily budget, because if it is exceeded, agents are paused until midnight UTC
- Try resetting the session

### "Session messages not loading"

- Refresh the page/app
- Check gateway connection status
- If the session is very long, try creating a new session

### "Interrupted by restart" or "Resumed after restart" in the activity feed

These two entries are normal. **Interrupted by restart** marks work that stopped
because of an app update, an operating-system restart, a crash, or a power loss.
**Resumed after restart** marks the same work starting again.

If you see **Interrupted by restart** with no matching **Resumed after restart**:

1. Open **Settings**, select **Automations**, and confirm **Continue work after restarts** is on.
2. Open the company, select **Settings**, and confirm that company is not overriding the setting to off.
3. Remember that work waiting for your approval is recorded but never resumed automatically, and neither are runs that failed for a reason other than the restart.

See [Continue Work After Restarts](continue-work-after-restarts.md).

### "Send a message to continue"

A thread that cannot be continued automatically, for example when the assistant
session has expired, shows **Send a message to continue** instead of resuming on
its own. Send a message in that thread to continue working. Your
original message is never sent twice.

---

## Billing & Payments

### "Payment failed"

- Verify your card details are correct
- Check with your bank for any blocks on international transactions
- Try a different payment method
- Contact support if the issue persists

### "Usage allocation exceeded"

- Your agent's token consumption has exceeded the allocation included in your subscription
- Additional usage is billed at usage rates. Check the Usage page in your dashboard for details.
- Switch to BYOK mode (use your own API keys) to route token costs directly to your provider
- Set daily budget limits to prevent unexpected overages

### "Plan not updated after payment"

- Stripe webhooks may have a slight delay (usually under 1 minute)
- Refresh the dashboard
- If still not updated after 5 minutes, contact support with your payment confirmation

---

## Performance

### "App slow on startup"

- License validation and gateway startup happen on launch, and this is normal
- Ensure your system meets minimum requirements
- Close unnecessary background apps

### "High CPU usage"

- If using always-on wake word (Porcupine), it uses minimal CPU but can be disabled
- Pause voice sessions when not in use
- Check for stuck gateway processes in Activity Monitor / Task Manager
- Recent versions reduced sustained gateway CPU during heavy activity, large session state is serialized more efficiently (compact, kept off the critical path) and background/scheduled work automatically backs off while the gateway is busy, so CPU spikes are less frequent and the app stays responsive under load

### "High memory usage"

- Long chat sessions accumulate memory
- Create new sessions periodically
- Restart the app to clear accumulated state

---

## Getting Help

- **Documentation**: You're here. Browse the knowledge base.
- **Support**: Visit the Support page at [neotask.ai/support](https://neotask.ai/support)
- **Contact**: Use the contact form on the website
- **Status**: Check service status at [neotask.ai](https://neotask.ai)

---

### [Trust & Security](https://neotask.ai/docs/trust-and-security)

## Product Controls

Use **Settings → Privacy and Data** for data choices, **Settings → Permissions** for Safe Mode, and **Approvals** for actions waiting on review.

![Session data sharing control, off by default](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-consent-1280.webp)

How Neotask protects your data. This page summarizes our security posture for customers and prospects. For our formal compliance status and how to request audit reports, see [Compliance Overview](compliance-overview). For data-protection rights, see [Data Privacy & Your Rights](data-privacy-and-your-rights).

## Our security principles

- **Your data is kept secure, durable, and consistent.** Your data is stored in a secure, managed cloud environment with automated, regularly-tested backups; any copy cached on your device for offline use is encrypted and kept in sync, so you always see accurate, up-to-date information.
- **Encrypt everywhere.** Sensitive data is encrypted in transit (TLS) and at rest. Credentials, tokens, and provider keys are encrypted at the field level before storage.
- **Least privilege + tenant isolation.** Every request is authenticated and scoped to your account from a verified token, never from client-supplied identifiers. One tenant can never see another's data.
- **Private by default.** Telemetry and error monitoring run with personal data collection disabled and redaction on by default.

## Data protection

| Area | What we do |
|---|---|
| Encryption in transit | All API and gateway traffic uses TLS. Plaintext transport is refused. |
| Encryption at rest | AES-256-GCM field-level encryption for credentials, OAuth tokens, provider keys, and other sensitive records, with versioned keys and backward-compatible key rotation. |
| Access control | Token-based authentication, role-based access within an account, and HMAC-signed requests (nonce + timestamp) for device and service calls. |
| Secrets management | No hardcoded secrets; production secrets are required at startup (the service fails closed if a required secret is missing) and are never logged. |
| Audit logging | Privileged and security-relevant actions are recorded to an append-only, tamper-evident audit log. |
| Monitoring | Error and availability monitoring with health checks and alerting; personal data is excluded from telemetry by default. |

## Infrastructure & hosting

- Hosted on established cloud providers with their own SOC 2 / ISO programs (database, application platform, object storage).
- Network hardening on the database tier (restricted access, encrypted connections).
- Backups with a tested restore procedure and a documented disaster-recovery plan.

## Secure development

Safe Mode requires an explicit confirmation before broader machine access is enabled.

![Safe Mode confirmation](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/safe-mode-confirmation-832.webp)

- Code review and change-management controls before changes reach production.
- Dependency and vulnerability management.
- A documented Secure SDLC and incident-response plan, with a defined breach-handling process.

## AI processing

Neotask routes AI requests through our gateway to model providers (see the [Subprocessors](subprocessors) list). We minimize what is sent, do not use your private content to train our own models, and document the providers that may process prompt content so you can make an informed choice.

## Reporting a security issue

If you believe you've found a vulnerability, contact **security@neotask.ai**. Please do not publicly disclose it until we've had a chance to investigate and remediate.

---

### [Usage Rate Breakdown](https://neotask.ai/docs/usage-rates)

## Read A Usage Record

1. Open **Usage**.
2. Choose the time range.
3. Review cost by provider, model, agent, and company.
4. Open a session when you need the request and tool history behind a charge.

![Usage overview with provider and model activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

This page details exactly how Neotask calculates usage-based charges. All users start with
$10 in one-time signup credits that are consumed by usage. Understanding these rates helps you
estimate costs and manage your credit balance effectively.

For an overview of plans and how billing works, see [Billing & Plans](billing.md).

---

## How Usage Is Calculated

Every time your agent processes a message, it consumes AI tokens (input and output). The cost
of those tokens depends on which AI model your agent uses. The system tracks four types of
tokens separately, each with its own rate:

- **Input tokens:** the prompt and context sent to the model
- **Output tokens:** the response the model generates
- **Cache read tokens:** tokens served from prompt cache (cheaper than input)
- **Cache write tokens:** tokens written into prompt cache for future reuse

### Token Cost Formula

```
Cost = (input tokens / 1,000,000) x input rate
     + (output tokens / 1,000,000) x output rate
     + (cache read tokens / 1,000,000) x cache read rate
     + (cache write tokens / 1,000,000) x cache write rate
```

Source: `server/src/config/modelPricing.ts`

---

## AI Model Token Rates

These are the per-token rates for each supported AI model. All rates are in **USD per 1 million tokens**.

### Anthropic Claude

| Model | Input | Output | Cache Read | Cache Write |
|-------|------:|-------:|-----------:|------------:|
| Claude Opus 4.6 | $15.00 | $75.00 | $1.50 | $18.75 |
| Claude Sonnet 4.5 | $3.00 | $15.00 | $0.30 | $3.75 |
| Claude Haiku 4.5 | $0.80 | $4.00 | $0.08 | $1.00 |
| Claude 3.5 Sonnet | $3.00 | $15.00 | $0.30 | $3.75 |
| Claude 3.5 Haiku | $0.80 | $4.00 | $0.08 | $1.00 |
| Claude 3 Opus | $15.00 | $75.00 | $1.50 | $18.75 |
| Claude 3 Sonnet | $3.00 | $15.00 | $0.30 | $3.75 |
| Claude 3 Haiku | $0.25 | $1.25 | $0.03 | $0.30 |

### OpenAI

| Model | Input | Output | Cache Read | Cache Write |
|-------|------:|-------:|-----------:|------------:|
| GPT-4o | $2.50 | $10.00 | $1.25 | $2.50 |
| GPT-4o Mini | $0.15 | $0.60 | $0.075 | $0.15 |
| GPT-4 Turbo | $10.00 | $30.00 | $5.00 | $10.00 |
| GPT-4 | $30.00 | $60.00 | $15.00 | $30.00 |
| o1 | $15.00 | $60.00 | $7.50 | $15.00 |
| o1-mini | $3.00 | $12.00 | $1.50 | $3.00 |
| o3-mini | $1.10 | $4.40 | $0.55 | $1.10 |

### Google Gemini

| Model | Input | Output | Cache Read | Cache Write |
|-------|------:|-------:|-----------:|------------:|
| Gemini 2.0 Flash | $0.10 | $0.40 | $0.025 | $0.10 |
| Gemini 1.5 Pro | $1.25 | $5.00 | $0.3125 | $1.25 |
| Gemini 1.5 Flash | $0.075 | $0.30 | $0.01875 | $0.075 |

Source: `server/src/config/modelPricing.ts` lines 8-42

---

## Platform Fees & Markups

There are several distinct fees that may apply depending on how your agent is being used. These
are **not** a single flat rate; each applies in different situations.

### 1. Credit Mode Markup (System Key): 20%

When using Neotask's managed API keys (System Key mode), a **20% markup** is applied on top
of the raw token cost. This covers API key management, automatic failover between providers,
model routing, and infrastructure.

If you use **BYOK mode** (Bring Your Own Keys), the base token cost goes directly to your AI
provider. However, the **20% platform fee still applies** and is deducted from your Neotask
credit balance. BYOK saves you the base API cost (since you pay your provider directly), but
the platform fee covers infrastructure, routing, failover, and orchestration.

Source: `server/src/config/modelPricing.ts` line 53, `CREDIT_MODE_MARKUP_PCT = 0.20`

### 2. Overage Platform Fee (Varies)

When your usage exceeds your included credit pool, the remaining cost becomes **overage**. A
platform fee is applied on top of the overage amount. The rate depends on your plan tier:

| Tier | Overage Fee |
|------|------------:|
| Individual | 25% |
| Business | 15% |
| Enterprise | 10% |

The overage fee is only charged on the amount that exceeds your credit pool. Usage within
your pool has no additional fee.

Overage is automatically charged to your card when the unsettled amount reaches **$10**
(Stripe minimum charge: $0.50).

Source: `server/src/config/planConfig.ts`, `overageFeePct` and `overageChargeThreshold`

### 3. Automation Markup: Progressive 50% Down to 33%

When your agents run **automated jobs** (scheduled tasks, cron jobs, recurring automations),
an additional automation markup is applied on top of the token cost. This is because automated
agents replace manual work. They run unattended, on schedule, handling tasks that would
otherwise require employees.

The automation markup uses **progressive brackets** (like income tax). You pay a higher rate
on your first dollars of automated usage and a lower rate as your automated spending grows.
There are no cliff effects; each dollar is charged at its bracket rate only.

| Cumulative Automated Spend (per billing cycle) | Marginal Markup Rate |
|------------------------------------------------|--------------------:|
| $0 - $10 | 50% |
| $10.01 - $25 | 45% |
| $25.01 - $50 | 40% |
| $50.01 - $100 | 37% |
| $100.01+ | 33% |

Source: `server/src/config/automationMarkup.ts` lines 38-44, `DEFAULT_AUTOMATION_BRACKETS`

### 4. Coding Task Markup: Flat 50%

Automated jobs that involve coding tools (file reads, writes, code execution, bash commands)
are charged at a **flat 50% markup**. The progressive reduction does **not** apply to coding
tasks; it is always 50% regardless of cumulative spend.

Coding tools: `exec`, `read`, `write`, `bash`, `code`

Source: `server/src/config/automationMarkup.ts` line 49, `CODING_TASK_MARKUP_RATE = 0.50`

---

## How Costs Stack

These fees can stack depending on the situation:

**Interactive chat (System Key mode):**
- Raw token cost + 20% credit mode markup
- No automation markup (not automated)

**Interactive chat (BYOK mode):**
- Raw token cost paid to your provider + 20% platform fee deducted from credits

**Automated cron job (System Key mode):**
- Raw token cost + 20% credit mode markup + automation markup (50%-33%)

**Automated coding task (System Key mode):**
- Raw token cost + 20% credit mode markup + 50% coding task markup

---

## Automation Billing Example

A scheduled cron job runs daily. The AI agent takes about 2-3 hours to complete its work.
The job consumes **$10 in raw tokens** (the actual cost to the AI provider).

**Breakdown:**

- **Token cost:** $10 (passed through to the AI provider; this is what the compute costs)
- **Automation markup:** $10 x 50% = $5 (platform fee for running the job unattended)
- **Total charge:** $15

The $10 covers the actual AI compute your agent consumed. The $5 is the automation fee. The
platform ran the job on schedule, monitored execution, handled retries, and delivered results,
all without anyone having to be at a computer.

**Why automation costs more:** Automated agents replace human labor. A cron job that checks
your analytics every morning, drafts reports, monitors inventory, or processes incoming orders
is doing work that would otherwise require a person's time. The automation markup reflects the
value of that unattended execution, and it **decreases the more you automate**, rewarding
scale. A tenant spending $100+/cycle on automated jobs pays only 33% markup instead of 50%.

---

## Top-Up Credits

You can purchase additional credits at any time to prepay for usage:

- **Minimum top-up:** $5
- **Maximum top-up:** $10,000
- **Credits never expire.** They remain in your account until used.
- **Auto top-up:** Optionally configure automatic top-ups when your balance drops below a
  threshold (default: $2 balance triggers a $10 top-up)

Top-up credits are added to the same balance as your signup credits. All credits draw from a
single balance. When your balance hits $0, your agent is paused unless you top up or enable
overage billing.

Source: `server/src/services/overageCharger.ts`, `server/src/services/balanceService.ts`

---

## Credit Deduction Order

When your agent completes a task, the cost is deducted in this order:

1. **Credit balance** (your $10 one-time signup grant plus any top-ups, all in a single balance that does not reset)
2. **Overage** (if enabled, charged to your card when the unsettled amount reaches $10)

Your credit balance is consumed first. When it reaches $0, your agent is paused unless overage
is enabled. With overage enabled, usage continues and is charged to your card automatically.

Source: `server/src/services/balanceService.ts` lines 240-286

---

## Budget Controls

Enterprise users have built-in budget enforcement:

- **Global daily budget:** Default $500/day (configurable). When exceeded, the agent gateway
  automatically shuts down to prevent runaway costs.
- **Per-agent spend limits:** Set a daily cap for each individual agent. When an agent hits
  its limit, its sessions are automatically paused.
- **Auto-shutoff:** Configurable. Enable or disable automatic gateway shutdown on budget exceed.
- **Real-time enforcement:** Budget is checked every 30 seconds.

Source: `Neotask-Electron/src/main/budgetEnforcer.ts`, `Neotask-Electron/src/main/agentSpendLimitStore.ts`

---

## Viewing Your Usage

The period chart shows how cost changes across the selected date range. Use it to locate a spike before opening the provider, agent, or session breakdown beneath it.

![Usage chart across a selected period](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/ledger-chart-1280.webp)

Track all of this in real time from the Usage page in the desktop app:

- **Token-level breakdown** by model, provider, agent, channel, and date
- **Cost breakdown:** input, output, cache read, cache write costs shown separately
- **Time granularity:** today, last 24h, 7 days, 30 days, 365 days, or custom range
- **Per-agent attribution:** see exactly which agent is consuming what
- **Per-channel attribution:** cost breakdown by messaging channel
- **Per-provider attribution:** cost by AI provider (Anthropic, OpenAI, Google)
- **CSV/JSON export** for external analysis

See [Billing & Plans](billing.md) for more on managing your subscription.

---

### [Voice Control](https://neotask.ai/docs/voice-control)

## Start Voice Control

1. Open **Settings → Voice**.
2. Choose the input device and confirm microphone access.
3. Start voice mode from the agent chat.
4. Review any action that pauses for approval.

![Wake word and microphone status ready for voice control](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/microphone-ready-1280.webp)

Neotask includes a full-featured voice control system that lets you interact with your AI agents entirely through speech. You can activate agents, issue commands, navigate the interface, and receive spoken responses, all hands-free.

---

## Overview

### Activation Methods

There are two ways to activate voice input:

1. **Always-listening wake word.** Speak a trigger phrase (e.g., "Hey Neotask") and the application begins listening. No keys to press.
2. **Keyboard shortcut.** Press a key combination to start voice input on demand.

You can choose your preferred method in **Settings > Wake Mode**.

### Voice Interaction Flow

Every voice interaction follows this cycle:

1. **Wake.** Activation via wake word or keyboard shortcut.
2. **Listen.** Neotask listens to your spoken input.
3. **Transcribe.** Speech is converted to text in real time.
4. **Think.** The AI processes your request and determines the appropriate actions.
5. **Speak.** The response is spoken back to you using natural text-to-speech.
6. **Listen.** The system returns to listening for your next command, keeping the conversation flowing.

---

## Wake Word Activation

### Default Wake Word

The default wake word is:

> **"Hey Neotask"**

Simply say this phrase, and Neotask will begin listening for your command.

### Custom Wake Words

You can set a custom wake word in **Settings > Wake Word**. Choose any short, distinct phrase that is easy for you to say and unlikely to occur in normal conversation.

### Performance

Wake word detection runs **entirely on your local machine**; no audio is sent to the cloud for wake word processing. The detection engine is optimized for **ultra-low CPU usage**, so it can remain active in the background without impacting system performance.

### Sensitivity

Wake word sensitivity is configurable. If you find the wake word triggers too easily (false positives) or not often enough (missed activations), adjust the sensitivity slider in **Settings > Wake Word > Sensitivity**.

---

## Keyboard Shortcut Activation

### Default Shortcuts

| Platform | Shortcut |
|---|---|
| macOS | `Cmd + Shift + Space` |
| Windows / Linux | `Ctrl + Shift + Space` |

### Customization

The keyboard shortcut is fully customizable. Go to **Settings > Wake Mode > Keyboard Shortcut** to set your preferred key combination.

---

## Voice Features

### Speech-to-Text

Neotask uses **Deepgram** for real-time speech-to-text transcription. Your spoken words appear as text in the conversation as you speak, with minimal latency.

### Text-to-Speech

Responses are spoken aloud using **ElevenLabs** natural text-to-speech technology. The voice library includes **100+ voices** spanning a wide range of styles.

### Voice Selection

Choose your preferred voice in **Settings > Voice**. You can filter voices by:

- **Gender:** Male, female, or neutral.
- **Accent:** American, British, Australian, and many more.
- **Age:** Young, middle-aged, or mature.

A **voice preview** button is available next to each voice so you can hear a sample before selecting it.

### Conversation Controls

- **Pause.** Pause the voice conversation at any time. The AI will stop listening and speaking until you resume.
- **Resume.** Continue the conversation from where you left off.

### File Attachments

You can attach files during a voice session. For example, say "I want to share a file" and use the attachment dialog, or drag and drop a file into the conversation window while voice mode is active. The AI can then reference and work with the attached file.

---

## Voice Commands

Neotask understands a wide range of natural language commands. Below are common categories with examples.

### Open Websites

| Example Command |
|---|
| "Open YouTube" |
| "Go to github.com" |
| "Open the Neotask documentation" |

### Search the Web

| Example Command |
|---|
| "Search for Python tutorials on Google" |
| "Look up the weather in San Francisco" |
| "Search Stack Overflow for React hooks" |

### Launch Applications

| Example Command |
|---|
| "Open Safari" |
| "Launch Finder" |
| "Open Visual Studio Code" |
| "Start Terminal" |

### Browser Control

| Example Command |
|---|
| "Scroll down" |
| "Go back" |
| "Refresh the page" |
| "Scroll to the top" |

### Agent Operations

| Example Command |
|---|
| "Create an agent called Research Assistant" |
| "Start the agent" |
| "Stop the agent" |
| "Show me agent status" |

### Multi-Command Chains

You can combine multiple instructions in a single spoken command:

| Example Command |
|---|
| "Create an agent called Data Analyzer, enable voice, and start it" |
| "Open YouTube and search for machine learning tutorials" |
| "Stop the agent and show me the session log" |

---

## Tool Execution During Voice

When your voice command triggers a tool or action, Neotask provides **real-time spoken feedback** so you know what is happening:

The briefing screen keeps the current action, supporting context, and next steps visible during the voice session.

![Voice briefing with governed next actions](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/voice-briefing-1280.webp)

- "I'm opening the file editor..."
- "Running the shell command now..."
- "Fetching the web page..."

### Supported Tool Actions

Tools that can be triggered by voice include:

- **Shell commands.** Execute terminal commands on your machine.
- **File operations.** Create, read, edit, and organize files.
- **Web requests.** Fetch data from URLs and APIs.

### Approval Workflow

When **Safe Mode** is enabled (on by default), sensitive actions require your explicit spoken or clicked approval before execution. Sensitive actions include:

- Deleting files or directories
- Deploying code or services
- Sending messages or emails on your behalf

The AI will describe the action and ask for confirmation before proceeding.

---

## Math Tutoring Mode

Neotask includes a specialized math tutoring mode that combines voice instruction with animated visualizations.

### How It Works

1. Ask for a math topic, for example, "Teach me about the unit circle."
2. The AI generates a **lesson plan** tailored to the topic.
3. Animated visualizations are rendered using **Manim** (the mathematical animation engine).
4. The lesson is delivered **section by section**, with spoken explanations synchronized to the visuals.

### Visualization Templates

The following built-in templates are available for instant animated lessons:

| Template | Description |
|---|---|
| **Unit Circle** | Visual walkthrough of the unit circle with angle and coordinate labels. |
| **Pythagorean Theorem** | Geometric proof animation with labeled squares on triangle sides. |
| **Taylor Series** | Step-by-step expansion showing polynomial approximation convergence. |
| **Quadratic Formula** | Derivation and graphical interpretation of roots. |
| **Sine / Cosine Waves** | Animated wave plots with amplitude, period, and phase annotations. |
| **Derivatives** | Tangent line animation illustrating instantaneous rate of change. |
| **Integrals** | Area-under-the-curve animation with Riemann sum progression. |
| **Graph Functions** | Plot any function with labeled axes, intercepts, and key features. |

### Progressive Teaching

Lessons are broken into sections. After each section, the AI pauses and asks if you are ready to continue, want to review, or have questions. This ensures you learn at your own pace.

---

## Supported Languages

Neotask supports voice interaction in **21 languages**:

| Language | Code |
|---|---|
| English | `en` |
| Spanish | `es` |
| French | `fr` |
| German | `de` |
| Italian | `it` |
| Portuguese | `pt` |
| Dutch | `nl` |
| Russian | `ru` |
| Chinese (Mandarin) | `zh` |
| Japanese | `ja` |
| Korean | `ko` |
| Arabic | `ar` |
| Hindi | `hi` |
| Turkish | `tr` |
| Polish | `pl` |
| Swedish | `sv` |
| Danish | `da` |
| Norwegian | `no` |
| Finnish | `fi` |
| Czech | `cs` |
| Romanian | `ro` |

You can change the voice language at any time in **Settings > Language**. Both speech recognition and text-to-speech will switch to the selected language.

---

## Voice Prompts

Voice prompts control how the AI assistant behaves and responds during voice conversations. Both are editable in **Settings > Voice Prompts**.

### System Prompt

The **system prompt** defines the overall personality and behavior of the voice assistant. It sets the tone, expertise level, and interaction style. For example, you can instruct the assistant to be concise and technical, or friendly and conversational.

### Response Prompt

The **response prompt** customizes how the assistant formats and delivers its spoken responses. Use this to control response length, level of detail, whether the assistant uses analogies, and other stylistic preferences.

Both prompts accept free-form text and take effect immediately for all subsequent voice interactions.

---

### [Activity History](https://neotask.ai/docs/neotask-desktop/activity-history)

Activity History helps Neotask understand recent work you do by hand on your Mac. With your permission, it turns supported app and website activity into a private timeline. You can ask an agent where you left off, prepare a standup, find recent work, or draft a standard operating procedure (SOP) from a repeated process.

Activity History observes your work. It never moves the mouse, types, or takes control of your computer. It is off by default, starts only after you turn it on, and does not create history for work completed before then.

## Availability

Activity History is available on Apple Silicon Macs in the current release. Intel Macs and the Windows and Linux desktop apps do not record computer activity yet. Activity History is unavailable in HIPAA deployments.

Before you can turn it on:

- **Memories must be enabled.** Memories make Activity History summaries available across chats and tasks.
- **Your organization must allow it.** On a company account, an owner or admin chooses which roles and members may opt in. Access never starts recording and never grants consent for you.
- **You must consent on this Mac.** Each Mac has its own consent and recording controls.
- **Global Privacy Control and Do Not Track must be off.** Either privacy signal blocks Activity History.
- **Neotask needs macOS Accessibility access.** This permission lets Neotask read supported app, window, and control information. Activity History does not request Screen Recording.

## Turn It On

1. Open the Neotask desktop app and select **Settings**.
2. Select **Activity History** in the General section.
3. Review the **Permissions** tab. Keep the default to record supported apps and websites, then add private apps and sites to the skip lists. For a smaller scope, choose selected apps and websites.
4. Read and accept both consent statements. One statement covers data processing. The other asks you to pause recording during calls, meetings, chats, and other communications with people.
5. Grant macOS Accessibility access when prompted, then return to Neotask and finish setup.
6. Turn on **Remember my computer activity** and save your changes. The status changes to **Recording** when every requirement is satisfied.

Recording is not retroactive. Nothing from before setup is added to your timeline.

## What Neotask Can Record

Activity History saves a small set of signals instead of a screen recording.

| Signal                       | What can appear in history                                                                                        | What is left out                                                                    |
| ---------------------------- | ----------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------- |
| Active app                   | The foreground app name                                                                                           | Activity from excluded apps, hidden system apps, and Neotask helper processes       |
| Window                       | A supported visible window title                                                                                  | Window pixels and screenshots                                                       |
| Browser page                 | The website domain and page path, such as `docs.example.com/projects/launch`                                      | URL query strings, page fragments, usernames, passwords, and private-browsing pages |
| Visible text and form values | Selected text and the current value of a focused element when the app exposes it through macOS Accessibility      | Secure password fields; text from excluded apps and websites                        |
| Keyboard activity            | A count of key presses and recognized shortcuts such as `cmd+c`                                                   | Key-by-key typed content and modifier-only presses                                  |
| Clicks                       | That a mouse button was clicked and, when available, the kind of control involved, such as a button or text field | Pointer coordinates, screenshots, and a guaranteed copy of the clicked label        |
| Context changes              | App switches, window changes, and supported browser page changes                                                  | Background network traffic and a record of every scroll position                    |

Activity History does not save your typing key by key. A visible document selection or form value can still appear when an app exposes it through Accessibility. Exclude messaging, health, financial, password, and other sensitive apps or websites before recording if you do not want that content in your history.

Activity History leaves out audio, microphone input, screenshots, clipboard contents, and private-browsing activity. Private browsing stays excluded even when the browser or website is otherwise allowed.

## Choose Which Apps And Websites Count

The **Permissions** tab provides separate controls for apps and websites:

- **Record all supported apps or websites** uses a skip list.
- **Record selected apps or websites** uses a record list.

For apps, select **Choose apps** and search or scroll through the apps installed on your Mac. Each choice shows the app's name and icon, and you can select several apps before closing the list. The picker works before you grant Accessibility access. Accessibility is still required when recording starts.

These personal settings control what the Mac records. Company settings do not add activity that your personal settings left out.

Browser activity must pass both controls. For example, allowing `example.com` does not record it when the browser itself is excluded. An exclusion always wins.

Enter websites as exact hostnames, such as `example.com` or `docs.example.com`. Neotask normalizes `www.`, but a rule for a parent hostname does not automatically cover each subdomain. Add the relevant subdomains separately.

You can also exclude an app from one of its timeline entries. Save permission changes before expecting the recorder to apply them.

## How Summaries Work

Activity is collected as short local segments. When new activity is ready, Neotask normally prepares a readable summary about every 10 minutes. It can also create larger six-hour and daily summaries so agents can answer broader questions without scanning a long list of individual events.

There are two separate kinds of processing:

- **Recording activity** does not call an AI model and does not use model tokens.
- **Writing readable summaries** uses your configured model provider. These calls use model tokens, and the selected activity sent for a summary may leave your Mac under that provider's terms and your Neotask configuration.

Summary work runs only when there is finalized activity to process. It is not a constant model call every few seconds.

## How Work Is Assigned To A Company

One recorder covers the Mac, even when your Neotask account has several companies. Neotask assigns each summary separately instead of treating all computer activity as work for the company page you last opened.

Each company can also have its own app and website matching rules. An admin can let the company use all supported apps and websites with skip lists, or limit it to selected apps and websites. These rules help with company assignment. They do not change what your Mac records.

It first checks useful context that can be explained to you:

- Apps and websites selected or skipped in each company's Activity History settings.
- A website domain connected with one of your companies.
- A file or repository inside a company workspace managed by Neotask.
- A work identity domain associated with the company.
- An active Neotask company or session when that context is available.
- Company corrections you made on earlier summaries.

A skip can settle a close match. For example, if one company skips Safari and another company allows it, Safari activity cannot be assigned to the company that skipped it. Neotask still considers the other available context before assigning the work.

When two or more companies are still plausible, a built-in local matching model can compare the activity description with the candidate companies and their approved task or project names. This matching stays on the Mac and does not spend provider credits. It runs only for a real tie between companies that already have supporting context.

The local tie-break waits when the Mac is on battery or under high processor or thermal load. Recording and ordinary summaries continue. If Neotask cannot make a responsible choice, the summary stays **Unattributed** or shows low confidence for you to review.

Open the summary's company control to assign or correct it. Your correction updates that summary and helps with similar close calls later. Unattributed work is not silently used as evidence for another company's SOP.

## How Agents Use Activity History

Agents can search the summaries through Activity History when you ask. Useful requests include:

- "What was I working on before lunch?"
- "Where did I leave off on the Acme renewal?"
- "Find the project I was reviewing yesterday."
- "Prepare a standup from today's work."
- "Draft an SOP from this repeated process."

History content serves only as reference material. A sentence on a web page or in a document cannot grant permission to make the agent run a tool. Activity History also does not start Computer Use or perform actions on your behalf.

For company work, only attributed summaries are eligible to inform that company's learning and planning settings. Owners and admins can turn those uses off independently from recording access.

## How SOP Learning Works

When SOP learning is allowed for a company, Neotask can look for repeated patterns in company-attributed summaries and suggest an SOP. Each suggestion includes supporting activity references for review.

An Activity History suggestion follows the company's normal approval policy. Approval creates an editable draft. It does not publish an active SOP or automatically execute the observed workflow. Ambiguous and unattributed work stays outside another company's SOP evidence until it is resolved.

## Record A Set Work Period With Watcher

Watcher records one set period of work, then checks that period for useful process suggestions. The default is five days. You can choose 30 minutes, one day, five days, seven days, or a custom time from 30 minutes to 30 days.

Use a short session to record one workflow. Use a longer session when you want Neotask to study a workday or work week.

To start a session:

1. Turn on Activity History, save your recording choices, and grant consent on the Mac.
2. Open the **History** tab in **Settings > Activity History**.
3. Choose a time under **Watcher session**, then select **Start Watcher**.
4. Do your work as usual. Watcher uses the same app, website, private-browsing, and skip-list rules as Activity History.
5. Return after the session ends to review the result tiles. They show work summaries, clicks recorded, keys typed, apps used, websites visited, SOP suggestions, workflow rule suggestions, and skill suggestions for that session.

One Watcher session can run on the same Mac at a time. Closing the Neotask window or quitting the desktop app does not end the session. The background Gateway service and macOS recorder keep it running. The Mac must be awake and your macOS session must be available to record activity. Time spent asleep or powered off is not added to the end of the session.

Select **End and prepare suggestions** when you have finished early. Select **Cancel session** when you want to stop without running the Watcher review. Canceling does not delete activity already saved in your ordinary history. Use **Clear activity history** if you also want to remove it.

When a session ends, Neotask first finishes its readable summaries. It then checks only the session's time period for repeated work. A session can include work for several companies. Each summary keeps its own company assignment, and work is considered only for the company it was assigned to. Unattributed work is left out of company SOP, rule, and skill evidence.

Watcher can finish with zero suggestions. A short session, unrelated tasks, weak company assignment, or too little repeated evidence may provide nothing worth proposing. Neotask does not fill the result with guesses.

Watcher follows the same review controls as the rest of Neotask:

- SOP and workflow rule suggestions go to the normal company review and approval flow.
- Skill suggestions go to Skill Workshop for inspection before they can be applied.
- A suggestion cannot publish itself, approve itself, or run a workflow.
- Company admins cannot start a member's Watcher session or read that member's raw activity.

Recording still uses no model tokens. Summary writing and the end-of-session review use your configured model provider and may use model tokens. Local company matching for a close attribution decision does not use provider credits.

## Pause Or Stop Recording

Use the Neotask menu bar icon to pause Activity History for 30 minutes, one hour, or until tomorrow. The menu shows when recording will resume, and you can resume sooner from the same place.

To stop future recording, turn off **Remember my computer activity** or withdraw consent in **Settings > Activity History**. An organization owner or admin can also disable the feature for the organization. Organization and platform controls override the setting on a Mac and stop its recorder.

Stopping recording leaves existing history in place. Use the deletion controls when you also want to remove saved activity.

## Review, Correct, Export, Or Delete

The **History** tab groups summaries by date and time of day. From there you can:

- Filter the timeline, including summaries that still need a company.
- Assign or correct a company.
- Open the local summary in Finder.
- Delete one summary and its company assignment record.
- Request an SOP draft for an attributed company.

**Export summaries** saves the summaries on that Mac as a portable data file. A complete Activity History data export combines this local file with your Neotask account export. If you used Activity History on several Macs, export from each one.

**Clear activity history** can remove the last 10 minutes, last hour, last 24 hours, or all history from that Mac. Clearing cannot be undone. Larger summaries can take up to six hours to reflect a clear. Clear each Mac separately when you want to remove device-local history from several computers.

## Where Activity History Data Lives

- Raw activity segments stay in Neotask's private storage on the Mac and normally expire after about 48 hours.
- Readable summaries stay on that Mac until you delete them.
- Watcher session status and result counts stay with your account for up to 90 days. Raw activity and summary text are not stored in that Watcher session record.
- Neotask stores the account and company records needed for settings, consent, company assignment, deletion, and SOP review. Raw activity files and local summary text are not uploaded with those records.
- macOS protects these files inside Neotask's private app storage. Neotask does not add a second layer of file encryption. Use FileVault and automatic screen locking on Macs that record work.

## Company Administration

The personal timeline and consent controls live in the desktop **Settings > Activity History** screen. Company controls live in a separate place.

Owners and admins can open a company in Neotask, select **Settings**, then select **Activity History** under Autonomy. The administration page lets them:

- Allow or disable Activity History across the organization.
- Choose which roles may opt in and add member-specific allow or block overrides.
- See whether a member separately opted in, without seeing that member's activity content.
- Enable or disable SOP learning and the use of approved insights in planning.
- See whether members may start timed Watcher sessions. Members start and control their own sessions on their Mac.
- Review **Company totals** for the last seven days. These totals combine participating members' assigned work and show clicks, keys, apps, websites, SOP suggestions, workflow rule suggestions, and skill suggestions.
- Review **Your activity** on the same page. This section shows the signed-in admin's work summaries, clicks, keys, apps, websites, SOP suggestions, workflow rule suggestions, and skill suggestions for work assigned to that company.
- Open **Unassigned summaries** to separate work still being prepared from finished work that needs a company choice. Waiting work is assigned after processing finishes. An admin can assign finished work to the company currently open.
- Choose all supported apps with an app skip list, or choose only the apps that belong to the company.
- Choose all supported websites with a website skip list, or choose only the websites that belong to the company.
- Review consent events tied to shortened device identifiers.

Admins cannot consent for a member or remotely start recording on that member's Mac. They cannot read local summary text, raw events, window titles, website paths, selected text, or local file paths from the administration page.

**Company totals** and **Your activity** are separate. Company totals combine eligible work from participating members. Your activity uses the signed-in account and the company that is open. An admin can check their own results on the company page without opening personal settings.

Company app and website rules only guide company matching. Each member's personal Activity History settings remain in control of recording. The company app picker uses the apps installed on the Mac where the admin opens Neotask.

The Unassigned summaries window shows the work period and its status. It does not show the member's activity text. **Still processing** means Neotask has not finished company assignment yet. **Needs a company** means processing finished without a strong match. Only finished work can be assigned from this window.

If another admin saves a policy change first, Neotask keeps your unsaved draft and asks you to refresh before trying again.

## Troubleshooting

### Activity History Is Missing Or Read-Only

Confirm that you are using the desktop app on an Apple Silicon Mac. On a company account, ask an owner or admin to check the organization, role, and member access settings. A platform or organization disable control cannot be overridden from the Mac.

Activity History is unavailable in HIPAA deployments. Global Privacy Control or Do Not Track also prevents it from being enabled.

### Accessibility Permission Is Missing

Open **System Settings > Privacy & Security > Accessibility**, enable Neotask, then return to Activity History and select **Finish setup** or **Try again**. Neotask turns recording off when this permission is denied.

### A Website Or App Does Not Appear

Check the app and website rules. Open **Choose apps** to confirm that the app is selected or absent from the skip list. If an installed app is missing, select **Try again** and confirm that the app is in a macOS Applications folder. Browser events must pass both rules. Confirm that you added the exact hostname and any needed subdomain. Private-browsing activity is always excluded.

### A Summary Has The Wrong Company

Open its company control and choose the correct company. If no company is appropriate, leave the summary unattributed. Owners and admins can review the company's selected and skipped apps and websites in company settings.

### A Watcher Session Stopped Or Did Not Finish

Confirm that Activity History is still on, company access and SOP suggestions are still allowed, and the Gateway service is running. The recorder stops when consent, macOS permission, company access, or its short Gateway authorization expires. Reopen Neotask to see the current session status. A failed final review is retried; it does not turn an unfinished suggestion into an approved item.

### Privacy Or Data-Rights Help

See [Data Privacy & Your Rights](data-privacy-and-your-rights.md) or email **privacy@neotask.ai**. For product help, contact **support@neotask.ai**.

---

### [Agent Spending](https://neotask.ai/docs/neotask-desktop/agent-payments)

## Review And Limit Agent Spend

1. Open **Usage** to see cost by agent and model.
2. Open the agent or company budget controls.
3. Set the permitted spend and review the projection.
4. Return to **Usage** after the next run to confirm the charge.

![Usage breakdown by agent and model](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/usage-main-1280.webp)

**Agent spending** lets your AI agents pay for things on your behalf, using a company-funded card that stays inside a hard cap you set, where **every single charge needs your explicit approval before any money moves.**

It is part of **Neotask Payments**, and you turn it on in **Settings → Payments → Agent spending**.

---

## What Agent Spending Is

When an agent is working on a task and needs to buy something, a subscription, an ad credit, a one-off purchase, it cannot reach into your bank account or use a saved card on file. Instead, agent spending gives it a controlled way to ask:

- The agent requests the purchase, describing the **merchant**, the **amount**, and **why** it needs to spend.
- **You** approve or decline that exact charge.
- Only after your approval does any money move, on a dedicated **Neotask Payments card** that is funded by a wallet you top up and bounded by a hard cap you set.

It is the **outbound** side of Neotask Payments: money going **out** so your agents can get things done. (The other side of Neotask Payments, collecting **revenue** from your own customers, is covered separately; see [Neotask Payments](neotask-payments).)

---

## Why It Is Safe

Agent spending is built around two safety features that can never be turned off:

### 1. Every charge needs your approval, no exceptions

A spend request is **always** routed to you for a human decision. This is true in **every** company mode, including **Fully Autonomous**. Even a company you have set to run completely on its own will still stop and ask you before spending a cent.

There is no setting, no rule, no "trusted app" exception, and no learned auto-approval that can let an agent spend money on its own. Spending sits **above** every automatic-approval path in the system, so nothing can route around it.

### 2. A hard spend cap you control

You set a **hard cap**, the absolute ceiling on what an agent can spend. No single charge can ever exceed it, regardless of what the agent requests. The real spendable amount is always the **lower** of your hard cap and your current wallet balance, so you are protected on both ends: a charge above your cap is declined, and a charge above your funded balance is declined.

### And there is no rush

An approval **waits for you**. There is no countdown and no auto-expiry that quietly approves a charge if you do not answer in time. A pending spend request simply sits and waits until you act on it, approve it tomorrow and it is still there; ignore it and nothing happens. Money only ever moves on a deliberate human "approve."

---

## Before You Begin, Connect Neotask Payments

Agent spending is built **on top of** Neotask Payments. Before the agent-spending controls appear, you must have a connected **Neotask Payments** account for the company.

Until Neotask Payments is connected, the Agent spending section stays hidden. Once it is connected, the section appears in **Settings → Payments**.

See [Neotask Payments](neotask-payments) to set up your account first.

| Requirement | Why |
|---|---|
| A connected **Neotask Payments** account | The wallet and card are issued under your Neotask Payments account |
| Company in **build mode** | Agent spending is a company-operation feature |
| You are an **admin** on the company | Funding and cap changes are administrator actions |

---

## How To Enable It

1. Open **Settings → Payments**.
2. Find the **Agent spending** section (it appears once Neotask Payments is connected).
3. Flip the **master toggle** on.
4. A confirmation appears explaining what you are enabling. It restates the key promise:

   > Your agents will be able to request purchases on a Neotask Payments card, up to the hard cap you set. **Every charge still needs your explicit approval before any money moves**, nothing is ever spent automatically. You can set the cap to 0 or turn this off at any time to freeze all agent spending.

5. Confirm with **Enable agent spending**.

When you first enable it, the cap starts at **0** until you set one, so spending is effectively frozen until you deliberately raise the cap. You are never one toggle away from open-ended spending.

---

## Setting The Hard Cap

The spend controls show the card balance, cap, and agent-spending state together.

![Agent spend controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/agent-spend-controls-1440.webp)

The **hard spend cap** is the most important control on the page. It is the absolute ceiling on any single charge.

- Enter a dollar amount in the **Hard spend cap** field and click **Save cap**.
- A charge can never exceed this cap, no matter what the agent asks for.
- The effective limit on any one purchase is always the **lower** of your cap and your wallet balance.

Set it to the most you would ever want a single agent purchase to be, and no higher.

---

## Topping Up The Wallet

The **wallet** is the pool of company funds that backs the card. Agents can only spend what the wallet holds.

- Enter an amount in the **Top up wallet** field and click **Top up**.
- The wallet balance updates to reflect the new funded total.
- Your **wallet balance** is shown on the page so you always know how much is available.

Because spending is capped by both the wallet balance **and** the hard cap, keeping the wallet modest is itself a safety lever, agents can never spend money that isn't there.

---

## What Happens When An Agent Wants To Spend

Here is the full flow, end to end:

1. **The agent asks.** While working a task, the agent decides it needs to buy something and submits a spend request, with the merchant, the amount, the currency, and a reason.
2. **It waits for you.** The request becomes a **pending approval**. The agent's purchase is paused, it cannot proceed until you decide. There is no timeout; the request waits indefinitely.
3. **You decide.** You **approve** or **decline** the charge, right in the app, or in a connected channel where you handle approvals.
4. **On approve, the charge runs once.** Only after your approval is a **single-use** Neotask Payments card prepared for that exact purchase, scoped to the approved amount. The agent uses it once to complete the checkout.
5. **The card matches the approval.** A charge is only allowed through if it matches your approval and stays within both the hard cap and the wallet balance. Anything over-cap, over-balance, or unmatched is **declined** automatically.
6. **It's recorded.** The completed charge shows the **real merchant** and the **app** it ran for, so your records always reflect exactly what was bought and on whose behalf.

If you **decline**, no money moves and the agent does not get the purchase, it carries on without it.

---

## How To Pause Or Freeze Spending

You are always one action away from stopping all agent spending. Two ways, either works instantly:

| To... | Do this | Effect |
|---|---|---|
| **Freeze spending but keep it set up** | Set the **hard cap to 0** | No charge can be approved or run, the ceiling is zero |
| **Turn agent spending off entirely** | Flip the **master toggle off** | Spending is disabled; the cap is set to 0 to freeze everything |

Turning the master toggle off automatically drops the cap to 0, so a single click fully freezes agent spending. You can re-enable it later, your wallet balance and settings are preserved, and you will be asked to acknowledge the terms again.

---

## Key Things To Remember

- **Agents can NEVER spend without your approval.** There is no mode, rule, or automation that bypasses this. Full autonomy does **not** mean autonomous spending.
- **The hard cap is absolute.** No charge ever exceeds it.
- **Approvals never expire.** A pending spend request waits for you, however long it takes.
- **You can freeze it instantly** by setting the cap to 0 or toggling it off.
- **Every charge is recorded** with the real merchant and the app it ran for.

---

## Related

- [Neotask Payments](neotask-payments)
- [Approvals & Safety](auto-companies-approvals-and-safety)
- [Company Modes](company-modes)

---

### [App Groups](https://neotask.ai/docs/neotask-desktop/app-groups)

## Create An App Group

1. Open **Apps**.
2. Select **New group**.
3. Add the connected services the agent should use together.
4. Review authentication for each service before saving.

![Apps workspace with functional groups and connected services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/apps-main-1280.webp)

## What Are App Groups?

App Groups are workflow bundles that organize your MCP app integrations into themed collections. Think of them like folders on your phone, you group related apps together and your agent uses them as a coordinated toolkit.

For example, you might create a "Sales" group with Salesforce, HubSpot, Gmail, and Google Calendar, or a "DevOps" group with GitHub, Jira, Docker Hub, and Slack.

Free users can connect apps and use them inside allowed chat, boards, and coding workflows. Creating automations from app groups, scheduling app-group work, or running company tasks from app groups requires a paid plan.

## Creating an App Group

### Step 1: Group Setup

Give your group a name, pick a color (8 preset options), and choose an emoji icon. Optionally enable **Workflow Memory** to let the group maintain shared context across all its apps.

### Step 2: Select Apps

Choose from 1200+ available MCP apps. Each app shows whether it needs credentials and what auth type it uses.

![Searchable picker for adding an app](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/app-picker-1280.webp)

### Step 3: Configure Credentials

For each app that requires authentication, you can:

- **Use existing credentials**, If you've already saved credentials for this app globally, the group can reuse them.
- **Enter new credentials**, Provide credentials specific to this group (e.g., a different Salesforce account for sales vs support).

The group creation modal shows the credential status for each app:
- ✓ Saved, Credentials already configured
- ✓ Using preset, Using existing global credentials
- ⚠ Missing, Credentials needed before the group can work

### Step 4: Per-App Configuration

For each app in the group, you can add a description explaining how the agent should use it in this workflow context (e.g., "Use Salesforce to look up customer accounts and update deal stages").

## Group Features

### Workflow Memory

When enabled, workflow memory lets all apps in the group share context. If your agent looks up a customer in Salesforce, that information is available when it drafts an email in Gmail, without you needing to repeat yourself.

### Color & Icon Coding

Groups display as color-coded cards with emoji icons for quick visual identification. Available colors include blue, red, green, orange, purple, pink, cyan, and amber.

### Group-Specific Credentials

Each group can have its own set of credentials, independent of global credentials. This means:

- Different Google accounts for personal vs work groups
- Separate API keys for dev vs production environments
- Isolated OAuth tokens per workflow

Credential lookup follows a chain: group-specific credentials first, then fallback to global credentials.

## Credential Management

### Supported Auth Types

| Type | Description |
|------|-------------|
| **API Key** | Single key field (e.g., OpenAI API key) |
| **OAuth** | Client ID + Client Secret pair |
| **Credentials** | Multi-field form (e.g., username + password + domain) |
| **None** | No authentication needed |

### Credential Storage

All credentials are encrypted with AES-256-GCM using a machine-derived key. They're stored locally at `~/.neotask/mcp-credentials.enc`.

### Credential Masking

When viewing saved credentials, sensitive values are masked:
- Short values: first 3 chars + `...` + last 2 chars
- Long values: first 6 chars + `...` + last 4 chars

### Legacy Migration

If you previously saved credentials using older formats (separate API key, OAuth, or custom URL stores), they're automatically migrated to the unified encrypted store.

## Use Cases

| Group | Apps | Purpose |
|-------|------|---------|
| **Sales** | Salesforce, HubSpot, Gmail, Calendar | CRM workflow |
| **DevOps** | GitHub, Jira, Docker Hub, Slack | Development pipeline |
| **Content** | WordPress, Canva, Google Docs, Buffer | Content creation |
| **Support** | Zendesk, Intercom, Notion, Slack | Customer service |
| **Finance** | QuickBooks, Stripe, Google Sheets | Financial tracking |
| **Marketing** | Mailchimp, Google Analytics, Facebook Ads | Campaign management |

---

### [Bring Your Own Key (BYOK)](https://neotask.ai/docs/neotask-desktop/byok)

## Overview

BYOK mode lets you use your own API keys from supported AI providers instead of Neotask's managed infrastructure. This gives you full control over which models you use and direct billing from your provider for the base token cost. Note that a **20% platform fee** and any applicable **automation markup** are still deducted from your Neotask credit balance, even in BYOK mode. You save on the base API cost, but platform fees still apply.

## How It Works

1. **Get an API key** from your preferred provider
2. **Add the key** in Neotask's settings
3. **Your agent** uses your key for all LLM calls
4. **Base token costs** go directly to your provider; platform fees are deducted from your Neotask credits

If you don't provide your own keys, Neotask uses its shared pooled keys with usage tracked and billed through your subscription plan.

![Own API Keys provider catalog in Neotask settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/own-keys-catalog-1280.webp)

## Add a provider key

1. Open **Settings**.
2. Select **Models**.
3. Choose **Own API Keys**.
4. Select the provider whose key you want to use.
5. Enter the fields requested by that provider and save.
6. Confirm the provider reports a ready state before starting a run.

The provider catalog in the app is the current list. The model picker updates after a provider is connected.

On the Free plan, BYOK only changes who pays the provider token bill. It does not remove Free-plan product gates. Free users still have the first `10` messages/events, then `5` messages/events per user-local calendar day, and they still cannot run company tasks, create custom agents, or create their own automations.

## Supported Providers

| Provider | Models | How to Get a Key |
|----------|--------|-----------------|
| **Anthropic** | Claude Opus 4.6, Claude Sonnet 4.1, Claude Haiku 4.5 | [console.anthropic.com](https://console.anthropic.com) |
| **OpenAI** | GPT-5.2, GPT-5.1, GPT-4 Turbo, GPT-4, DALL-E 3 | [platform.openai.com](https://platform.openai.com) |
| **Google AI** | Gemini 3 Pro, Gemini 2 Flash, Gemini 1 Pro | [aistudio.google.com](https://aistudio.google.com) |
| **OpenRouter** | 100+ models from multiple providers | [openrouter.ai](https://openrouter.ai) |
| **DeepSeek** | DeepSeek Chat, DeepSeek Coder | [platform.deepseek.com](https://platform.deepseek.com) |

## Key Security

- **Server-side storage**, Keys are encrypted and stored on the Neotask license server, not in local plaintext files
- **HMAC-signed requests**, Key management operations are signed with your device secret
- **Masked display**, Only the last 4 characters of your key are visible in the UI
- **Device binding**, Keys are associated with your device + license for additional security

## Benefits of BYOK

| Feature | Managed (Default) | BYOK |
|---------|-------------------|------|
| Setup required | None | Get API key from provider |
| Billing | Through Neotask plan | Base cost direct to provider; 20% platform fee + automation markup from credits |
| Model access | Models included in plan | All models your key supports |
| Pricing | Included in subscription | Provider rates + 20% platform fee from credits |
| Rate limits | Shared pool limits | Your account's limits |
| Cost control | Daily budget in app | Provider dashboard + app budget |

## When to Use BYOK

- **You want specific models** not available on the managed plan
- **You have existing API credits** or enterprise agreements with providers
- **You want higher rate limits** than the shared pool provides
- **You prefer direct billing** from the provider
- **Your organization requires** keys to stay within your cloud account

BYOK provider balance must be valid for every allowed Free message. If the provider key runs out of credits or hits a billing limit, Free usage stops until the provider account is funded or the user switches to hosted credits with enough Neotask balance.

---

### [Company Modes](https://neotask.ai/docs/neotask-desktop/company-modes)

## Check The Active Company Controls

1. Open the company and select **Plan**.
2. Read the current autonomy and heartbeat status in the controls rail.
3. Change the mode only after reviewing approval defaults and budget limits.
4. Confirm the new mode before triggering work.

![Company plan with autonomy and heartbeat controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-autonomy-1280.webp)

Every autonomous company runs in one of two modes. **Build mode** is for a brand-new company you are creating from an idea and growing toward launch and revenue. **Manage mode** is for an established company whose ongoing operations you want run and improved week after week. The mode tells the AI what *kind* of help it is giving, and it changes how every weekly plan is written.

---

## The two modes at a glance

**Build mode** treats your company as a brand-new venture with nothing connected yet. The AI plans like a founder's right hand: a week of forward-moving build work, research, drafting, producing content, and getting ready to launch, where every task can run immediately with no setup from you.

**Manage mode** treats your company as a running business. The AI plans like a chief-of-staff or operations lead: a weekly operations review that keeps what is working, drops what is not, and adds outreach or sales only where they genuinely fit your business.

There is also a **Both** setting, which behaves like Build mode while you are still pre-launch and automatically shifts to Manage mode once your company has shipped a product. If you are unsure, Both is the safe choice, it never treats a company as further along than it really is.

---

## Build vs Manage

| | **Build mode** | **Manage mode** |
|---|---|---|
| **For** | A brand-new company being built from an idea | An established company already operating |
| **The AI acts as** | A founder's right hand getting you to launch | A chief-of-staff running weekly operations |
| **What a week looks like** | A sequence of one-time build tasks that compound toward launch | A revised set of recurring operations that keep the business running |
| **Tools it uses** | Only built-in, no-setup capabilities (see below) | Your connected apps, integrations, and installed skills |
| **Outreach & sales** | None, nothing is sent externally and nothing is sold yet | Added only where they fit your actual business and stage |
| **Goal** | Reach a real, launchable product | Keep what works, fix what does not, and improve results |

---

## Build mode in detail

Build mode is for a company that exists as an idea but has not launched. Its guiding rule is that **every task must be completable right now, with zero setup from you.** A brand-new company has not connected any accounts yet, so the AI will not plan anything that would require you to link a service, paste a key, or sign in somewhere first. Every task is ready to run the moment the plan is created.

To honor that, Build mode works only with the capabilities that are built in and need no accounts:

- generating images
- writing and publishing posts on your company's own built-in blog (the only publishing channel in this mode)
- drafting content such as emails and documents, saved as files rather than sent
- researching and summarizing information from the web
- using your company's own internal files and memory
- a small set of no-setup helpers (for example, working with PDFs or local notes)

A typical Build week is a handful of one-time tasks spread across the days so progress compounds, research, then drafting, then producing, then preparing to launch, rather than a fixed daily routine. The very first task is usually to write and publish your company's first blog post right away, so there is visible progress from day one.

What Build mode deliberately will **not** do: it will not run cold-email campaigns, daily performance reports, or ad-management loops, and it will not try to sell a product that does not exist yet. Those belong to an established business, not a brand-new build. When your company grows past the early stage, it moves out of Build mode automatically.

---

## Manage mode in detail

Manage mode is for a company that is already up and running. Each week the AI performs an **operations review**: it looks at how the business actually works today, keeps the operations that are producing results, adjusts or drops the ones that are not, and adds new operations only where they genuinely fit.

Because the company is established, Manage mode is allowed to use everything you have connected, your apps, integrations, and any skills you have installed. It reasons across the whole business: strategy, product, content, marketing, sales, operations, and finance, adapting to how your company really operates rather than following a single fixed script.

A Manage week leans toward recurring operations on a sensible cadence, and it always includes a check on the health of the business and a review of performance and metrics. Importantly, Manage mode does **not** force the same cold-email-and-daily-report routine onto every company. Outreach and sales tasks appear only when they suit your business and the stage it is at, never as a blanket default.

---

## Product readiness, why your company won't sell before it's ready

Independent of the mode, your company moves through stages as it matures: **idea → pre-launch → launched → selling.** It advances on its own as real progress happens:

- It leaves **idea** once website work begins or the first build task is finished.
- It becomes **pre-launch** while the product is being built.
- It becomes **launched** once the website is live and enough core build work has shipped.
- It only reaches **selling** once it is launched **and** payments are connected.

Selling is held back until your company is genuinely ready. Before it ever takes a real customer payment, every condition on its own "ready to sell" checklist, a working live page, connected payments, fulfillment in place, and anything else specific to your business, must be confirmed by real evidence, not just assumed. Until then, any selling task is automatically held back. (Your company can still legitimately collect a waitlist or pre-orders in the meantime when that is the right move.) This keeps a company from trying to sell something before it actually exists or before it can take money safely.

---

## Where you set and see the mode

You set and change the operating mode on the **Profile** tab of your company. The Profile tab is where you manage your company's identity and core business information, and it is where you manage **build/manage mode** alongside the company stage. You can switch a company between **Build**, **Manage**, and **Both** there at any time, and adjust the company stage to match where the business really is.

The mode is usually set for you when you first create the company: starting from an existing company website sets it to Manage, while starting from a new idea sets it to Build. You only need to revisit the Profile tab if you want to change it later.

---

## Related

- [Planning & Weekly Updates](planning)
- [Onboarding & Plan Setup](auto-companies-onboarding)
- [Operations, Controls & Settings](auto-companies-operations-controls-and-settings)

---

### [Custom Skills And Plugins](https://neotask.ai/docs/neotask-desktop/custom-skills-and-plugins)

Custom skills and plugins are capabilities that you or your agents create, or
that you choose to import from another supported app. Neotask keeps them
separate from built-in and managed catalog content so ownership, review,
setup, and storage remain clear.

This page describes the durable custom-artifact flow. If that service is not
enabled for your workspace, or a save cannot be finalized, Neotask reports a
pending or failed state instead of claiming that the item is safely stored.

## What Counts As Custom

- **Custom skills** include skills created with **Create Skill**, skills an
  agent creates or learns for your workspace, and portable skills selected in
  an external-app import.
- **Custom plugins** include plugin packages you create and compatible plugin
  packages you import. Source from a foreign plugin format is kept as an inert
  conversion draft, not silently installed.
- **Built-in skills and plugins** ship with Neotask. Their code is not uploaded
  into customer artifact storage.
- **Neotask and company managed catalog plugins** continue to use their
  established GitHub and managed release flow. Importing a local plugin does
  not create a repository or push code to GitHub.

## Where An Item Appears

The destination you choose is part of the saved target record and, for a
canonical item, its assignment:

| Destination               | Skill location                      | Plugin location                                 |
| ------------------------- | ----------------------------------- | ----------------------------------------------- |
| Main personal workspace   | Your **My Skills**                  | **Add-ons**, as metadata with runtime isolation |
| Selected standalone agent | That agent's **My Skills**          | **Add-ons** metadata; no executable assignment  |
| Company                   | That company's **Skills** workspace | **Add-ons** metadata; no executable assignment  |

Personal, agent, and company assignments are not inferred from a local folder.
If the target cannot be verified, Neotask does not publish the item to another
scope.

Customer plugin code does not execute in the current desktop Gateway process.
Its source, provenance, review state, and declared setup requirements remain
visible in Add-ons, but its card shows **Runtime isolation required** and does
not offer Add, credential-save, or binary-install actions. A future executable
customer plugin runtime must be isolated, killable, and scoped to the exact
assignment and target. Managed built-in, catalog, and company-released plugins
keep their established trusted release and runtime paths.

For customer plugins, **metadata-only** describes the card and runtime
registration. It does not mean that retained source bytes are discarded: an
eligible source package can remain in private cloud backup and in the inert
local quarantine described below, without becoming installed or executable.

## What A Card Shows

My Skills and Add-ons receive display metadata only, such as the name, summary,
source, target, version, lifecycle state, dependency names, and setup status.
Cards do not receive or render:

- skill instructions, plugin source, support files, or package archives;
- raw manifests or package scripts;
- local filesystem paths, private object keys, or signed download URLs;
- API keys, OAuth tokens, environment values, or secret-derived hashes.

This keeps the readable catalog view separate from the files that are
inspected, stored, and materialized for runtime use.

## Save, Review, And Install Lifecycle

1. Neotask builds a bounded package for the exact destination and computes its
   checksum.
2. A successful durable save uploads the bytes through a short-lived,
   exact-object upload authorization into private encrypted storage. The app
   never receives general AWS credentials.
3. The package is checked for integrity, unsafe archive paths, secrets,
   compatibility, dependency declarations, and the expected skill or plugin
   family.
4. A compatible custom skill can enter review and setup. Customer plugin
   packages can enter review, but executable packages remain metadata-only
   with **Runtime isolation required**. A foreign or source-only plugin remains
   inert as **Conversion required** until a separately inspected Neotask plugin
   package exists.
5. For a compatible custom skill, the local copy is materialized only for an
   authorized assignment and exact version. Neotask verifies the signed
   manifest and bytes, then restores the complete skill file tree before use.
   Gateway normally runs the skill from that verified local tree; it does not
   stream code from cloud storage at execution time. Customer plugin code is
   not materialized for execution while runtime isolation is required.

Common card states include **Draft**, **Conversion required**, **Review
pending**, **Available**, **Needs setup**, **Needs authentication**,
**Installed on this device**, **Failed**, **Revoked**, and **Archived**. The
displayed state is meant to describe the actual lifecycle, not just the
presence of a file on one computer. **Installed on this device** does not apply
to executable customer plugin code while runtime isolation is required.

## Dependencies And Credentials

A custom item can declare requirements without embedding their values in the
package.

- **External binaries for custom skills** use the managed component catalog
  when Neotask knows an approved source, version, and integrity rule.
  Installation requires the displayed user action and an exact readback
  check. Unknown installers and package-provided shell commands are not run
  automatically. Customer plugin binary setup remains unavailable while its
  executable runtime is isolated.
- **API keys and environment values for custom skills** are stored in the
  encrypted credential store for the exact user, agent, or company assignment.
  The card receives only the field name and whether setup is complete. Neotask
  does not accept new customer plugin credentials while plugin execution is
  unavailable. A **Clear credentials** action remains available to erase
  values saved by an earlier version.
- **OAuth, local-app, and MCP requirements** are declarations and setup status
  only unless Neotask already has a separately registered authority for that
  exact connection. Importing a skill or plugin does not copy a source app's
  general login, grant new OAuth authority, or change MCP runtime behavior.

A custom skill stays in **Needs setup** or **Needs authentication** until every
required dependency is satisfied. A customer plugin stays inactive with
**Runtime isolation required**, even when all declared dependencies are
present. Metadata alone never marks an item installed or executable.

## Storage And Sync

For a successful durable save, MongoDB and Construct store the target and, for
canonical versions, the assignment, along with lifecycle, approval, provenance,
dependency metadata, and audit state. Private S3-backed object storage holds
user-created, imported, and Create Skill package or source bytes under opaque
tenant and target namespaces. It is the durable byte authority and cloud backup
for those customer artifacts. Built-in bytes remain part of the Neotask
release/build or managed catalog and are not moved into customer storage.
Managed GitHub/catalog paths are unchanged.

For new customer-created or imported bytes, Neotask does not treat an old
metadata-only registration as a completed durable save. This does not convert
or bulk-upload existing installed metadata: built-in, managed catalog/GitHub,
scoped, and activity-synced installation records keep their existing metadata
and release paths.

For an active custom skill assignment, Neotask downloads only the exact signed
version, verifies its manifest, digest, files, and modes, and materializes its
complete local working tree, including `SKILL.md`, references, templates,
scripts, and assets, with safe executable modes preserved. That local tree is
the normal Gateway execution source. Create, approval, and support-file updates
do not report success until the exact tree is verified locally, and they never
substitute an unverified direct local write when cloud materialization fails.
The durable version lets another authorized clean device restore the same tree,
while local presence alone never creates ownership or assignment authority.
Customer plugin source may be retained durably for review, but its code/bundle
remains metadata-only and inert until an isolated runtime exists.

On another authorized device, retained customer plugin source can also be
restored as the exact verified package archive and a receipt in a private local
quarantine backup. That backup sits outside every plugin-discovery folder. It
is not extracted into a plugin tree, registered, sent to the page, or executed;
it exists for recovery and a later explicit review/conversion flow only.

After a successful durable save, another authorized desktop can discover the
saved item. For a reviewed item with an active assignment, it can request the
exact approved version. If cloud reconciliation is unavailable, Neotask keeps
an honest retry or unavailable state and does not erase the last verified local
copy or clear its in-memory exact credentials merely because authentication
expired or a remote response is empty or transient. Only a definitive response
that the exact assignment is forbidden, absent, or gone can deauthorize the
local lifecycle state.

## Imports And Provenance

An imported custom item keeps its supported source provenance, original
timestamps, destination association, and content identity. Neotask records the
import and later artifact lifecycle in its own ledger. It does not copy the
source provider's private telemetry, hidden reasoning, or internal
administrative audit logs.

The source app is read locally during import. It is not contacted merely to
copy local files. Any later connection to an external service is governed by
that connection's own authorization and setup flow.

The supported migration flow for Claude, Hermes, and Codex personal or
default-agent custom skills publishes them through the same durable
customer-artifact path and then materializes the verified local tree. An item
is not reported as migrated while authorization, review, setup, authentication,
or integrity checks are unresolved. Claude project skills and commands remain
explicitly local-only: they keep their safe project-local copy behavior and are
not represented as cloud-backed artifacts. Managed Codex plugins, catalog
items, GitHub content, and built-ins remain on their established paths.

When an import finds only a remote marketplace plugin reference, Neotask saves
sanitized reference metadata as an inert Add-ons draft so it remains visible
on another authorized device. It does not clone, download, install, or execute
the referenced plugin. Conversion still requires a separately inspected
Neotask plugin package.

See [Import From Other AI Apps](external-agent-import.md) for supported sources,
selection, audit-trail details, and trust boundaries.

## Create Skill And Coding Mode

**Create Skill** uses the same custom-artifact rules and saves to the active
personal, standalone-agent, or company target. Coding mode can prepare a
custom plugin source draft for review. Saving that draft does not create a
GitHub repository and does not make the code executable; it enters the same
private inspection and conversion lifecycle described above.

## Removing Or Revoking An Item

Use **Disable** on an exact custom-artifact card to remove its activation
authority from this workspace. The durable assignment is changed first, then
desktop reconciliation removes or quarantines its local runtime copy. The card,
source, and immutable version remain available so **Enable** or **Reinstall**
can restore the same authorized version on this or another device. Disabling
also retires credentials bound to that exact assignment atomically with the
desired-state change; it never removes another assignment's credentials.
An earlier in-flight credential request cannot restore those values after the
assignment has been disabled, revoked, or erased. Enable restores assignment
availability, but it does not resurrect retired credential values. Neotask
reports the local change as complete only after a complete current
reconciliation succeeds; if that pass is unavailable or incomplete, the card
stays pending instead of claiming that the device already converged.

Deleting an owning target is a separate erasure lifecycle. Tenant deletion,
company deletion (including company-owned agent and coding-project targets),
standalone-agent deletion, tenant-member offboarding, and coding-project
deletion fence new artifact work before removing matching metadata and every
stored object version in the target's opaque namespace. A remote deletion
failure retains opaque retry debt instead of reporting erasure complete. Shared
immutable versions remain while another active assignment still references
them. Even a target deleted before its first upload keeps an erased storage
marker, so a late retry cannot silently recreate authority for that deleted
target id.

Physical deletion of one individual artifact, including source/version
reference-count cleanup in object storage, is not currently a card action.
**Disable** is therefore reversible and must not be described as permanent
artifact deletion. The separate **Clear credentials** action removes values for
an exact authorized assignment; it does not delete the artifact record. An
absent card or failed network read is never treated as an implicit deletion.

## Related

- [Skills](skills.md)
- [Skill Learning](skill-learning.md)
- [Import From Other AI Apps](external-agent-import.md)
- [Neotask Code](neotask-code.md)
- [Security](security.md)

---

### [Import From Other AI Apps](https://neotask.ai/docs/neotask-desktop/external-agent-import)

Neotask Desktop can bring supported local setup and recent conversation history
into Neotask without changing the source app's files. The import screen shows
only apps for which it finds eligible local material on the current device.
This detected-only behavior keeps unavailable or empty providers out of the
picker.

## Start An Import

1. Open **Settings → Import**. You can also choose **Import agent** during
   onboarding or from Add Agent.
2. Select one detected app.
3. Review the **Tools & setup**, **Projects**, and **Chat sessions** groups.
4. Choose whether to create an agent, add to an existing agent, or apply
   eligible shared setup globally. Company Settings keeps the import inside
   that company.
5. Run the import and review the result report, including anything skipped or
   requiring attention.

Neotask imports one provider at a time. It does not offer a folder picker or
accept an arbitrary source root from the import screen.

## Supported Local Sources

| App               | Local material used by the Desktop importer                                                                                                         | Conversation destination                                                                                                                           |
| ----------------- | --------------------------------------------------------------------------------------------------------------------------------------------------- | -------------------------------------------------------------------------------------------------------------------------------------------------- |
| **ChatGPT**       | Eligible local Desktop and Work sessions                                                                                                            | Project-backed Work sessions go to Code; genuinely projectless local sessions go to Chat. Cloud-only ChatGPT history is not a local import source. |
| **Claude Code**   | Eligible local setup, projects, and recent sessions                                                                                                 | Code                                                                                                                                               |
| **Claude Cowork** | Eligible local Cowork setup and conversations                                                                                                       | Chat                                                                                                                                               |
| **Codex**         | Eligible Codex CLI setup, projects, and recent coding sessions                                                                                      | Code                                                                                                                                               |
| **Cursor**        | Eligible local setup, projects, and recent sessions                                                                                                 | Code                                                                                                                                               |
| **Grok Build**    | Eligible Grok Build 1.0.5 setup and recent local sessions from an absolute `GROK_HOME`, or `~/.grok` by default                                     | A verified Git repository or explicit source workspace goes to Code; an ordinary or projectless working directory goes to Chat.                    |
| **Kimi Code**     | Current Kimi Code 0.38 data from an absolute `KIMI_CODE_HOME`, or `~/.kimi-code` by default, plus deduplicated eligible legacy data under `~/.kimi` | A repository or explicit Kimi workspace goes to Code; a genuinely projectless conversation goes to Chat.                                           |

An app does not appear merely because its software is installed. It appears
only when Neotask detects at least one eligible item. Credential-only,
cache-only, log-only, or otherwise non-importable state does not create a
provider row.

## What Can Be Imported

Available items depend on what the source app stores and what Neotask can map
safely. Eligible items can include:

- instructions, rules, and supported customization settings;
- skills, commands, agents, and safe plugin components;
- hooks and other executable setup, subject to the review rules below;
- MCP server definitions and transferable MCP-specific credentials;
- project associations; and
- recent active session history, including original titles and dates and
  visible user, assistant, tool-call, and tool-result activity.

Session import is bounded to the most recent 30 days and up to 50 eligible
sessions per provider. Oversized, malformed, changing-during-read, or
out-of-scope files are skipped instead of being partially trusted.

### Kimi Code current and legacy formats

For Kimi Code 0.38, Neotask recognizes the current session metadata and the
main conversation wire, as well as the legacy workspace index and legacy
session context layout. When equivalent current and legacy items overlap, the
current format wins and the result is deduplicated.

Neotask reads active main-conversation history. Archived or deleted Kimi
sessions are omitted before transcript reading. Visible parent orchestration,
including a visible tool call and its result, stays in the parent transcript;
private subordinate-agent wires are not read or flattened into that
conversation.

### Grok Build format

For Grok Build 1.0.5, Neotask reads a bounded local session summary and prefers
the session's bounded updates stream. It uses bounded chat history only when
the updates stream is unavailable. Hidden or subagent sessions and raw
reasoning records are excluded. Detection is local and read-only; it does not
query xAI for Grok history.

## Code And Chat Routing

Tool use alone never decides whether a conversation belongs in Code. Neotask
uses source-owned workspace information and verifies the local project
boundary:

- a verified repository or explicit source workspace is associated with the
  matching Neotask Code project; and
- a genuinely projectless conversation is imported into Agent Chat.

This prevents a projectless conversation that happened to call a tool from
being placed in an unrelated coding project. Project identities are stable on
the device, so the same project can converge across supported coding providers
without creating one duplicate project per provider.

## Audit Trail And Provenance

**Does imported state keep its audit trail?**

Yes. Imported state preserves its source provenance, original timestamps,
session/project associations, and visible user, assistant, and tool activity.
Neotask also records the import in its own ledger. It does not copy the
provider’s private telemetry, hidden reasoning, or internal administrative
audit logs.

The ledger uses opaque, target-aware source identities to support retries and
sync without placing raw local paths or secret values in public reports. It
records import provenance and outcome; it is not a copy of the source
provider's private operational logging system.

## Executable Setup And Trust

An import is not permission to inherit another app's trust decisions. Neotask
Safe Mode, approvals, project containment, and company boundaries remain
authoritative.

- Source folder, plugin, sandbox, shell, approval, and permission trust is not
  copied.
- Unsupported or trust-affecting settings remain review information rather
  than silently changing Neotask policy.
- Remote plugin installation metadata and executable project hooks remain
  review-only.
- Kimi plugin hooks that depend on Kimi-specific working-directory or
  environment behavior remain review-only.
- Grok project or plugin executable components are not activated when the
  source folder is not trusted. A home-level disabled-MCP decision continues
  to disable the same server in project and plugin sources.

Review the result report before enabling or recreating any item that was kept
in review-only state.

## MCP Definitions And Credentials

MCP import has separate definition, credential, materialization, connection,
and verification steps. Neotask can transfer supported static credential
values found in an MCP definition, such as environment or header values, and
can adopt an MCP-specific OAuth grant only when the source exposes a portable
record bound to the exact server name and canonical URL.

Portability is provider-specific:

- Kimi Code 0.38 and Grok Build 1.0.5 have supported exact-key MCP credential
  records that Neotask can validate.
- An access token is not treated as universally portable. Refresh is available
  only when bounded, usable token-endpoint and client metadata is present.
- Source-app account login, unrelated grants, OS-keychain-only material,
  device-bound credentials, opaque broker state, and malformed or mismatched
  OAuth records are not copied.
- If a grant is unavailable, rejected, expired, or non-portable, reconnect the
  MCP server through Neotask's fresh local authorization flow.

Imported MCP credentials are target-scoped and encrypted at rest with
AES-256-GCM. Secret values and secret-derived hashes do not enter renderer
payloads, the Site, import reports, screenshots, persisted public config, or
logs. A bearer header is constructed only in memory for the bound connection.
Neotask reports an MCP import as successful only after the definition is
materialized and the connection path is checked; a missing executable or
required fresh login appears as an item needing attention.

## Manual Import And Autosync

A manual import applies the items you reviewed for one provider. After a
successful personal import, **Keep imports in sync** can periodically recheck
an enabled provider and apply safe, idempotent updates to its existing target.
Use **Check again** when you want an immediate re-scan.

Autosync does not activate new or changed MCP definitions, rotate MCP
credentials, or bypass review for executable or trust-affecting content. MCP
changes require a new manual review and import. Company-scoped imports follow
the controls shown in Company Settings.

Running the same unchanged import again does not duplicate agents, projects,
sessions, messages, or setup. If an imported transcript was edited locally,
Neotask does not overwrite that work. Append-only source changes can extend an
imported transcript when its existing content remains an exact untouched
prefix. If a Kimi or Grok source rewinds or branches, Neotask may replace only
unmodified imported source rows that Neotask still owns. A local edit or an
ownership or authority mismatch is reported as a conflict that needs review.

## Paths, Custom Homes, And Source Safety

Neotask uses the platform's native path rules on macOS, Windows, and Linux. A
custom `KIMI_CODE_HOME` or `GROK_HOME` must be absolute. Relative homes,
filesystem roots, broad home-directory project guesses, unsafe links,
protected paths, and files that resolve outside their approved source boundary
are rejected or skipped.

If a custom Kimi or Grok home is nested inside a project, Neotask excludes that
runtime-data tree from project hydration. Detection is read-only and does not
modify the source app's home or session files.

## Verify An Import

After the result report completes:

1. Open **Code** and confirm each repository or explicit workspace session is
   under the expected project.
2. Open **Chat** and confirm genuinely projectless conversations appear there.
3. Open a conversation and check its title, original date, message order, and
   visible tool activity.
4. Review all **needs attention**, **review-only**, and skipped rows.
5. For an imported MCP server, confirm its connection status and run a
   harmless read-only tool action. If authentication is requested, complete a
   fresh Neotask connection instead of copying credentials manually.
6. Run the same unchanged import once more when you need an idempotency check;
   the report should show existing items rather than create duplicates.

## Troubleshooting

### The provider is not listed

- Confirm the source app has eligible local setup or a recent active session
  on this device. Installation alone is not enough.
- For Kimi, confirm `KIMI_CODE_HOME` is an absolute path when set, or that the
  current or legacy default home contains eligible material.
- For Grok Build, confirm `GROK_HOME` is absolute when set and contains
  eligible Grok Build 1.0.5 material.
- Credential-only, archived-only, deleted-only, hidden-only, cache-only, or
  log-only homes are intentionally not shown.
- Select **Check again** after the source app has finished writing its files.

### A session went to Chat instead of Code

Neotask did not find a verified repository or explicit source workspace for
that session. Open the source project once so the provider records its
workspace association, then run a manual import again. An arbitrary working
directory or tool call is not enough to create a Code project.

### An item was skipped or needs review

Open the result row for the specific reason. Common causes include an archived
or hidden session, an old session outside the import window, an unsafe or
changing path, an oversized file, an untrusted executable component, a local
edit conflict, a missing MCP executable, or OAuth that requires fresh
authorization.

### The MCP server imported but is not connected

Check that the server executable or URL is available, then use **Reconnect**
or the offered authorization action. Neotask does not reuse provider-account
login or a credential that cannot be proven portable for that exact server.

## Related

- [Custom Skills And Plugins](custom-skills-and-plugins.md)
- [Neotask Code](neotask-code.md)
- [Neotask Code: Safety and Approvals](neotask-code-safety.md)
- [Supported Apps](mcp-integrations.md)
- [Security](security.md)
- [Troubleshooting](troubleshooting.md)

---

### [Models & Providers](https://neotask.ai/docs/neotask-desktop/models-providers)

## Choose the model Neotask uses

Neotask can use hosted models from supported providers or local models running on your computer. You can change the default model from **Settings > Models**.

You can also choose a different model for an agent or autonomous company. The selected provider and model appear with the run so you can see what handled the work.

## Hosted models

Open **Settings > Models** to browse the providers and models available to your account.

![Models settings showing subscription, API key, and Neotask credit connection modes](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/connection-options-1280.webp)

Neotask supports two ways to connect hosted models:

- **Credits Mode** uses your Neotask credit balance. Supported hosted providers do not require a local API key.
- **Own API Keys** connects a provider account that you manage. Provider usage is billed through that account. Neotask plan and platform charges still apply.

The available catalog can change as providers release models or retire older versions. The model picker in Settings is the current list for your account.

### Connect a model source

1. Open **Settings**.
2. Select **Models**.
3. Choose **CLI & OAuth**, **Own API Keys**, or **Neotask Credits**.
4. Select a provider and complete the connection shown in the app.
5. Return to **Models** and confirm the provider shows as ready.

The catalog shown in **Neotask Credits** is the source of truth for the providers and models available through your credit balance.

![Neotask Credits catalog showing models from NVIDIA and other providers](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/credits-catalog-nvidia-1280.webp)

## Local models

Local models run on this computer, or on another computer on your local network that you add as a custom server. In **Settings > Models**, **Local models** is its own choice, the fourth card beside **Credits**, **Own API Keys** and **OAuth**; it is not part of Own API Keys. Choosing it shows the Local models page, which detects Ollama, LM Studio, Jan, Docker Model Runner and custom OpenAI-compatible servers running on this computer and lists the models each supported local model source offers. To use a server on another computer, select **Add a custom server** and confirm that it is on your local network. Only use a computer you trust: your prompts are sent to it.

Choosing **Local models** makes a connected local model your default on this computer and leaves your hosted mode (Credits, Own API Keys or OAuth) as it was. Choosing a hosted card again returns to your hosted default. Local models also appear in every model picker with a **Local** label, whichever card is chosen, so you can choose one for an agent, team or task.

Local models are always free. Neotask never bills them, and they never count against your plan limits. Quality and speed depend on this computer. Hosted models stay available.

### Set up a local model

1. Open **Settings > Models** and choose the **Local models** card.
2. If no local model is connected yet, setup opens. Otherwise select **Set up a local model**.
3. Follow the steps shown. Neotask installs Ollama if it is missing and downloads a Qwen model sized to this computer's memory.
4. Neotask checks that the model can call tools and use a context of at least 32K tokens before it connects the model.
5. Select the model as your default or for the work where you want to use it.

The local model source must stay running while Neotask uses its models.

## Set the default model

1. Open **Settings > Models**.
2. Choose a provider.
3. Select a model.
4. Set it as the default.

The default handles work that has no separate model assignment.

## Use your own provider key

Open **Own API Keys** to see the current provider catalog. Select a provider, enter the requested credential, save it, and confirm the provider reports a ready state.

![Own API Keys provider catalog in Models settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/own-keys-catalog-1280.webp)

## Assign a model to specific work

You can choose a different model for an agent or autonomous company. This lets you keep one default while using another model for selected work.

Open the model assignment control for the agent or company. Choose a provider and model. Save the assignment. New runs from that agent or company will use the selected model.

## Smart Model Routing

Smart Model Routing can choose from the models available to your account. Your provider connections and model settings determine the choices it can use.

## Check which model ran

Open the active run or session history. The run record shows the provider and model used for that session. Usage and cost remain attached to the session when those details are available.

## Related guides

- [Bring Your Own Key](/docs/neotask-desktop/byok)
- [Billing & Plans](/docs/neotask-desktop/billing)
- [Agents](/docs/neotask-desktop/agents)
- [Autonomous Companies](/docs/neotask-desktop/auto-companies)

---

### [Neotask Code](https://neotask.ai/docs/neotask-desktop/neotask-code)

## Start And Review A Coding Task

1. Open **Code**.
2. Select or create a project.
3. Describe the change in the project chat.
4. Review changed files and verification output before committing.

![Coding workspace with project chat and changed files](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/code-main-1280.webp)

Neotask Code is the built-in coding agent inside the Neotask desktop app. You describe what you want in plain language, and it writes and edits real code in a real project on your machine, runs your build and tests, and (when you connect a repository) commits and pushes your work to GitHub. Everything runs inside the same Neotask app you already use for agents, companies, and channels, so your plan, credits, and approval settings all apply.

This page explains what Neotask Code is, how projects and conversations work, and the different coding engines you can run it on. For the git and GitHub side, see [Neotask Code: Git and GitHub](neotask-code-git.md). For how approvals and containment keep it safe, see [Neotask Code: Safety and Approvals](neotask-code-safety.md). To bring eligible setup and recent local history from Claude Code, Codex, Cursor, Grok Build, Kimi Code, and other detected apps into Neotask, see [Import From Other AI Apps](/docs/neotask-desktop/external-agent-import).

---

## What You Can Build

Neotask Code is a general-purpose coding assistant. You can use it to:

- Start a new app or script from scratch.
- Open an existing project and make changes across many files.
- Fix bugs, refactor, and add features by describing the outcome you want.
- Run your project's build, test, and lint commands and act on the results.
- Preview a web project against its own local development server.
- Connect a GitHub repository and commit, push, branch, and open pull requests directly from the conversation.

The agent works the way a developer would: it reads the relevant files, makes edits, runs commands to check its work, and shows you every change in the conversation as it happens.

---

## Projects, Conversations, and Working Copies

Neotask Code is organized around **projects**. A project is a self-contained workspace on your computer that holds one codebase and everything the agent needs to work on it.

- **Project**: a single coding workspace. Each project lives in its own folder that Neotask manages, and the agent is confined to that folder. You can create a project from an empty folder, from a local folder you already have, or from a GitHub repository.
- **Conversation**: the chat thread where you talk to the agent about a project. Each turn you send is a coding request; the agent responds by editing files and running commands, and the whole transcript stays with the project so you can follow what was done and pick up where you left off.
- **Working copies (worktrees)**: within a project you can create additional working copies so you can keep more than one branch checked out at the same time, side by side, without disturbing your main working copy. This is useful when you want to try an idea on a separate branch while your main copy stays untouched. Working copies can be created and removed as you need them.
- **Checkpoints**: Neotask Code can take a checkpoint of the project so you have a known-good point to return to. If a change goes in a direction you do not want, you can restore the project to an earlier checkpoint. See [Neotask Code: Safety and Approvals](neotask-code-safety.md) for how checkpoints fit into the safety model.

Each project keeps its own history, its own connected repository, and its own settings, so work on one project never leaks into another.

A conversation also remembers which repository, branch, and working copy it belongs to, so a restart does not move your work. Every turn runs inside that session's working copy. If the working copy folder is missing, Neotask repairs it before running, or stops with a clear message if it cannot. Cleanup never deletes uncommitted work, and a development server from an earlier run is stopped or adopted safely instead of being left holding the port. Whether a coding session reopens by itself after a restart is controlled by **Continue work after restarts** in **Settings** -> **Automations**, which is on by default. See [Continue Work After Restarts](continue-work-after-restarts.md).

---

## Choosing a Coding Engine

Neotask Code can run on more than one underlying coding engine. The engine is the agent loop that actually does the work; you pick one per project, and you can see a short description of each option where you choose it. What changes between engines is which account and models the work runs on.

| Engine                      | Runs on                               | Models                                   | Billing                                                                             |
| --------------------------- | ------------------------------------- | ---------------------------------------- | ----------------------------------------------------------------------------------- |
| **Neotask Code (built-in)** | Neotask, through your current sign-in | Any model eligible in your plan and mode | Your Neotask credits, or your connected provider login, or your own key (see below) |
| **Claude Code**             | Your connected Claude account         | Claude models from your account          | Your own Claude subscription                                                        |
| **Codex**                   | Your connected OpenAI account         | OpenAI models from your account          | Your own OpenAI subscription                                                        |
| **Cursor**                  | Your own Cursor account               | Models offered by Cursor                 | Your Cursor subscription (never uses Neotask credits)                               |

The built-in **Neotask Code** engine is the default and the simplest to start with: it works with whatever sign-in you already use for the rest of Neotask, and you can pick from the models your plan makes available. The other engines let you bring an outside coding tool you already pay for and run it through Neotask instead of a separate window, so your projects, approvals, and git connections all stay in one place. Choosing one of them runs your work through that tool on your existing account for it; you do not create a second, separate Neotask login.

### How billing works per engine

- **Built-in Neotask Code** follows the same billing as the rest of the app. Depending on how your workspace is set up, that is your Neotask credits, a coding provider you have signed into, or your own API key if you use Bring Your Own Key. See [Billing and Plans](billing.md), [Usage Rates](usage-rates.md), and [Bring Your Own Key](byok.md).
- **Claude Code, Codex, and Cursor** run on the outside account you sign into for each of them. Their usage is billed by that vendor to your account with them, on their models, not through Neotask credits.

You can change which engine a project uses; starting a new engine begins a fresh conversation context for that project.

---

## Getting Started

1. Open the coding area in the Neotask desktop app.
2. Create a project: start from an empty folder, open a folder you already have, or connect a GitHub repository. See [Neotask Code: Git and GitHub](neotask-code-git.md) for the repository options.
3. Pick a coding engine (or keep the default built-in engine) and, for the built-in engine, choose a model.
4. Describe what you want in the conversation. The agent reads your project, makes changes, and runs commands, showing each step as it goes.
5. Review the changes. When you are ready, ask it to commit and push, or open a pull request, if you have connected a repository.

The activity feed keeps source changes and verification output with the coding conversation.

![Code work with source changes and verification](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/work-code-1280.webp)

Coding is available on the Free plan within the Free message quota, and without limit on paid plans. See [Getting Started](getting-started.md) and [Free Plan](free-plan.md) for what each plan includes.

---

## Availability

- **Built-in Neotask Code, Claude Code, Codex, and Cursor engines** are available today.
- **Multi-repository projects** (one project that holds several connected repositories at once, so a single conversation can work across, for example, a server and its client together) are shipping soon. Today each project connects to one repository.

---

## Related

- [Continue Work After Restarts](continue-work-after-restarts.md)
- [Custom Skills And Plugins](custom-skills-and-plugins.md)
- [Import From Other AI Apps](/docs/neotask-desktop/external-agent-import)
- [Neotask Code: Git and GitHub](neotask-code-git.md)
- [Neotask Code: Safety and Approvals](neotask-code-safety.md)
- [Billing and Plans](billing.md)
- [Bring Your Own Key](byok.md)
- [Security](security.md)

---

### [Neotask Code: Git and GitHub](https://neotask.ai/docs/neotask-desktop/neotask-code-git)

## Review Before Commit Or Pull Request

1. Open the coding project.
2. Select **Changed files**.
3. Review the source diff and verification results.
4. Confirm the branch and repository.
5. Commit, push, or open the pull request after the review is complete.

![Coding workspace with repository changes](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/code-main-1280.webp)

Neotask Code does real version control. It works with git the way a developer would: it can connect to a GitHub repository, create branches, commit and push your work, and open pull requests, all from inside the coding conversation. This page covers how to connect an account, how each repository option works, and what the agent can do with git once you are connected.

For the wider picture, see [Neotask Code](neotask-code.md). For how git actions are gated and kept safe, see [Neotask Code: Safety and Approvals](neotask-code-safety.md).

---

## Connecting a GitHub Account

Neotask Code needs a GitHub account before it can clone private repositories, push, or open pull requests. There are two ways to connect one.

- **Sign in through Neotask.** Use the guided GitHub sign-in in the coding area. This opens GitHub in your browser, you approve access, and Neotask stores the connection securely. This is the simplest path if you do not already use the GitHub command line.
- **Use a GitHub account already on your computer.** If you have signed into GitHub with the GitHub CLI on your machine, Neotask can detect those accounts and use one of them. When more than one account is available, you choose which one to use, and Neotask tells you why an account can or cannot be used.

You can connect an account at two levels:

- **For your main agent or workspace**, so it is available across your coding projects.
- **Per project**, so a specific project pushes as a specific account. This is useful when different projects belong to different GitHub identities.

Your GitHub credentials are stored encrypted, and each project uses only the account you connected for it. Neotask never switches your machine's default GitHub account behind your back; it uses the specific account you chose for each action.

---

## Starting a Project From a Repository

When you create a coding project, you have three repository options:

| Option | What happens |
| --- | --- |
| **Connect an existing repository** | Neotask clones a GitHub repository you already have into a new managed project workspace, so the agent can work on it. |
| **Publish a new repository** | Start from an empty or local project, then create a brand-new GitHub repository from it and push the first commit. |
| **Work locally with no repository** | Start from a local folder with no GitHub connection. You can connect or publish a repository later. |

When you connect an existing repository, Neotask can look it up by choosing from your account's repositories or by pasting its URL, and it reads the available branches so you can pick which one to start on.

---

## Branches and Working Copies

- **Branches.** The agent can create a new branch and switch the working copy to a different branch. Where you choose a branch, Neotask shows the branches that already exist rather than asking you to type a name, so you do not push to the wrong place by accident.
- **Working copies (worktrees).** Within a project you can create additional working copies, each checked out to its own branch, so you can keep more than one line of work going at the same time without disturbing your main copy. Working copies can be created and removed as needed.

### After A Restart

Each coding session stays bound to its own repository, branch, and working copy, so an app update, an operating-system restart, a crash, or a power loss cannot move your work to the wrong branch.

- Every turn runs inside that session's working copy.
- If the working copy folder is missing, Neotask repairs it before running. If it cannot repair it, it stops with a clear message instead of running somewhere else.
- Working copy registrations that no longer point at a real folder are listed for you instead of hidden.
- Cleanup never deletes uncommitted work.
- A development server from a previous run is stopped or adopted safely. It is never left holding the port.

Whether the session reopens on its own is controlled by **Continue work after restarts** in **Settings** -> **Automations**, which is on by default. See [Continue Work After Restarts](continue-work-after-restarts.md).

---

## Committing, Pushing, and Pull Requests

Review the verified source changes in the activity feed before creating the commit or pull request.

![Verified code activity and source changes](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/work-code-1280.webp)

Once a repository is connected, you can ask the agent to handle version control as part of the conversation. Supported actions include:

- **Commit** the current changes with a message.
- **Push** commits to the connected repository.
- **Commit and push** in one step.
- **Open a pull request**, including commit, push, and pull request together in one step.
- **Pull** the latest changes from the repository.
- **Publish a new repository** from a local project.

You do not have to leave Neotask to run these; you ask for them in plain language and review what the agent proposes. Reading history and status is immediate, while actions that change your repository or push to GitHub ask for your approval first. See [Neotask Code: Safety and Approvals](neotask-code-safety.md) for exactly which git actions run freely and which ask.

### Commit identity

Commits the agent makes on your behalf are attributed clearly so your history stays honest. Commits you write yourself are left untouched, and every push uses the specific GitHub account you connected for that project, so your work is pushed under the identity you expect.

---

## Multi-Repository Projects (Shipping Soon)

Today a project connects to one repository. Support for a single project that holds several connected repositories at once, so one coding conversation can work across, for example, a backend and a frontend together while keeping each repository's branches, accounts, and history separate, is shipping soon. Until then, use a separate project per repository.

---

## Related

- [Neotask Code](neotask-code.md)
- [Continue Work After Restarts](continue-work-after-restarts.md)
- [Neotask Code: Safety and Approvals](neotask-code-safety.md)
- [Security](security.md)
- [Integrations](integrations.md)

---

### [Neotask Code: Safety and Approvals](https://neotask.ai/docs/neotask-desktop/neotask-code-safety)

## Review A Sensitive Command

1. Read the command and the reason it was flagged.
2. Check the project and working directory.
3. Approve only when the target and effect are correct.
4. Deny the request when the scope is broader than intended.

![Destructive command guard with the requested command and review context](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/destructive-guard-1280.webp)

Neotask Code is designed to move fast on the routine work of coding while stopping to ask you before anything with real consequences. It does this with a clear rule: safe, everyday actions inside your project run on their own, and anything risky, anything that reaches outside your project, or anything the system cannot confidently classify pauses for your approval. The default is to ask when in doubt.

This page explains what runs freely, what asks first, and the boundaries that keep the agent contained. For the overall feature, see [Neotask Code](neotask-code.md).

---

## What Runs Freely, and What Asks First

A coding conversation is something you start and watch. So that it does not stall on every small step, Neotask Code lets the ordinary parts of coding run without an approval prompt, while keeping a gate in front of everything that could affect systems or data outside the immediate work.

**Runs on its own, inside your project workspace:**

- Reading, searching, and listing files in the project.
- Writing and editing files in the project.
- Read-only inspection commands (for example listing files or printing file contents).
- Common build, test, and lint commands, so the agent can check its own work as it goes.

**Asks for your approval first:**

- Git and GitHub actions that change your repository or push, such as pushing, publishing a repository, or opening a pull request. Force and destructive git operations require an explicit, higher-level approval, and some especially dangerous git overrides are blocked outright.
- Network access to anything outside your project.
- Installing or publishing packages.
- Destructive operations.
- Any command that tries to reach a file outside the project workspace.
- Anything the system cannot confidently classify as safe.

This last point matters: the system fails closed. If a command does not clearly fall into the safe set, it is treated as needing approval rather than being allowed through.

---

## Destructive Command Guard (Shipping Soon)

Beyond the approval rules above, Neotask Code is adding a dedicated guard for catastrophic, irreversible commands. It sits in front of your approval settings, not inside them: certain classes of command are stopped before they can ever be offered as something to approve.

**Blocked outright, in every mode, with no approval that can allow them:**

- Deleting files or folders outside your project workspace, anywhere on your computer.
- Erasing or reformatting a disk.
- On Windows, the equivalent system-destroying commands, such as deleting shadow copies or formatting a drive.
- Destructive commands hidden inside a script or an inline command, not just ones typed directly.

**Escalates to your approval settings instead of running:**

- Dropping or wiping a database.
- Tearing down cloud infrastructure, containers, or clusters.
- Deleting secrets or credentials.

These are serious actions that are sometimes exactly what you intend, so rather than blocking them outright, the guard routes them to your normal approval flow so you can confirm before anything happens.

**Always on in Fully Autonomous mode.** Even when a company runs with the least approval friction, the guard's outright-blocked commands stay blocked. No autonomy setting, and no always-approve rule, can allow a catastrophic command to run unattended. See [Approvals and Safety](auto-companies-approvals-and-safety.md) for how autonomy modes work.

**Never silent.** Whenever the guard stops a command, it always shows up as a visible, labeled entry in the conversation, explaining what was blocked and why. A blocked command is never dropped quietly in the background.

This capability is shipping soon. Until it ships, the approval rules described above are what govern destructive and out-of-workspace commands.

---

## Always Approve

When an approval appears, you can approve it once, or choose to always approve that kind of action so similar actions in the future run without prompting you again. This lets you keep tight control at first and loosen it for the specific actions you have come to trust, without turning off oversight everywhere. Approvals wait for you; they do not silently expire or auto-decide on your behalf.

---

## How Coding Fits Your Autonomy Settings

The approval detail names the app, scope, owner, and review window before the action is released.

![Approval detail for a guarded action](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-detail-1280.webp)

Coding respects the same approval and autonomy controls as the rest of Neotask. The approval settings that govern how much the AI can do on its own, from fully supervised to more autonomous, apply to coding work too, so coding does not become a way around the oversight you have set elsewhere. See [Approvals and Safety](auto-companies-approvals-and-safety.md) for how those approval modes work across the platform.

---

## The Agent Stays Inside the Project

Every coding project is a contained workspace on your computer, and the agent cannot touch files outside it.

- File reads, writes, and edits are confined to the project's own folder.
- Any command argument that points at a path outside the project workspace is stopped and sent for approval rather than being run.
- Attempts to escape the workspace, for example by referencing a parent directory or an absolute path outside the project, are caught before they run.

The result is that a coding conversation can only affect the one project you are working on. It cannot wander into the rest of your files.

---

## Browser Access Is Limited to Your Own Dev Server

When a project runs a local development server, the agent can open a browser view of that server to check its own work, for example to see the page it just built. That browser access is deliberately limited:

- It can open your project's own development server on your computer (a local loopback address, on any port) or simply look at the page already open.
- It cannot navigate to an external website on its own. Any attempt to visit a different, outside address pauses for your approval first.

So the browser is there to preview what the agent is building, not to browse the open internet unsupervised.

---

## Version Control Is Your Safety Net

Beyond approvals, git and checkpoints give you a way back.

- **Checkpoints.** Neotask Code can capture a checkpoint of the project as a known-good point. If a change goes somewhere you do not want, you can restore the project to an earlier checkpoint.
- **Commits and branches.** Because the agent works with real git, your committed history is a durable record, and you can keep experimental work on a separate branch or working copy so your main line of work stays clean. See [Neotask Code: Git and GitHub](neotask-code-git.md).

---

## Your Push Identity Stays Yours

Pushes and pull requests use the specific GitHub account you connected for that project, and commits the agent makes on your behalf are attributed clearly. Neotask does not change your machine's default GitHub account behind your back, and it does not push under an identity you did not choose. See [Neotask Code: Git and GitHub](neotask-code-git.md) for how accounts are connected.

---

## Related

- [Neotask Code](neotask-code.md)
- [Neotask Code: Git and GitHub](neotask-code-git.md)
- [Approvals and Safety](auto-companies-approvals-and-safety.md)
- [Security](security.md)
- [Trust and Security](trust-and-security.md)

---

### [Neotask Payments](https://neotask.ai/docs/neotask-desktop/neotask-payments)

## Connect And Review Payments

1. Open the company and select **Payments**.
2. Choose **Set up payments** when no account is connected.
3. Complete the provider setup and return to Neotask.
4. Confirm the connected state before accepting payments.
5. Use the overview and ledger to trace charges and payouts.

![Payments setup and account status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/enterprise/2026-08-12-r4/payments-overview-1440.webp)

**Neotask Payments** is the built-in way for your autonomous company to get paid. Sign up once, and your company can accept money from customers with payouts going straight to your own bank account.

---

## What Neotask Payments Is

When Neotask builds and runs a company for you, that company often needs to **sell something**, a product, a service, a subscription. Neotask Payments is the rail that lets it actually collect that money.

In plain terms:

- Your customers pay your company.
- Neotask Payments handles the **checkout** and processing.
- The money is paid out to **your company's own bank account**.
- Neotask keeps the books so everything is tracked.

It is designed so you do almost nothing. You connect your payment account once during setup, and from then on your agents handle the rest.

---

## Who It's For

Neotask Payments is for **build-mode companies**, companies that Neotask creates and operates for you (as opposed to companies where Neotask only manages something you already run). If your company is meant to sell to customers and receive income, this is the feature that makes that possible.

If your company doesn't take payments, you can simply skip it.

---

## How To Connect A Payment Account

There are two places you can set this up. Both use the **same** secure connect flow, so it doesn't matter which one you use.

### During onboarding (the Payments step)

When you create a build-mode company, onboarding includes a **Payments** step:

1. You'll see the **Neotask Payments** card with a **Set up Neotask Payments** button.
2. Clicking it opens a **secure setup window** where you provide the details needed to verify your business and add the bank account you want to be paid into.
3. When you finish those hosted steps, Neotask **detects it automatically**, you don't have to come back and click anything in most cases. There's also an **"I've finished, check status"** button if you'd like to refresh it yourself.

### Later, in Settings → Payments

Every build-mode company has a **Payments** section in its **Settings** tab. It shows the same Neotask Payments card with a **Set up Neotask Payments** (or **Resume setup**) button. Use it any time after onboarding to connect, finish, or review your payment account.

---

## What "Connected" Means

Connecting isn't just "I clicked the button." Your account is only marked **Connected** once it can genuinely take and move money. Behind the scenes there are two things that have to be switched on:

| Capability | What it means |
|------------|----------------|
| **Charges enabled** | Your company can accept payments from customers. |
| **Payouts enabled** | Money your company earns can be paid out to your bank. |

Only when **both** are on does Neotask show the green **Connected** pill. Until then, your company can't collect payments yet.

Your connection can be in one of a few states:

| Status | What it means for you |
|--------|------------------------|
| **Not started** | You haven't begun setup yet. |
| **In progress** | Setup is started but the verification steps aren't finished. Finish them in the secure window. |
| **Pending** | You submitted your details; they're being reviewed. |
| **Connected** | Everything is on, your company can accept payments and receive payouts. |
| **Needs attention** | Something is blocking your account (e.g. more information is required) and needs you to resolve it. |
| **Disabled** | The payment account was disconnected. |

The status you see always reflects the **real, verified state**, Neotask reads it from the source of truth, never a hopeful local guess. While setup is in progress, Neotask quietly re-checks for you so the moment you're done, the card flips to **Connected**.

---

## How Money Flows To Your Bank

Once you're connected, the flow is simple:

1. A customer pays your company.
2. The payment is processed through Neotask Payments.
3. **The money belongs to your company** and is paid out to the bank account you added during setup.

Your earnings land in **your own account**, Neotask is the rail, not the destination.

---

## The Platform Fee

Open **Transactions** to trace the payment, fee, and net amount recorded for each charge.

![Payments activity and recorded charges](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/ledger-payments-1280.webp)

Neotask charges a **15% platform fee** on payments your company collects through Neotask Payments. Concretely:

- Your company keeps **85%** of each payment.
- Neotask's **15%** fee is taken from the payment automatically.

A couple of things worth knowing:

- The 15% fee is **separate from** the standard payment-processing fee that any card payment incurs.
- The fee is calculated on the **gross amount** of each payment and recorded for you, so there's a clear trail of what was collected, what the fee was, and what was paid through to your company.

---

## Managing Or Disconnecting

You manage Neotask Payments from the company's **Settings → Payments** section at any time:

- **Review your status**, see whether you're Connected and whether charges/payouts are enabled.
- **Finish or resume setup**, if you started but didn't complete verification, pick it back up here.
- **Update your details**, re-open the secure setup window to change business or bank information.

If a payment account is ever **disconnected** or **disabled**, the card reflects that, and your company stops being able to take payments until it's reconnected.

---

## You Can Set This Up Later

Neotask Payments is **deferrable**, you don't have to finish it during onboarding. In the Payments step you'll see a **"Set up later"** option. If you choose it:

- Onboarding stops blocking on the Payments step, so you can finish creating your company.
- **Your company can't accept payments until Neotask Payments is connected.**
- Any **payment-related tasks** your company would do are kept as **drafts**, they won't run for real until you've connected.

When you're ready, just go to **Settings → Payments** and complete setup. Choosing "Set up later" never means you're connected; it only postpones the step.

---

## Related

- [Agent Spending](agent-payments)
- [Billing & Plans](billing)
- [Company Modes](company-modes)
- [Auto, Company Dashboard](auto-companies-dashboard)
- [Subprocessors](subprocessors)

---

### [Nightly Rundown](https://neotask.ai/docs/neotask-desktop/nightly-rundown)

## Review Tomorrow's Work

1. Open **Nightly Rundown** from the company workspace.
2. Review completed work, unresolved items, and suggested tasks.
3. Approve or edit work that requires a person.
4. Open **Plan** to see the resulting schedule.

![Nightly Rundown summary and suggested work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/nightly-summary-1280.webp)

The **Nightly Rundown** is your company's autonomous end-of-day engine. Every night it runs on its own, reviews everything your company did, actively checks your live tools and tomorrow's readiness, **plans and (when you allow it) schedules the next day's work**, then delivers the whole briefing to you by email and in your chosen channel, with one place to approve anything extra it found.

It's not just a summary screen you open. It's a per-company job that fires on a schedule, thinks about tomorrow, turns its best ideas into real tasks (capped and gated to your comfort level), and even assigns *you* the steps only a human can do, so your company keeps moving overnight and you wake up to a clear, actionable plan.

> **Note:** The Nightly Rundown is part of a feature rollout and is controlled by a per-company setting, so it may not be switched on for every company yet. Everything below describes how it behaves when it's enabled.

---

## How It Works at a Glance

1. **It runs itself, nightly.** Each company gets its own nightly schedule (default **10:00 PM in the company's local time**, configurable). The job fires on that schedule every night, including weekends, without you opening anything and without slowing down your company's other work.
2. **It reviews today + inspects your live tools.** It gathers your real last-24-hours activity and runs a read-only AI review that actively probes your connected apps, payments, email, and integration health to catch fresh signals.
3. **It plans tomorrow.** A mini planner turns the review's findings into a capped set of real next-day tasks, sequenced around your fixed plan and your company's mode.
4. **It respects your autonomy level.** On autopilot it auto-approves up to your nightly cap and schedules the work; in supervised mode it queues those tasks as proposals for you to approve.
5. **It assigns you the human steps.** When something genuinely needs a person, connect an app, upload a file, give feedback, ship a physical item, it creates a task for *you* that holds the dependent agent work until you've done it.
6. **It delivers + saves.** It saves the rundown to your account (the database is the source of truth), emails you the briefing, and can post it, with interactive approve/deny buttons, into a connected Slack, Teams, or Discord channel.
7. **It's instant when you open it.** Opening the Nightly Rundown shows the already-generated briefing immediately, no waiting, no errors.

---

## When It Runs

The Nightly Rundown runs on a **dedicated per-company nightly schedule**, by default **10:00 PM in the company's local timezone**, and you can change the time (see *Turning It On & Setting the Time*).

- **On its own schedule, every night.** A real scheduled job fires nightly in your company's local time. It runs even on quiet days and weekends, so your company keeps up with support email, payments, light content, and product work.
- **Spread across the fleet.** Nightly jobs are deterministically staggered across the window (and naturally spread across each company's local timezone), so nothing piles up at the same instant.
- **Off the hot path.** The heavy review + planning runs in the background, separate from your company's normal task loop, a long nightly run never delays your scheduled tasks.
- **Instant on open.** Clicking the **Nightly Rundown** button on the company dashboard opens that night's already-generated briefing immediately. If a rundown hasn't been generated yet, you see the day's summary plus a *"generating"* state while it's prepared, never a "not ready" error and never a failed screen.
- **Once per day.** The rundown is built once per upcoming business day. Reopening it shows the same review for tomorrow rather than regenerating, and a retry for the same day never duplicates tasks or re-sends your email.

---

## What It Reviews

Before it shows you anything, the Nightly Rundown gathers your company's **real activity**, it never invents a summary. The end-of-day digest reads the **last 24 hours**; the planner additionally considers a **configurable lookback window** (see *Lookback Window* below) so its recommendations are grounded in the right amount of history.

| It looks at | So it can show you |
|---|---|
| **Tasks that ran** (from your authoritative task-run history) | How many tasks succeeded and how many failed |
| **Files & documents created** | What your company actually produced today |
| **Approvals processed** + items still waiting | What's been decided and what still needs you |
| **Requires-action items** | Anything blocking a task until you act |
| **Today's activity & errors** | Highlights worth knowing and problems to look at |
| **Your live schedule** (real scheduled jobs) | Exactly what is set to run tomorrow |
| **Connected vs. disconnected integrations** | Apps that are broken or expired and could stall tomorrow's work |
| **Your active plan** | The fixed work tomorrow is built around |
| **Billing context** | Today's task-fee spend, your credit balance, and any tasks blocked for balance |

Every billing read is safeguarded, if it can't be loaded, the rundown still shows everything else rather than failing.

### Active Live-Tool Inspection

Beyond reading your stored activity, the nightly review **actively inspects your live tools each night, read-only**, it looks at things like Gmail, your payments/Stripe state, your connected MCP apps, and overall integration health to surface fresh signals (a customer email that needs a reply, a payment that needs attention, an integration that quietly expired). These probes are **strictly read-only**, the review can look, but it never sends, charges, or changes anything on its own. If a probe fails (for example, a disconnected integration), that becomes a **visible integration-health finding** in your rundown rather than being silently dropped.

---

## Lookback Window

You choose how much history the nightly planner weighs when it reasons about tomorrow:

| Preset | Roughly how far back | Best for |
|---|---|---|
| **Daily** | ~3 days | Fast-moving companies that want tight, recent-focused planning |
| **Weekly** *(default)* | ~2 weeks | The balanced default, enough context without over-reaching |
| **Monthly** | ~1 month | Slower cycles where monthly patterns matter |
| **Quarterly** | ~1 quarter | Big-picture, longer-horizon reasoning |

The window is always bounded, so even a quarterly lookback stays fast and never dumps unbounded history. The day's *summary* is always the last 24 hours, the lookback only widens what the *planner* considers.

---

## The Next-Day Planner

Open **Tasks** after the planner finishes to review the resulting work and the reason each task was added, continued, changed, or dropped.

![Next week replanned in the task workspace](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-11-r1/hero-week-replanned-1280.webp)

This is what makes the Nightly Rundown an engine rather than a digest. After the review, a **mini planner** proposes net-new work worth doing that isn't already in your fixed plan, and, when you allow it, **turns those proposals into real next-day tasks** that are scheduled to run tomorrow.

- **Capped on purpose.** It creates at most your **maximum dynamic tasks per night** (default **3**, adjustable). If the review surfaces more ideas than the cap, it keeps the highest-priority ones. Each task costs roughly a normal task run, so the cap keeps spend predictable (≈$3/night at the default).
- **Mode-aware.** A build-mode company's next-day tasks lean toward building toward revenue; a manage-mode company's lean toward operations. Work that isn't ready yet (for example, selling before the product is ready) is held behind its prerequisites instead of firing prematurely.
- **Budget-aware & idempotent.** It respects your budget, and a re-run for the same day reconciles to the same set, it never double-creates tomorrow's tasks.

This replaces the old behavior where suggestions were only recommendations that never became work.

---

## How Generated Tasks Are Gated (Your Autonomy Level)

The next-day tasks follow the **same autonomy tier and approval system** the rest of your company uses, there's no separate, parallel approval path:

- **Autopilot / full-autonomy company:** the nightly tasks (up to your cap) are **auto-approved and scheduled** to run tomorrow.
- **Supervised / balanced company:** each nightly task becomes a **pending proposal** in your approvals queue. Nothing runs until you approve it, and approvals **never expire**, so nothing is lost if you wait.

You stay in control of the comfort level; the nightly engine simply plugs into it.

---

## Tasks the AI Assigns to *You* (Human-Gated Work)

Some essential steps can only be done by a person, connecting an account, uploading a deliverable or file, giving a decision/feedback, or a real-world action like shipping a physical item. The Nightly Rundown can **detect when a human is genuinely required** and create a task **for you**, then **hold the dependent agent work behind it** until you've completed it, **even in full autopilot**.

- The human step shows up as a clear to-do (connect / provide / confirm / do).
- The downstream agent task that depends on it won't run until you finish the human step.
- Completing your step **releases** the dependent agent task on the next cycle.

This way the AI never silently stalls or pretends a blocked-on-you task is done, it asks you for exactly what it needs and keeps everything else moving. (This shares the same human-task system the weekly planner uses, so your human to-dos live in one place.)

---

## Delivery: Email + Your Channel

When the rundown is generated and saved, it reaches you two ways:

- **The morning email.** Saving the rundown triggers a single owner email, your end-of-day/morning briefing, sent once per night for that day (a retry won't double-send). It uses your existing email path; no new third party is involved.
- **Interactive channel delivery.** If you've pointed the rundown at a connected channel (Slack, Microsoft Teams, or Discord), it posts the briefing there with **interactive Approve / Deny / Always-Approve buttons** on each proposed task. You can:
  - **Route by channel or by user**, send the rundown (and its approvals) to a shared channel or directly to a specific person.
  - **Control who can approve**, a per-user permission check means only authorized people can act on the buttons; an unauthorized press is refused.
  - **Approve right from the message**, a button press resolves the underlying approval through the normal approval system; **Always-Approve** turns a recurring yes into a standing rule so you stop being asked.

Everything stays consistent: the email, the channel message, and the in-app modal show the same plan and the same approvals.

---

## What You See in the App

Opening the Nightly Rundown shows a single, unified next-day plan with clear sections:

- **Today's Summary**, Tasks, Failed, Pending, and Documents counts, plus highlight chips (e.g. *"12 tasks completed," "3 files created," "$0.42 in task fees"*) and watch-out chips (e.g. *"2 tasks failed or blocked," "1 action item needs attention"*).
- **Overview**, Tasks Run, Scheduled, Setup Needed, and Approvals at a glance.
- **Tomorrow's Plan & Readiness**, one combined list of everything set to run tomorrow: your **fixed scheduled tasks**, the **new dynamic tasks** the planner created, and any **human-gated tasks** (shown as blocked-on-you), each with a color-coded readiness label. Click any row to jump straight to that task.
- **Requires Your Action**, tasks paused until you do something (connect an app, upload a file, make a decision).
- **Pending Approvals**, approve or deny each one, or **Approve All / Deny All** at once. Approvals never expire.
- **Gateway Review**, the AI's written summary of the day, with **Highlights** and **Watchouts**.
- **Suggested Tomorrow Work**, cards for extra ideas surfaced outside your fixed plan (see below).
- **Looks Good**, dismiss the rundown when you're done.

The header shows the date, the company name, and which AI model produced the review.

---

## Tomorrow's Readiness Labels

Each task set to run tomorrow gets a label so you know, at a glance, whether it will just run or wait for you:

| Label | What it means |
|---|---|
| **Auto** | Runs without asking. The company (or this task) is set to auto-approve, or the task is low-risk enough to run on its own. |
| **Scheduled** | Will run as planned, no approval needed. |
| **Likely Approval** | Will probably ask for your OK first, the company is in a supervised mode, or the task is higher-risk or set to always require approval. |
| **Pending Approval** | An approval is **already waiting** for this task right now. |
| **Blocked on You** | A human step must be completed first (connect / upload / decide / do) before the dependent task can run. |
| **Ready / Setup Needed / Blocked** | Runtime states, ready to go, needs an app connected first, or currently blocked. |

These labels are derived from your real autonomy settings, live approval records, and human-task gating, so they reflect what will actually happen, not a guess.

---

## Suggested Tomorrow Work

Alongside the tasks it schedules, the review also surfaces ideas as cards, **recommendations that explain *why*** they're worth doing, how likely they are to need approval, and which screen to open to act on them. Suggestions come in seven kinds:

| Type | What it suggests |
|---|---|
| **Plan gap** | Useful work your fixed plan missed |
| **Follow-up** | A natural next step to something that already happened |
| **Integration fix** | A disconnected or broken app to reconnect before it blocks tomorrow's work |
| **Analysis** | A one-off analytical or diagnostic task |
| **Content opportunity** | A content or SEO opportunity to capitalize on |
| **Approval follow-up** | A recurring approval pattern that could become a standing rule |
| **Risk review** | A prompt to investigate failed, blocked, or error events |

Each carries a **priority** (low / medium / high) and an **approval likelihood** (none / maybe / likely) so you can triage fast.

---

## How It's Generated

The Nightly Rundown is produced by a **real, read-only AI review** running on Neotask's gateway, on the nightly schedule. The review can inspect your live tools, connected apps, and activity, but it **never changes anything or runs a task on its own**, only the planner (gated by your autonomy level) schedules work. It returns a structured summary, highlights, watchouts, and a prioritized set of suggested-tomorrow-work items, and it self-corrects if its first draft comes back malformed.

The review uses whichever AI model your company is set to use. Failure is always **visible, never silent**: if a review comes back empty or unusable, the engine reports it and **leaves your last good rundown intact**, it never overwrites a working briefing with an empty one. Each rundown is saved to your account (the database is the source of truth), so it's consistent across devices and kept as a record.

---

## Turning It On & Setting the Time

The Nightly Rundown is configured in **Notification Preferences** for each company:

- **Nightly Rundown toggle**, turns the nightly engine on or off for the company.
- **Time**, when it runs each night (default **10:00 PM**, company-local).
- **Timezone**, the timezone the schedule fires in (defaults to the company's profile timezone).
- **Max dynamic tasks per night**, the cap on next-day tasks the planner may create (default **3**).
- **Lookback preset**, Daily / Weekly *(default)* / Monthly / Quarterly.
- **Review depth**, how actively the nightly review probes your live tools.
- **Channel delivery**, which connected channel (and which users) the rundown and its approvals are sent to, plus who's allowed to approve.

> *"Receive a nightly summary of all activity, tasks run, and pending items, and let Neotask plan and tee up tomorrow's work."*

Every setting is saved to your account (the database is the source of truth), so it follows you to any device. When the engine is off, no nightly job runs and nothing is scheduled, though you can still open the **Nightly Rundown** button on demand.

---

## Related

- [Approvals & Safety](./auto-companies-approvals-and-safety.md), how approvals, autonomy modes, and risk levels decide what runs on its own
- [Planning & Weekly Updates](./planning.md), the fixed weekly plan the nightly engine checks readiness against and complements
- [Company Modes](./company-modes.md), build vs. manage, which shapes the next-day plan
- [Agent Spending](./agent-payments.md), task fees, budgets, and the per-night cap
- [Channels](./channels.md), connecting Slack, Teams, or Discord for interactive delivery
- [Company Dashboard](./auto-companies-dashboard.md), where the Nightly Rundown button lives

---

### [Planning & Weekly Updates](https://neotask.ai/docs/neotask-desktop/planning)

## Review The Plan And Schedule

1. Open the company and select **Plan**.
2. Review the generated tasks, recurring work, and quiet hours.

![Company Plan showing the active schedule, recurring work, quiet hours, and operating controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v6/2026-08-11-r1/ui/active-plan-1280.webp)

3. Select **Tasks**, then open a task to inspect its definition and run history.

![Company Tasks showing recurring work with schedules, connected apps, run status, and spend](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-11-r1/tasks-recurring-1280.webp)

![Opened company task showing its integrations, skills, schedule, status, and completed run](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r6/company-task-run-detail-1280.webp)

4. Return after the weekly refresh to review added, continued, and dropped work.

Your company starts with a full week of work planned for it, and that plan **refreshes itself every week** so the company keeps making forward progress instead of going quiet after the first few days. This guide explains how the first plan is created, how the weekly refresh works, and where you can see exactly when the next refresh will happen.

> **Note:** The automatic weekly refresh is part of a feature rollout, so it may not be switched on for every company yet. Everything below describes how planning behaves when the weekly refresh is active.

---

## The First Week's Plan

When you finish onboarding a new company, Neotask immediately writes its **first week of work** as a real schedule of tasks, not a vague to-do list. The plan is built specifically for a brand-new, pre-launch company, so every task can be done right away with the built-in tools, without you having to connect anything first.

A few things are true about that first plan by design:

- It is **one week of forward work**, not a permanent loop. Most tasks happen once; at most a couple repeat (like a single weekly blog post).
- The very first task is to **write and publish your company's first blog post**, and it is set to run right away rather than waiting for a scheduled day.
- The rest of the tasks are spread across **Monday through Friday**, each with a specific day and time, so the week is paced rather than stacked on one day.
- The week is **sequenced to compound**, research first, then drafting, then producing, then preparing to launch, so each day builds on the one before it.

---

## "Your Website Already Exists", The Planner Doesn't Rebuild What's Done

A common failure for an autonomous company is to keep re-doing the same setup work. Neotask's planner avoids this by being **grounded in what your company has already shipped** before it writes a single task.

Every generated company already has a live website that includes:

| Already exists | What it means for planning |
|---|---|
| A **landing page** (hero + content sections) | The planner extends it, never rebuilds it |
| A **working email-capture waitlist** | A real, functioning signup form that is already collecting people, the planner grows it, it does not re-create it |
| A **blog** | The publishing channel, the planner adds posts, it does not "build a blog" |

On top of that, a fixed set of starter deliverables runs automatically (your first blog post, an initial website SEO/copy pass, starter brand assets, a first content batch, a prospect list, and a first briefing). The planner is explicitly told **not to repeat any of these**.

So instead of re-creating finished work, the planner authors genuinely **new, additive** work: a new page or section the site doesn't have yet (an FAQ, a docs page, a case-studies page), defining the offer and pricing, drafting foundational documents, deepening content, growing the existing waitlist, and preparing the launch.

---

## How The Weekly Refresh Works

The most important part of the system is that **the plan does not run out**. Once a week, Neotask reviews the company and writes a **fresh, revised week of real scheduled tasks** to replace the old one.

This weekly review is not a note-taking exercise. It produces actual tasks that update your **task list**, your **schedule view**, and your **plan**, the same way the original plan did, with each task armed to run on its day and time.

After the review, open **Tasks** to see which work was added, continued, changed, or dropped. That record explains how the current week differs from the prior plan.

![Week replanned with added, continued, and dropped work](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v5/2026-08-11-r1/hero-week-replanned-1280.webp)

### What The Weekly Review Looks At

Before writing next week, the review grounds itself in your company's **real current state**, it never invents a week from scratch:

| It reads | Why |
|---|---|
| **What actually ran** this past week | To see which tasks succeeded, which failed, and what produced results |
| The company's **memory** | To know what has been learned and recorded |
| The **live site** as it exists now | To know what has already shipped and what is already collecting signups |

From that, it **doubles down on what worked, fixes or replaces what failed, and only adds net-new forward work**. It will not re-create anything already built, not the live site, not the working waitlist, not the blog, and not the starter tasks that already ran.

### What You Get Out Of It

The result is a revised week that:

- replaces last week's plan as the company's **current plan**,
- lands tasks on the **weekly calendar with specific days and times**, and
- is **armed automatically** so the tasks run on schedule without you having to re-approve them.

If the review can't produce a coherent week for some reason, it does **nothing** rather than leaving you with a broken or empty plan, and it records that it couldn't, so the situation is visible instead of silent.

---

## When The Refresh Happens

The weekly refresh runs **once a week, on Sunday evening, in your company's own timezone**, roughly between 5:00 PM and 7:00 PM local time. (The exact minute is spread slightly across that window so that many companies don't all refresh at the same instant; it stays consistent for your company.)

Because it runs Sunday evening, your company starts each new week with a **freshly planned, up-to-date arc of work** already on the calendar.

The refresh runs in the background and does not interrupt anything the company is currently doing.

---

## How A Plan Becomes Scheduled Work

It helps to know the difference between the two stages of planning, because the weekly refresh repeats the same flow the first plan went through:

1. **The plan is written.** A week of tasks is authored, each with a category, a goal, and, for everything except the immediate first blog, a specific weekday and clock time.
2. **The tasks are armed.** Each scheduled task is placed on the company's calendar so it actually fires on its day and time, automatically, without needing a fresh approval each week.

The first plan does this at onboarding. The **weekly refresh does exactly the same thing every Sunday evening**, it writes the revised week and arms it, which is why the new plan simply takes over and the company keeps running without a gap.

---

## Seeing The Next Planning Time

You don't have to guess when the next refresh will happen. Open the company and go to the **Plan** tab. Below the weekly schedule, you'll see a small disclosure that reads something like:

> **Every Sunday at 5:30 PM**, Weekly review & plan next week
> Reviews this week's memory, completed work, and progress, then plans next week's build arc. Runs automatically.

That time is read from the **real, live schedule**, so it reflects exactly when your company's next plan will be written, in your company's timezone. If no weekly refresh is currently scheduled, the disclosure simply doesn't appear.

---

## Where Planning Shows Up

| Surface | What you see |
|---|---|
| **Plan** tab | The current plan, the weekly schedule grid (days × times), the "runs first" blog card, and the next-planning-time disclosure |
| **Tasks** tab | The same tasks, one at a time, with runtime status and per-task settings |
| **Schedule view** | Each task placed on its specific day and time across the week |

When the weekly refresh runs, all of these update together, the new tasks appear on the schedule, in the task list, and in the plan view at once.

---

## What This Means For You

- You don't have to re-plan your company every week, it re-plans itself.
- The plan stays **current**, reacting to what actually happened rather than running a stale script.
- The company keeps **moving forward**: building on wins, repairing misses, and adding new work each week.
- You can always check the **Plan** tab to see what's coming and exactly when the next plan will be written.

You remain in control. The plan is yours to refine at any time, connect missing tools, edit tasks, pause or resume work, and the weekly refresh builds forward from wherever the company actually is.

---

## Related

- [Onboarding & Plan Setup](auto-companies-onboarding)
- [Company Modes](company-modes)
- [Company Dashboard](auto-companies-dashboard)

---

### [Skill Learning](https://neotask.ai/docs/neotask-desktop/skill-learning)

Skill Learning lets an autonomous company's agent turn its own successful work into a reusable capability. When the agent completes a task in a new way that works, it can propose a skill describing how to do it again. Once that skill is live, every future run for that company can use it, so the agent keeps getting better at your specific business over time.

## What Skill Learning Is

A learned skill is a small, self-contained procedure the agent writes for itself after finishing real work, in the same format as a normal skill (a `SKILL.md` file with a description and instructions). Skills can come from the agent noticing a repeatable pattern at the end of a task, from a scheduled review of past work, or from a weekly cleanup pass (see below). Every proposal goes through an automated content check before it can ever be applied, so a learned skill can never smuggle in credentials or unsafe instructions.

Learned skills are scoped to a single company. They are not shared between companies and do not change what a different company's agent can do.

## Turning Skill Learning On or Off

Open your company's **Settings**, then go to **Skill Learning** under the Autonomy section. The **Learn new skills** toggle controls whether the agent is allowed to propose new skills from its work.

- **On by default.** Every company starts with skill learning enabled.
- **Turning it off pauses new proposals only.** Skills that are already live keep working. Disabling the toggle stops the agent from writing new ones; it does not remove what it already learned.

## How Approval Works

Whether a proposed skill needs your sign-off depends on the company's autonomy tier, set in the Approvals section of the same settings area:

| Autonomy tier | What happens to a new proposal |
|---|---|
| **Supervised** | Requires your approval before it goes live. |
| **Balanced** | Requires your approval before it goes live. |
| **Autopilot** | Applies automatically and appears in the activity feed. |

Under Supervised or Balanced, a proposed skill sits as a pending item until you approve or reject it. It never expires or times out on its own. Under Autopilot, the agent writes the skill immediately without waiting on you, but the write is never silent: it is recorded as an auditable event so you can see exactly what was added and when.

Changing the company's autonomy tier updates this behavior automatically; you don't need to do anything else in the Skill Learning section for the new tier to take effect.

## Reviewing Pending Skills

The **Pending skills** card, shown right on the Skill Learning settings page, is where you review skills waiting on a decision. For each one you'll see:

- The skill's name and a chip describing what kind of change it is (new skill, addition, or revision to an existing one).
- A status chip: Pending, Applied, Rejected, or Quarantined.
- A short title, the agent's stated reason for proposing it, and when it was proposed.

Click **Inspect** to expand a proposal and see the reason in full along with the exact content that would be written if you approve it. From there, **Approve** makes the skill live for the agent; **Reject** discards the proposal. Both actions ask for a quick confirmation first. If nothing is waiting on you, the card simply says there is nothing to review.

A **Quarantined** status means the automated content check flagged something in the proposal. Quarantined items can be inspected but cannot be approved or rejected from this card, since they were never allowed to reach the pending state in the normal path.

The card refreshes automatically in the background, and you can also refresh it manually at any time.

![Skill proposal showing the review controls and proposed instructions](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/skill-proposal-1280.webp)

## Weekly Skill Consolidation

Once a week, Neotask can review a company's learned skills and merge ones that overlap into a single, cleaner skill, in addition to retiring skills that have gone stale. This keeps the skill set from growing cluttered as the agent learns more over time. Nothing is ever deleted outright. Retired skills are archived and remain recoverable rather than removed.

This runs on its own cadence in the background (no more than once a week, and only once a company has learned enough skills for consolidation to be worthwhile) as part of skill learning, and only while skill learning itself is turned on.

To opt out, go to **Skill Learning** in company settings and turn off **Weekly skill consolidation**. This toggle is separate from the main **Learn new skills** toggle. Turning off consolidation still lets the agent propose new skills; it just stops the periodic merge/cleanup pass. Like the main toggle, it is on by default, and it is grayed out (not available to change) whenever skill learning as a whole is off, since there would be nothing to consolidate.

## Durability: Skills Survive a Reinstall

Learned skills are not just a local file on your machine. Every time a skill is applied or its status changes, Neotask saves the full record to your company's account on our servers. That durable, per-company copy is the source of truth, not the local copy on disk. The local copy is only a working cache.

Practically, this means if you reinstall the desktop app, reset your machine, or move to a new device, your company's learned skills come back automatically once you sign back in. They are rebuilt from the server-side copy, not lost with the old install.

## Related

- [Custom Skills And Plugins](custom-skills-and-plugins.md)
- [Skills](skills)
- [Webhook Triggers](webhooks)
- [Auto - Approvals & Safety](auto-companies-approvals-and-safety)
- [Auto - Operations, Controls & Settings](auto-companies-operations-controls-and-settings)

---

### [What's used for training, and what never is](https://neotask.ai/docs/neotask-desktop/training-data)

## Check The Current Choice

Open **Settings → Privacy and Data**. The main session-data control shows whether sharing is enabled. Company controls remain unavailable until the main control is enabled.

![Session data sharing control, off by default](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-consent-1280.webp)

When you turn on **"Share usage data to train smarter agents"** (it's on by default during onboarding, and you can turn it off anytime in **Settings → Privacy & Data**), Neotask uses sanitized data from your agent runs to make the models behind your agents smarter. This page explains exactly what is shared, what is never shared, and the controls you keep. For the broader picture see [Data Privacy & Your Rights](data-privacy-and-your-rights) and [Trust & Security](trust-and-security).

## What is used when sharing is on

With sharing enabled, we use **sanitized content from your agent runs**:

- **Prompts and responses**, what you and your agents said, after on-device sanitization (below).
- **Tool activity**, which tools and apps your agents used and how those calls went, again sanitized.
- **Run outcomes**, whether tasks succeeded, were retried, or needed a human, so models learn what "good" looks like for real workflows.

That's it. Training data comes only from this sanitized agent-run stream, never from your account records, billing data, or credential stores.

## What is never used

**Your credentials, passwords, API keys, and connected-app logins are never part of training data.** Three separate layers make sure of it:

1. **Masked on your device, before anything leaves it.** The Neotask engine that runs on your machine applies its canonical secret-redaction pipeline to every trace *on-device*. Secrets, tokens, keys, and credential-shaped values are masked before the data is stored or shipped anywhere. What leaves your machine is already sanitized.
2. **Credentials live in a separate, encrypted world.** The logins and tokens for apps you connect are stored in dedicated encrypted stores (AES-256-GCM), fully isolated from the training pipeline. The export pipeline has no access to those stores. It never reads them, so even if we wanted to include a credential, the pipeline never sees one.
3. **Identifiers are pseudonymized.** Before anything is exported to training infrastructure, account, company, session, and task identifiers are replaced with per-customer pseudonyms (keyed one-way hashes). Training infrastructure never receives your real identifiers, and those pseudonyms are destroyed when your data is erased.

Additionally:

- **HIPAA deployments and HIPAA customers are excluded outright**, no training export at all, regardless of any toggle.
- **No consent, no export.** Sharing is checked per customer on every export run; if it's off, nothing ships.

## Custom models for your companies

Sharing usage data does more than improve Neotask generally: it lets us **train models tuned to your specific companies and workflows**: agents that learn your SOPs, your tone, and the way your operations actually run. If you're interested in a custom per-workspace model, contact **sales@neotask.ai**.

## Your control

When the main sharing control is enabled, use the company list to keep sharing off for a specific company. The company row shows the current choice before you change it.

![Company-specific training controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r3/privacy-company-controls-1280.webp)

- **Turn it off anytime** in **Settings → Privacy & Data** in the desktop app. From that moment, your new agent runs are not used for training.
- **Deleting your account erases everything.** Account deletion runs a full erasure cascade that covers your agent-run traces and their stored content, and the pseudonym keys that could ever link exported data back to you.
- Questions? **privacy@neotask.ai**.

<!-- Internal accuracy breadcrumb (verified 2026-07-17, fi/158-training-terms-docs):
     - On-device masking before storage/shipping: neotask-gateway extensions/llm-tracing,
       src/sdk-runtime.ts (canonical redaction pipeline sanitizeDiagnosticPayload → redactSecrets,
       "applied before anything is stored"), src/trace-sink.ts, src/llm-trace-finalize-worker.runtime.ts
       (imports src/agents/payload-redaction.js + src/logging/redact.js). The gateway runs on the user's machine.
     - Encrypted credential stores, AES-256-GCM, separate from training: server/src/utils/crypto.ts
       (ALGO = 'aes-256-gcm', versioned keyring), server/src/models/TenantSecret.ts,
       server/src/models/ExternalAccount.ts. Neither server/src/workers/inferenceExportWorker.ts nor
       server/src/services/inferenceOtlpExporter.ts imports any credential model or decrypt helper.
     - Pseudonymized identifiers (per-tenant HMAC) before export: server/src/services/inferenceOtlpExporter.ts
       ("Identity attributes carry PSEUDONYMS ONLY (per-tenant HMAC keys)"),
       server/src/workers/inferenceExportWorker.ts (EXPORT_PSEUDONYM_KEY gate).
     - HIPAA exclusion: server/src/workers/inferenceExportWorker.ts (isHipaaMode() blocks the whole
       deployment; per-tenant Tenant.hipaaTenant skip; blockedBy 'hipaa-deployment' / 'hipaa-tenant').
     - Per-tenant/per-company consent gate on every run: inferenceExportWorker.ts
       (tenantsSkipped reason 'no-consent'; consentedCompanyIds filter).
     - Erasure cascade covers trace spans + stored content: server/src/services/tenantErasure.ts
       (proved in tenantErasure.cov.test.ts, deletes llm-trace contentRef objects).
     - Server-held encryption deliberately phrased "stored encrypted, isolated", keys are server-side
       (versioned keyring in crypto.ts), NOT user-held; do not claim user-only decryption. -->

---

### [Webhook Triggers](https://neotask.ai/docs/neotask-desktop/webhooks)

## Create And Test A Webhook

1. Open **Settings → Automations → Webhooks**.
2. Select **New webhook**.
3. Choose the event and agent or company destination.
4. Copy the endpoint and signing secret into the sending service.
5. Send a test event and confirm it in the events log.

![Webhook subscriptions with copy, test, verify, edit, and enable controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/webhook-subscriptions-1196.webp)

Webhook triggers let an external system kick off work in your autonomous company by sending it an HTTP request. When something happens elsewhere, such as a GitHub push, a support ticket, or an event from your own service, that system POSTs to a URL that belongs to your company, and Neotask turns it into a prompt for your agent to act on.

## What a Webhook Trigger Is

Each webhook trigger is a **route**: a company-scoped URL, a signature scheme used to verify who's allowed to send to it, an optional filter on which event types to accept, and a prompt template that turns the incoming payload into instructions for your agent. When a verified delivery arrives, Neotask renders the prompt and queues it; the desktop app picks it up and fires an agent run from it.

## Creating a Route

Open your company's **Settings**, then go to **Webhooks** under Integrations, and create a new route. You'll provide:

- **Name**: a label for the route (a route slug is derived from it automatically, and you can edit the slug directly).
- **Signature verifier**: which signing scheme the sender uses (see below).
- **Event filter**: optional, see [Filtering by Event Type](#filtering-by-event-type).
- **Prompt template**: the text sent to your agent, with placeholders filled in from the incoming payload.

### Provider Presets

Instead of picking a signature scheme and building a prompt template from scratch, you can start from a curated preset for a specific provider, GitHub, Stripe, Resend, Clerk, and JotForm among them. A preset pre-selects the right signature scheme and pre-fills a starter prompt template with placeholders that match that provider's real event shape. Neotask remembers which preset a route was created from so **Send test event** (see [Testing Your Webhook](#testing-your-webhook)) can use realistic sample data for that exact provider rather than a generic one for the underlying scheme.

Two presets are worth calling out because they don't map to a new signature scheme of their own:

- **Resend** and **Clerk** both use the **Svix-style** scheme under the hood, no dedicated scheme was needed for either, since both providers deliver their webhooks the same way any Svix-based sender does. The secret direction for both is **the provider mints it, you paste it back into Neotask**: add your route's URL as an endpoint in the Resend or Clerk dashboard, copy the signing secret it shows you, and paste it into your route's settings.
- **JotForm** has no webhook signing capability of any kind, confirmed directly on JotForm's own support forum, where JotForm's staff recommend a shared token in the webhook URL as the workaround. A JotForm route therefore uses the **URL token** scheme instead of a signature scheme, and the JotForm preset also handles a format difference: JotForm posts your form submission as `multipart/form-data`, not JSON, with the submission answers riding inside a `rawRequest` field. Neotask parses that automatically for any URL-token route receiving a multipart delivery, so your prompt template's placeholders resolve against the submission's parsed fields exactly like they would for a JSON provider. See the URL token row below and the JotForm note under [Which direction the secret travels](#which-direction-the-secret-travels) for the full detail.

Choosing a preset is a convenience, every field it fills in is still a normal, editable route setting afterward, and you can always build a route from scratch by picking a signature scheme directly instead of a preset.

## The Webhook URL

Each route gets its own public URL in this form:

```
https://neotask.ai/api/hooks/<companyId>/<routeSlug>
```

Give this URL to the external system you want to trigger your agent from. It only accepts `POST` requests, and every request must be correctly signed. There is no unsigned/open mode. The route also answers a bodyless `HEAD` probe of the same URL, some providers, such as Intercom, send a `HEAD` request to validate the URL before they let you save the webhook, so the endpoint replies `200` when the URL resolves to an enabled route and `404` otherwise, without ever verifying, storing, or rate-limiting anything (a `HEAD` carries no payload).

## Signature Verification

Every route picks one signature scheme when it's created, and the endpoint verifies every delivery against it. Neotask supports sixteen schemes:

| Scheme | How it verifies |
|---|---|
| **GitHub-style** | Reads the `X-Hub-Signature-256` header (`sha256=<hexdigest>`), an HMAC-SHA256 of the raw request body keyed with your route's secret. This is the same scheme GitHub itself uses for repository webhooks. |
| **GitLab** | Reads the `X-Gitlab-Token` header and compares it against your route's secret. This is GitLab's own webhook scheme: a shared secret token sent verbatim, not a computed signature. There is no timestamp in this scheme, so no replay window applies; a redelivered event is instead deduplicated by its delivery id. |
| **Stripe** | Reads the `Stripe-Signature` header (`t=<timestamp>,v1=<signature>`), an HMAC-SHA256 of the timestamp and raw body together, keyed with your Stripe signing secret (the whole `whsec_...` string). Deliveries older than five minutes are rejected as a replay-protection measure. During a Stripe secret rotation the header can carry more than one `v1` entry, and any matching one is accepted. |
| **Shopify** | Reads the `X-Shopify-Hmac-Sha256` header, a base64 HMAC-SHA256 of the raw request body keyed with your route's secret. No timestamp exists in this scheme, so replay protection comes from delivery id deduplication. |
| **Jira (token in URL)** | Standard Jira webhooks are unsigned, so this scheme uses a shared token carried in the webhook URL itself: append `?token=<secret>` to your route's URL when you register it in Jira. The endpoint compares the token against your route's secret and rejects anything else. |
| **Svix-style** | Reads `webhook-id`, `webhook-timestamp`, and `webhook-signature` headers, verifying an HMAC-SHA256 signature over the id, timestamp, and raw body together. Deliveries older than five minutes are rejected as a replay-protection measure. Use this for any sender built on the Svix webhook standard. |
| **Generic timestamped HMAC** | For senders that don't speak any of the formats above. Reads an `X-Webhook-Timestamp` header and an `X-Webhook-Signature` header (a hex HMAC-SHA256 of the timestamp and raw body together), with the same five-minute replay window. |
| **Calendly** | Reads the `Calendly-Webhook-Signature` header, in the same `t=<timestamp>,v1=<signature>` shape as Stripe, an HMAC-SHA256 of the timestamp and raw body together, keyed with your route's secret. Deliveries older than five minutes are rejected. |
| **Cal.com** | Reads the `X-Cal-Signature-256` header, a hex HMAC-SHA256 of the raw request body keyed with your route's secret. No timestamp exists in this scheme, so replay protection comes from delivery id deduplication. |
| **Typeform** | Reads the `Typeform-Signature` header (`sha256=<base64digest>`), a base64 HMAC-SHA256 of the raw request body keyed with your route's secret. No timestamp exists in this scheme. Typeform sends no signature at all if the webhook has no secret configured, so an unsigned request is rejected the same as a wrongly-signed one. |
| **Tally** | Reads the `Tally-Signature` header, a raw base64 HMAC-SHA256 digest of the raw request body (no prefix) keyed with your route's secret. No timestamp exists in this scheme. |
| **Vercel** | Reads the `x-vercel-signature` header, a hex HMAC-SHA1 of the raw request body keyed with your route's secret. No timestamp exists in this scheme. (Vercel's own scheme intentionally uses SHA1, not SHA256.) |
| **HubSpot** | Reads the `X-HubSpot-Signature-v3` header (a base64 HMAC-SHA256) together with the `X-HubSpot-Request-Timestamp` header, over a signed string that concatenates the HTTP method, the exact public request URL, the raw request body, and the timestamp, keyed with your route's secret. Deliveries whose timestamp is more than five minutes old are rejected. Because the signature covers the exact URL HubSpot called, this scheme depends on your route's public URL matching what's registered in HubSpot's app settings byte-for-byte, if you ever front this endpoint with an additional proxy that rewrites the host or path, verification will fail closed rather than silently accept a mismatched request. |
| **Intercom** | Reads the `X-Hub-Signature` header (`sha1=<hexdigest>`), an HMAC-SHA1 of the raw request body keyed with your route's secret. No timestamp exists in this scheme. |
| **Linear** | Reads the `Linear-Signature` header, a hex HMAC-SHA256 of the raw request body keyed with your route's secret. Replay protection here works differently from every other scheme: Linear puts a `webhookTimestamp` field inside the JSON body itself rather than a header, and Neotask rejects anything more than one minute old (a tighter window than the five minutes used elsewhere). |
| **URL token (generic)** | Reads the `?token=<secret>` query-string parameter and compares it against your route's secret, the same mechanism as the Jira scheme above, generalized for any provider that offers no signing of its own. JotForm is the motivating example: JotForm has no webhook signing capability at all, so a JotForm route uses this scheme instead. Because there is no body or header signature to check, this scheme's security rests entirely on keeping the full URL, including the token, secret, and only ever sending it over HTTPS. Unlike the "Generic timestamped HMAC" scheme above (for senders that CAN sign but don't match a named provider), this scheme is for senders that CAN'T sign at all. |

## The Signing Secret

When you create a route, Neotask generates a secret and shows it to you exactly once, with a clear warning that it will not be shown again. Copy it into your sending system right away. If you lose it, you can rotate the secret from the route's settings, which generates a new one and shows it once in the same way; the old secret immediately stops working.

The secret is never displayed again after that moment, in the UI or in any export. Only whether a route has one is shown afterward.

### Which direction the secret travels

Who mints the secret depends on the provider, and the route settings support all three directions:

- **Neotask mints it, you paste it into the provider**: GitHub, GitLab, Calendly, Cal.com, Typeform, Tally, and custom senders (Svix-style and generic HMAC). Create the route, copy the secret Neotask shows you once, and paste it into the provider's webhook configuration as its signing secret or secret token. (Calendly's standard integration path works this way; a separate OAuth-app integration style has Calendly generate a signing key per application instead, which doesn't apply to a single route's secret.)
- **The provider mints it, you paste it back into Neotask**: Stripe, Shopify, Vercel, HubSpot, Intercom, Linear, **Resend**, and **Clerk**. Each of these generates its own secret rather than letting you choose one, Stripe's signing secret (`whsec_...`) when you register the endpoint, Shopify's shared secret shown below the webhooks list in Settings > Notifications, Vercel's shown once when the webhook is created, HubSpot's app Client Secret from its Auth settings, Intercom's app Client Secret from the Developer Hub, Linear's shown on the webhook's detail page, and Resend's and Clerk's own Svix-backed signing secrets (also `whsec_...`) shown once when you add your route's URL as an endpoint in their dashboards. Register your route's URL with the provider first, copy the secret it shows you, then paste it into your route's settings in Neotask. Setting a secret this way replaces the previous one immediately, and Neotask never displays it back to you afterward.
- **The secret rides in the URL**: Jira, and any route using the generic **URL token** scheme, most notably **JotForm**, which has no webhook signing of any kind. Because these providers carry no signature, the route's secret instead acts as an access token appended to the URL you give the provider (`?token=<secret>`). Treat that full URL as sensitive, exactly as you would the secret itself, anyone who has the URL can trigger the route.

For a provider-mints-it route, the route's secret is set to a Neotask-generated placeholder the moment you create it, so verification against a real delivery fails closed, every delivery is rejected with `401`, until you paste the provider's own secret back in. Neotask tracks whether that paste-back has happened, and the desktop app warns on a route created from a provider-mints-it preset (or manually pointed at one of those verifiers) until it detects a successful paste-back, so it's obvious a route is silently rejecting every delivery rather than discovering it only after the provider reports failed deliveries. Rotating a route's secret (even for a provider-mints-it route) always generates a new Neotask placeholder and re-arms this warning, since the newly generated secret is not the provider's real signing secret either, paste the provider's secret back in again after a rotation.

## Testing Your Webhook

Every route has a **Send test event** action in its settings. It runs your prompt template against a small, realistic sample payload for the route's provider (for example, a pull request event for a GitHub route or a successful payment event for a Stripe route) and queues the result as a pending event, marked as a test so you can tell it apart in the events log. The response shows you the exact rendered prompt your agent would receive, which makes it the quickest way to check that your placeholders resolve the way you expect before pointing the real sender at the route.

If your route was created from a specific [provider preset](#provider-presets) such as Resend, Clerk, or JotForm, the test event uses a sample shaped for that exact provider, a JotForm route's test, for example, renders against a realistic parsed submission rather than the generic URL-token sample. A route created without a preset (or from an unrecognized one) falls back to a sample for its underlying signature scheme, exactly as before.

Test events go through the same template rendering, the same queue, and the same rate limit as real deliveries, but they skip signature verification, because you trigger them from inside Neotask rather than from the external sender. A disabled route refuses to send test events until you re-enable it.

## Filtering by Event Type

By default a route accepts every delivery that passes signature verification. If you only want certain event types to reach your agent, add them to the route's event filter as a comma-separated list. Neotask determines the event type of each delivery the same way GitHub-style clients do: from an `X-GitHub-Event` header if present, otherwise from an `event_type` or `type` field in the payload itself. A filtered-out delivery is acknowledged but never queued for your agent.

## Prompt Templates and Placeholders

The prompt template is what your agent actually receives, built from the verified payload. Use `{dot.path}` placeholders to pull values out of the incoming JSON, for example `{action}` or `{repository.full_name}`. A few rules to keep in mind:

- Placeholders resolve against the parsed payload only; a value you insert is never re-scanned for more placeholders, so payload content can't inject additional substitutions.
- If a path doesn't exist in the payload, the placeholder is left in the prompt as literal text rather than silently disappearing, so a broken template is easy to spot.
- Long values are trimmed (each inserted value up to 2,000 characters), and the whole rendered prompt is capped at 16,384 characters.

## Delivery Status and the Events Log

After sending a test, open the event list and confirm its delivery state before registering the real sender.

![Recent webhook event with its delivery state](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/webhook-events-1196.webp)

The **Webhooks** section also shows a log of recent deliveries for each route, grouped by status:

- **Pending**: verified and queued, waiting for the desktop app to run it.
- **Consumed**: picked up and run.
- **Failed**: picked up but the run did not complete successfully.

Not every delivery becomes a logged event. A request can also be turned away or acknowledged without ever reaching the queue: an unsigned or incorrectly signed request is rejected, a redelivery of something already received is acknowledged without creating a second event, a delivery excluded by your event filter is acknowledged but not queued, and a route sending too many requests too quickly is rate-limited.

## Rate Limits and Payload Size

Each route accepts up to 30 deliveries per minute by default; requests beyond that are rejected until the window resets. Each delivery is also capped at 1 MiB, and larger payloads are rejected outright.

## Run Policy

Each route also carries a run policy that governs the run a verified delivery triggers: whether you're notified when it fires (off by default), an optional pinned model for that run, and an optional daily cap on how many runs the route can trigger. Leaving all of this unset behaves exactly as described above, a verified delivery is queued and picked up by the desktop app with no extra gate. The policy is captured on each event at the moment it's received, so an edit to a route's run policy only affects deliveries from that point forward, never events already sitting in the queue. Enforcement of the run policy (the delivery notification, the model pin, and the daily cap) is handled by the desktop app.

There is no per-route pre-run approval setting: a webhook-triggered run is governed in-run by the same approval settings and autonomy tier as the company's autonomous runs, so it asks for approval exactly where any other autonomous action would.

## HIPAA Mode

If your company has HIPAA mode enabled, inbound webhook deliveries are acknowledged with a normal success response, but the payload is never processed or stored. This check happens before anything else, including signature verification, because an external sender isn't a party Neotask has a signed agreement with to receive protected health information. In practice this means webhook triggers are effectively inert for a company in HIPAA mode: deliveries won't be rejected outright, but they also won't reach your agent.

## Main Agent Webhooks

Webhook triggers aren't only for a company's agent. You can also create a route on your account's main agent, so an external system can kick off work that isn't tied to any specific company.

Main agent routes work the same way as company routes: the same signature verifier choices, the same optional event filter, the same prompt template placeholders, the same rate limits and payload cap, and the same once-only secret display. The differences are where you create them and the shape of the URL.

Create a main agent route from your account-level **Webhooks** settings, alongside your company's onboarding surfaces. Each route gets its own public URL in this form:

```
https://neotask.ai/api/hooks/main/<publicId>
```

The `<publicId>` is a random, opaque identifier Neotask generates for the route; it doesn't reveal your account or tenant. Deliveries, signature verification, filtering, idempotency, rate limiting, and the events log all work exactly as described above for company routes, just scoped to your main agent instead of a company.

## Related

- [Skill Learning](skill-learning)
- [Automation & Scheduling](automation)
- [Auto - Apps, Integrations & Files](auto-companies-apps-integrations-and-files)

---

### [Agents](https://neotask.ai/docs/neotask-gateway/agents)

## See An Agent At Work

1. Open **Agents**.
2. Select an agent.
3. Use **Chat** to start work or continue a session.
4. Open the activity drawer to inspect the apps and tool calls used during the run.

![Main agent chat with an active work session](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/chat-main-1280.webp)

## What Are Agents?

The overview brings scheduled work, session totals, and current agent activity together. Use it to inspect the workspace before opening an individual session.

![Agent overview with sessions, scheduled work, and activity](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/overview-main-1280.webp)

Agents are isolated AI assistants that reason through problems, execute tasks, and communicate across multiple platforms. Each agent has its own workspace, session history, model configuration, and channel bindings.

## Agent Isolation

Every agent gets:

- **Workspace**, A dedicated directory for files, notes, and agent-specific data
- **Sessions**, Independent conversation history and transcripts
- **Configuration**, Model selection, tool permissions, spend limits
- **Identity**, A unique persona defined in the agent's `AGENTS.md` file
- **Auth Profiles**, Per-agent API key and authentication configurations

## Agent Personas

Each agent's workspace contains an `AGENTS.md` file that defines personality, instructions, and behavioral guidelines. This persistent system prompt shapes how the agent responds, tone, domain expertise, formatting, tool usage, and safety policies.

## Multi-Agent Routing

When running multiple agents, inbound messages are routed automatically based on **bindings**:

- **Channel binding**, Route all messages from a platform (e.g., all Telegram → `personal` agent)
- **Server/workspace binding**, Route a specific Discord server or Slack workspace to a dedicated agent
- **Chat type binding**, Route private messages to one agent and group messages to another
- **Contact/peer binding**, Route messages from specific contacts to specific agents

Bindings are evaluated in priority order. First match wins. Unmatched messages go to the default agent.

## Sub-Agents & Delegation

Agents can spawn sub-agents for specialized tasks:

- **Task delegation**, Break down complex tasks and delegate parts to specialized sub-agents
- **Isolated execution**, Sub-agents run in their own session context
- **Ping-pong conversations**, Agents can have back-and-forth exchanges (up to 5 turns) to collaborate
- **Result announcement**, Sub-agent results can be posted back to the original channel

## Teams & Orchestration

Multi-agent teams enable coordinated workflows:

- **Parallel execution**, Run multiple agents simultaneously on different aspects of a task
- **Sequential pipelines**, Chain agents where each builds on the previous output
- **Hybrid mode**, Combine parallel and sequential execution
- **Task tracking**, Monitor status of each agent's work (queued, running, done, error, canceled)

## Tool Profiles

Control what each agent can do:

| Profile | Tools Available |
|---------|----------------|
| **minimal** | Session status only |
| **coding** | File I/O, execution, sessions, memory, images |
| **messaging** | Message sending, session history |
| **full** | All tools (default) |

Custom allow/deny lists can be set per agent and per model provider.

## Safe Mode

Safe Mode provides an execution sandbox:

- **Master toggle**, Enable or disable globally
- **Per-agent controls**, Disable for trusted agents, keep on for others
- **Execution approval**, Sensitive operations require explicit user approval
- **Policy sync**, Settings synchronized across all connected clients in real time

## Agent Memory

Each agent has access to a memory system:

- **MEMORY.md**, Persistent notes and important information across sessions
- **Memory files**, Additional structured knowledge in `memory/`
- **Vector search**, Search memory using natural language queries

See [Sessions & Memory](sessions-memory.md) for details.

---

### [Automation & Scheduling](https://neotask.ai/docs/neotask-gateway/automation)

## Create And Check An Automation

1. Open **Automations** in the agent workspace.
2. Select **New**.
3. Choose a schedule, event, or one-time run.
4. Set the task and delivery destination.
5. Save it, then open the run history after the first execution.

![Automations page with active schedules and recent runs](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/automations-main-1280.webp)

## Cron Jobs

Schedule agent wakeups with precise timing. Cron jobs persist across Gateway restarts and support flexible scheduling patterns.

### Schedule Types

| Type | Description | Example |
|------|-------------|---------|
| **Cron expression** | Standard cron syntax with timezone | `0 7 * * *` (daily at 7 AM) |
| **Interval** | Repeat at fixed intervals | `30m`, `2h`, `1d` |
| **One-shot** | Run once at a specific time | `2026-03-15T16:00:00Z` |

### Session Targets

- **Main session**, Enqueue as a system event in the agent's main conversation
- **Isolated session**, Run in a dedicated session (`cron:<jobId>`) to avoid contaminating main conversations

### Delivery Modes

| Mode | Description |
|------|-------------|
| **None** | Silent execution (no output posted) |
| **Announce** | Post the result back to a specific channel |
| **Webhook** | HTTP POST the result to an external URL |

### One-Shot Reminders

Set a reminder for a specific time. The agent wakes up, processes the reminder, and optionally deletes the job after running.

### Recurring Jobs

Schedule repeating tasks with cron expressions and timezone support. Jobs track their execution history (success, failure, timing).

---

## Webhooks

Receive events from external services and route them to agents.

### Gmail Pub/Sub

Integrate with Gmail's push notification system to trigger agent actions when emails arrive:

1. Set up a GCP Pub/Sub topic and subscription
2. Configure Gmail label filtering
3. Point push delivery to your webhook URL
4. Neotask receives messages and enqueues agent events

### Custom Webhooks

Register HTTP endpoints that trigger agent turns when called. Useful for integrating with CI/CD pipelines, monitoring systems, or any service that can send HTTP requests.

---

## System Events

System events are signals that wake an agent for processing:

- **Immediate**, Process on the next heartbeat
- **Next heartbeat**, Queue for the next scheduled check-in

Use system events for reactive workflows, agents wake up when something happens rather than on a fixed schedule.

---

## Heartbeat

The heartbeat is a periodic signal (default ~15 seconds) that gives agents an opportunity to process pending work:

- Check for new system events
- Process queued messages
- Run pending automations

The heartbeat can be enabled, disabled, or configured per-agent.

---

## Hooks

Event-driven automation that triggers shell commands or scripts in response to Gateway events:

- **On message**, Execute when a message is received
- **On agent turn**, Execute before or after an agent processes a message
- **On channel event**, Execute on channel-specific events (join, leave, reaction)

Hooks run on the Gateway host and have access to event context data.

---

## Use Cases

| Scenario | Solution |
|----------|----------|
| Daily morning briefing | Cron job at 7 AM → agent summarizes overnight updates |
| Weekly report generation | Cron job every Monday → agent generates and emails report |
| Email monitoring | Gmail webhook → agent processes incoming emails |
| Build notifications | CI/CD webhook → agent announces build results on Discord |
| Hourly data check | Interval-based cron → agent checks dashboard and alerts on anomalies |
| Scheduled social posts | Cron job → agent drafts and posts content at optimal times |

---

### [Channels](https://neotask.ai/docs/neotask-gateway/channels)

## Connect And Verify A Channel

1. Open **Channels** from the agent workspace.
2. Select **Add channel**.
3. Choose the messaging service and finish its authentication flow.
4. Run the connection tests.
5. Set who can message the agent and which actions require approval.

![Channel picker with supported messaging services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/channel-picker-1120.webp)

## What Are Channels?

Channels are the messaging platforms your agents connect to. By connecting a channel, your agent can listen for incoming messages and respond in real time on that platform.

Neotask supports **20+ messaging platforms** out of the box.

## Supported Channels

### Core Channels (Built-In)

| Channel | Description |
|---------|-------------|
| **WhatsApp** | Personal and business accounts via WhatsApp Web protocol. Groups, reactions, media, threads, edit/unsend, forwarding. |
| **Telegram** | Bot API integration. Groups, channels, threads, reactions, pinning. Optional webhook ingestion. |
| **Discord** | Full server integration. Channels, roles, threads, voice channel awareness, per-server agent binding. |
| **Slack** | Workspace app via OAuth. Channels, threads, reactions, file sharing. |
| **Signal** | Encrypted DMs and groups via Signal protocol. Phone pairing via QR. |
| **Google Chat** | Spaces, threads, reactions, card responses via webhook. |
| **iMessage** | Full SMS/iMessage support via BlueBubbles on macOS. Edit, unsend, effects. |
| **IRC** | Classic IRC with channels, DMs, SASL auth, multi-server support. |

### Plugin Channels

| Channel | Description |
|---------|-------------|
| **Microsoft Teams** | Workspace integration via plugin. |
| **Matrix** | Decentralized messaging protocol. |
| **Nostr** | Decentralized social protocol DMs. |
| **Zalo** | Vietnamese messaging platform (Bot API and personal account). |
| **Feishu/Lark** | Enterprise messaging platform. |
| **LINE** | LINE Messaging API integration. |
| **Nextcloud Talk** | Self-hosted messaging via Nextcloud. |
| **Mattermost** | Open-source team messaging. |
| **Tlon** | Urbit-based messaging. |
| **Twitch** | Live stream chat via IRC. |

## Channel Features

### Multi-Account Support

Use the connection test after adding an account. A passed test confirms that the saved channel account can be reached before you configure routing.

![Passed channel connection tests](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/channel-tests-1280.webp)
Run multiple accounts per channel. For example, manage several WhatsApp numbers or Telegram bots from the same Gateway.

### Media Support
Send and receive images, videos, audio, documents, stickers, and other media types. Media handling varies by platform but most channels support rich media in both directions.

### Group & Channel Management
Agents can participate in group conversations, respond to mentions, manage threads, and interact with channel-specific features (reactions, polls, pins).

### Message Threading
On platforms that support it (Slack, Discord, Telegram), agents can create threads, reply within threads, and maintain thread-specific context.

### Message Editing & Reactions
Agents can edit their own messages, unsend messages, and add reactions on platforms that support these features.

### Broadcast Groups
Send announcements to multiple channels simultaneously. Configure broadcast groups to keep multiple platforms in sync.

## Channel Routing

### Default Behavior
By default, the main agent receives all inbound messages from all channels.

### Multi-Agent Routing
With multiple agents, messages are routed based on bindings:

- **By channel**, All WhatsApp messages → personal agent
- **By server**, Discord server A → support agent, server B → dev agent
- **By workspace**, Slack workspace → team agent
- **By chat type**, Private DMs → assistant, group messages → team bot

### DM Access Control

Slack access settings define which channels and direct messages the agent can use. Review these controls when a connected account works but a specific destination does not.

![Slack access and approval settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/slack-access-1280.webp)
Configure which contacts can message your agent directly. Options include allowing all DMs, only known contacts, or a specific allowlist.

### Group Mention Gating
In group chats, configure whether agents respond to all messages or only when mentioned.

## Pairing & Authentication

Each channel has its own authentication method:

| Channel | Auth Method |
|---------|-------------|
| WhatsApp | QR code scan |
| Telegram | Bot token |
| Discord | Bot token |
| Slack | OAuth (workspace app) |
| Signal | Phone pairing via QR |
| Google Chat | Webhook URL |
| iMessage | BlueBubbles API key |
| IRC | Nick + optional password |

---

### [Companion Apps](https://neotask.ai/docs/neotask-gateway/companion-apps)

## Confirm The Desktop Connection First

Before pairing another device, open **Settings → Gateway** on the desktop app and confirm that the current Gateway is connected and healthy.

![Gateway connection path used by the desktop and companion apps](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-connection-path-1280.webp)

## What Are Companion Apps?

Companion apps are native applications for iOS, Android, and macOS that connect to your Gateway as **nodes**. Nodes expose device capabilities, camera, canvas, screen recording, location, voice, and system execution, to your agents.

Think of it this way: the Gateway is the brain, and companion apps are the hands and eyes on your devices.

## How Nodes Work

1. **Connect**, The companion app connects to your Gateway over WebSocket
2. **Pair**, New devices go through a pairing approval process
3. **Register capabilities**, The app tells the Gateway what it can do (camera, canvas, location, etc.)
4. **Agents invoke**, During conversations, agents can request the node to take a photo, record screen, render a canvas, and more

Nodes are **not gateways**, they don't run their own AI. They're peripheral devices that extend what your agents can do.

## macOS Companion App

The macOS app lives in your menu bar and serves as both a companion node and a gateway manager.

### Capabilities
- **Canvas**, Render interactive web content in a native window
- **Camera**, Take photos and video clips (front/back)
- **Screen Recording**, Full-screen or per-app recording
- **System Execution**, Run shell commands on the Mac (gated by exec approvals)
- **Notifications**, Send native macOS notifications with priority options
- **Location**, GPS when available

### Additional Features
- **Gateway Management**, Start, stop, and monitor the Gateway
- **Permission Ownership**, Manages macOS TCC permissions (Notifications, Accessibility, Screen Recording, Microphone, Speech Recognition, AppleScript)
- **Deep Links**, `neotask://` URL scheme for launching agents with parameters
- **CLI Install**, Installs the `neotask` command-line tool

## iOS App

The iOS app connects as a node to your Gateway, bringing mobile capabilities to your agents.

### Capabilities
- **Canvas**, SwiftUI-based interactive content rendering
- **Camera**, Photos and video clips (front/back cameras)
- **Audio Recording**, Microphone capture
- **Location**, GPS with accuracy selection
- **Notifications**, Native iOS push notifications

### Features
- **Discovery**, Finds your Gateway via Bonjour (LAN), Tailscale (VPN), or manual entry
- **Pairing**, Secure device pairing with approval from the Gateway
- **Voice Wake**, Wake word activation for hands-free interaction
- **Talk Mode**, Continuous voice conversation
- **Share Extension**, Share content from other apps directly to your agent
- **A2UI**, Agent-driven UI rendering on the device canvas

### Limitations
- Foreground-first design (iOS suspends WebSocket connections in background)
- Strict background command restrictions
- Currently in internal preview

## Android App

The Android app connects as a node with capabilities similar to iOS.

### Capabilities
- **Canvas**, Jetpack Compose-based UI rendering
- **Camera**, Photos and video (front/back, permission-gated)
- **Screen Recording**, System screen capture (Android 10+)
- **Location**, GPS with accuracy modes
- **SMS Sending**, On devices with telephony support
- **Chat & History**, View conversation history on device

### Features
- **Foreground Service**, Persistent notification keeps the connection alive
- **Shared Sessions**, Same session across all connected nodes
- **Modern Android**, Requires Android 12+ (API 31)

## Headless Node Host

Run a node without any UI, ideal for build servers, remote machines, or headless Linux boxes.

### Capabilities
- **System Execution**, Run commands on the node host
- **Command Allowlist**, Configure which commands the node can execute
- **Persistent Service**, Install as launchd/systemd service

### Use Cases
- CI/CD build agents, Let your AI agent trigger builds on a remote server
- Home lab automation, Run commands on a Raspberry Pi or home server
- Multi-machine orchestration, Coordinate tasks across multiple machines

## Device Pairing

### Pairing Flow
1. Companion app presents its device identity (fingerprint + public key)
2. Gateway issues a pairing approval request
3. You approve the device through the desktop app or dashboard
4. Gateway issues a device token for future connections

### Trust Model
- **Local devices** (loopback) are auto-approved for convenience
- **Non-local devices** require explicit approval
- Device tokens are stored locally and reused on reconnect

## Exec Approvals

For nodes that can run system commands (macOS, headless hosts), exec approvals control what's allowed:

- **Allowlist mode**, Only pre-approved commands can execute
- **Ask mode**, Unknown commands prompt you for approval
- **Full mode**, No restrictions (use with caution)

Configure per-node allowlists to precisely control what each device can do.

---

### [Deployment](https://neotask.ai/docs/neotask-gateway/deployment)

Neotask can run anywhere, from your laptop to a Kubernetes cluster. Choose the deployment method that fits your needs.

After deployment, open **Settings → Gateway → Health** in the desktop app. Confirm that the service, runtime, and connection checks pass before connecting channels or running agents.

![Gateway health checks after deployment](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-health-1280.webp)

## Local Installation

The simplest way to get started. Install Neotask directly on macOS, Linux, or Windows (via WSL2). The Gateway runs as a background service managed by launchd (macOS) or systemd (Linux).

**Requirements:** Node.js 22+

**What you get:**
- Gateway running as a background service
- Automatic startup on boot
- macOS companion app (on macOS)
- All CLI tools

## Docker

Run the Gateway in a Docker container with persistent config and workspace volumes. Ideal for VPS deployments or when you want isolation from your host system.

**Features:**
- Pre-built images based on Node 22 (Debian)
- Optional Chromium + Xvfb for browser automation (adds ~300MB)
- Docker Compose setup with persistent volumes
- Health check endpoint at `/health`
- Runs as non-root `node` user
- Agent sandbox support (Docker-in-Docker for isolated execution)

**Persistent State:**
- Configuration: `~/.neotask/neotask.json`
- Workspace: `~/.neotask/workspace/`
- Sessions: `~/.neotask/agents/*/sessions/`
- Channel state: `~/.neotask/whatsapp/`, etc.

## Kubernetes

Deploy to Kubernetes using the StatefulSet pattern for persistent storage. Community-maintained Helm charts are available.

**Typical setup:**
- **StatefulSet** for the Gateway (persistent sessions and config)
- **ConfigMap** for `neotask.json` configuration
- **Secret** for tokens and API keys
- **PersistentVolumeClaim** for session data and workspace
- **Service** (ClusterIP or LoadBalancer) for WebSocket access

## Fly.io

One-command deployment to Fly.io with persistent volumes and automatic HTTPS.

**Features:**
- x86 VM with configurable RAM (default 2048MB)
- Persistent `/data` volume for config and sessions
- Automatic TLS termination
- SSH access for configuration
- Webhook support via ngrok (optional)

## Hetzner VPS

Run on a dedicated Hetzner VPS with Docker Compose. Full infrastructure-as-code support with community Terraform modules.

**Setup includes:**
- Docker Compose configuration
- Persistent directories for config, workspace, and channel state
- Custom Dockerfile for baking in required binaries
- Firewall and security hardening

## Render

Blueprint-based deployment on Render with automatic builds and 1GB persistent disk.

## GCP Compute Engine

Docker-based deployment on Google Cloud with persistent disk and startup automation.

## Podman

Rootless container execution with Podman and systemd Quadlet support. Ideal for environments where Docker daemon access is restricted.

## Choosing a Deployment

| Need | Recommended |
|------|-------------|
| Just trying it out | Local install |
| Personal use on your Mac | Local + macOS companion app |
| Always-on server | Docker on VPS (Hetzner, DigitalOcean) |
| Production with scaling | Kubernetes |
| Quick cloud deploy | Fly.io or Render |
| Rootless containers | Podman |

## Remote Access

By default, the Gateway only listens on localhost (127.0.0.1). For remote access:

- **Tailscale**, VPN mesh network with automatic DNS. The Gateway can bind to your Tailnet for secure access from anywhere. Tailscale Serve can expose the dashboard with HTTPS.
- **SSH Tunnel**, Forward the Gateway port through SSH for secure remote access without exposing ports.
- **LAN Bind**, Bind to your local network for access from other devices on the same network. Automatic discovery via Bonjour/mDNS.

See [Gateway](gateway.md) for configuration details.

---

### [Gateway](https://neotask.ai/docs/neotask-gateway/gateway)

## Check The Gateway From The Desktop App

1. Open **Settings**.
2. Select **Gateway**.
3. Check the connection path and runtime status.
4. Open **Gateway preferences** before changing the host, port, or launch behavior.

![Gateway settings with the active connection path](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-connection-path-1280.webp)

## What Is the Gateway?

The Gateway is the heart of Neotask, a single long-running service that manages all agent sessions, messaging channels, tool execution, and device connections. Everything flows through the Gateway.

It's a WebSocket server that listens for connections from clients (desktop app, mobile apps, web dashboard) and manages the entire AI assistant lifecycle.

## Architecture

### Components

- **WebSocket Server**, Accepts client connections (default port 18789)
- **HTTP Server**, Serves the Control UI, canvas host, and health endpoints
- **Channel Adapters**, Maintain connections to messaging platforms (WhatsApp, Telegram, Discord, etc.)
- **Session Manager**, Persists and manages agent conversation state
- **Cron Scheduler**, Runs scheduled jobs and heartbeat wakeups
- **Plugin Runtime**, Loads and manages plugin lifecycle
- **Browser Service**, Controls Chromium instances for web automation

### Binding Modes

| Mode | Description |
|------|-------------|
| **Loopback** | Default. Only accessible from localhost (127.0.0.1) |
| **LAN** | Accessible on your local network. Auto-discoverable via Bonjour/mDNS |
| **Tailnet** | Accessible through Tailscale VPN with MagicDNS |
| **Custom** | Bind to a specific address |

### Authentication

| Method | Description |
|--------|-------------|
| **Token** | WebSocket bearer token (UUID or custom string) |
| **Password** | Password-based auth (bcrypt hashed) |
| **Trusted Proxy** | For reverse proxies with pre-auth headers |
| **Local Trust** | Loopback connections auto-approved |

## Configuration

Gateway preferences contain the local runtime and launch controls. Confirm the active connection path before changing them.

![Gateway preferences with local runtime controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-preferences-1280.webp)

The Gateway is configured via `neotask.json` (JSON5 format). Key sections:

| Section | Controls |
|---------|----------|
| `agents` | Agent list, defaults, workspaces |
| `gateway` | Port, bind mode, auth, reload behavior |
| `channels` | Per-channel config and credentials |
| `tools` | Tool permissions, profiles, security |
| `plugins` | Plugin loading, enabling, per-plugin config |
| `models` | Provider auth, fallbacks, provider settings |
| `skills` | Skill directories, per-skill configuration |
| `browser` | Browser profiles, ports, headless mode |
| `cron` | Scheduler configuration |
| `session` | Reset rules, compaction thresholds |
| `messages` | TTS settings, media policies |
| `memory` | Memory indexing configuration |
| `security` | Sandboxing, elevated mode, approvals |

### Hot Reload

Configuration changes can be applied without restarting:

| Mode | Behavior |
|------|----------|
| **Hybrid** (default) | Hot-apply safe changes, restart for others |
| **Hot** | Only apply hot-safe changes, ignore others |
| **Restart** | Restart on any reload-required change |
| **Off** | No automatic reload |

### Restart Recovery

After an interruption, an app or Gateway update, a host restart, a crash, or a power loss, the Gateway can resume the work that was in flight: interrupted scheduled runs, interactive agent threads, and coding sessions rebound to their own repository, branch, and worktree. Three environment variables control it.

| Variable | Values | Default | Controls |
|---|---|---|---|
| `NEOTASK_RESUME_AFTER_RESTART` | `1` or `0` | Set per launch by the desktop app | Master switch for restart recovery in the managed service |
| `NEOTASK_THREAD_CONTINUATION` | `0`, `false`, `no`, `off` to disable | On | Continuing interactive agent threads on the same session after a restart |
| `NEOTASK_CODING_WORKTREE_BINDING` | `0`, `false`, `no`, `off` to disable | On | Binding each coding session to its repository, branch, and worktree, plus worktree repair and safe cleanup |

Notes for operators:

- `NEOTASK_RESUME_AFTER_RESTART` is written into the managed service environment by the Neotask desktop app from the account-wide and per-company **Continue work after restarts** setting.
- `NEOTASK_THREAD_CONTINUATION` and `NEOTASK_CODING_WORKTREE_BINDING` are opt-out switches. Leave them unset to keep the shipped behavior; set one to `0`, `false`, `no`, or `off` to disable that half of recovery while leaving the other in place.
- Scheduled run resumption is bounded. Completed work is never re-executed, and blog and email side effects are idempotent across a retry.
- Pending approvals are never resumed by any of these switches. They stay pending.

## Remote Access

### Tailscale (Recommended)

Tailscale provides secure, zero-config remote access:

- Bind the Gateway to your Tailnet
- Access from any device on the same Tailscale network
- Automatic DNS via MagicDNS
- Tailscale Serve can expose the dashboard with HTTPS

### SSH Tunnel

Forward the Gateway port through SSH:

- Simple, secure, no configuration changes needed
- Works with any SSH-capable machine
- Supports the full WebSocket protocol and Control UI

### Security Rules

- **Loopback + SSH/Tailscale** is the safest combination
- Non-loopback binding requires auth token configuration
- TLS fingerprint pinning available for extra security
- Browser control is disabled for remote connections by default

## Health Monitoring

### Health Endpoint

The `/health` HTTP endpoint returns:

- Gateway status
- RPC liveness
- Per-channel health (connected, auth status, account count)
- Browser readiness
- Active agents and connected nodes
- Uptime

### Probes

- **RPC liveness**, WebSocket connect + echo test
- **Channel health**, Auth check + API ping per channel
- **Browser readiness**, Port check, process check
- **Node connectivity**, Last heartbeat timestamp

## Discovery

### Bonjour/mDNS

When bound to LAN, the Gateway advertises itself via multicast DNS:

- Service type: `_neotask-gw._tcp`
- Includes port, transport, TLS status, and Tailscale DNS info
- Companion apps discover the Gateway automatically

### Wide-Area Discovery

Discover gateways across your network with configurable timeouts. Supports both local Bonjour and Tailscale DNS discovery.

## Multiple Gateways

You can run multiple Gateway instances on the same machine using profiles:

- Each profile has separate config, workspace, sessions, and state
- Different ports for each Gateway
- Useful for isolation (personal vs work) or testing

## Diagnostics

### Doctor

The built-in diagnostic tool checks:

- Configuration validity
- File permissions
- Channel connectivity
- Model auth status
- Node.js version compatibility
- Common misconfigurations

It can automatically fix many issues when given permission.

### Logging

Gateway logs are written to daily JSONL files:

- Per-completion token usage tracking
- Channel event logs
- Error logging with structured context
- Configurable log levels

---

### [Models & Providers](https://neotask.ai/docs/neotask-gateway/models-providers)

## Choose How Models Are Connected

Open **Settings → Models**. The provider overview shows connected subscriptions, hosted credits, and your own keys in one place.

![Model connection options for subscriptions, credits, and provider keys](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/connection-options-1280.webp)

1. Choose **Subscriptions** to connect a supported account.
2. Choose **Credits** to browse the hosted model catalog.
3. Choose **Your keys** to add a provider key.
4. Select the model or routing default for the agent.

![Hosted model catalog showing NVIDIA and other providers](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/credits-catalog-nvidia-1280.webp)

## Overview

The Neotask gateway supports a wide range of AI model providers. You can use cloud-hosted models from Anthropic, OpenAI, and Google, run local models via Ollama or vLLM, or connect to an OpenAI-compatible API endpoint in direct BYOK mode.

## Supported Providers

### Cloud Providers

| Provider | Models | Auth |
|----------|--------|------|
| **Anthropic** | Claude Opus 4.6, Claude Sonnet 4.1, Claude Haiku 4.5 | API key or setup-token |
| **OpenAI** | GPT-5.2, GPT-5.1, GPT-4 Turbo, GPT-4 | API key |
| **Google (Gemini)** | Gemini 3 Pro, Gemini 2 Flash, Gemini 1 Pro | API key or OAuth |
| **Together AI** | Various open-weight models (Llama, Mixtral, etc.) | API key |
| **Moonshot (Kimi)** | Moonshot v1 (32K, 128K context) | API key |
| **OpenRouter** | 100+ models from multiple providers | API key |
| **DeepSeek** | DeepSeek Chat, DeepSeek Coder | API key |
| **OpenCode Zen** | Claude variants, custom-tuned models | API key |

### Local Model Providers

| Provider | Description |
|----------|-------------|
| **Ollama** | Run open-weight models locally (Llama, Mistral, etc.) |
| **vLLM** | High-throughput local inference server |
| **LiteLLM** | Unified proxy for 100+ providers |

### Custom Providers

In direct BYOK mode, connect an OpenAI-compatible or Anthropic-compatible API by specifying a base URL and your own API key. This supports self-hosted inference servers, enterprise proxy endpoints, and third-party model APIs without sending a Neotask-managed provider key to that custom host.

Hosted-credit requests use Neotask's server passthrough instead. That path accepts only the official HTTPS hosts assigned to the selected provider, rejects provider/host mismatches, and does not follow upstream redirects. Custom base URLs are never accepted on the hosted-credit path.

## Model Configuration

### Primary Model
Set a default model for all agents. Each agent can override this with its own primary model.

### Fallback Chains
Configure a prioritized list of fallback models. If the primary model fails (rate limit, downtime, auth error), Neotask automatically tries the next model in the chain.

### Image Models
Separate model configuration for image analysis and generation. Supports Claude Vision, GPT-4 Vision, and DALL-E 3.

### Model Aliases
Create shortcuts for long model names. For example, alias `fast` to `anthropic/claude-haiku-4-5` for quick reference.

### Per-Agent Overrides
Each agent can have its own model configuration, independent of the global default.

### Model Allowlists
Restrict which models agents can use. Useful for cost control or compliance.

## API Key Management

Switch to the provider-key catalog when you want to connect credentials that you manage. Each provider card opens its own setup flow.

![Bring your own key provider catalog](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/models/2026-08-13-r4/own-keys-catalog-1280.webp)

### Key Sources (Priority Order)
1. Live override key (highest priority)
2. Comma-separated key list (rotation)
3. Primary key
4. Numbered keys (key_1, key_2, etc.)

### Automatic Key Rotation
When a rate limit is hit, Neotask automatically rotates to the next available key. Non-rate-limit failures return errors immediately without rotation.

### Per-Provider Configuration
Each provider can have its own set of API keys, auth profiles, and model-specific settings.

## Provider Auth Methods

| Method | Description |
|--------|-------------|
| **API Key** | Standard bearer token authentication |
| **Setup Token** | Interactive auth flow (Anthropic) |
| **OAuth** | Delegated auth with refresh tokens (Google, GitHub Copilot) |
| **Token Paste** | Manual token entry for specialized providers |

## Model Selection Priority

When an agent needs to make an LLM call, models are selected in this order:

1. Per-agent override (if set)
2. Global default model
3. Model allowlist (if configured, limits options)
4. First available model (fallback)

## Provider-Specific Tool Restrictions

You can restrict which tools are available for specific models or providers. For example, limit a less-trusted model to file operations only, while giving Claude full tool access.

---

### [Platform Overview](https://neotask.ai/docs/neotask-gateway/platform-overview)

The desktop app connects to the Neotask Gateway automatically. Open **Settings → Gateway** to see the active connection, local runtime, and health checks in one place.

![Gateway connection path showing the desktop app, local runtime, and hosted services](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-connection-path-1280.webp)

1. Open **Settings**.
2. Select **Gateway**.
3. Confirm that the connection path shows the expected runtime.
4. Open **Health** when a connection or tool is unavailable.

![Gateway health checks with connection and runtime status](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r1/gateway-health-1280.webp)

## What Is Neotask?

The **Gateway** is the core engine that powers Neotask. It's a single long-running service that manages everything: agent sessions, messaging integrations, tool execution, scheduled automations, and device connections.

Think of it as the brain that your Neotask desktop app, mobile apps, and web dashboard all connect to. While you interact through those interfaces, the Gateway is doing the heavy lifting behind the scenes.

## Architecture

Neotask uses a hub-and-spoke model:

- **The Gateway** is the central hub, a WebSocket server that runs on your machine (or in the cloud). It manages all agent sessions, routes messages between channels, orchestrates tool execution, and persists state.
- **Clients** (the desktop app, mobile apps, web dashboard, or terminal) connect to the Gateway and send/receive messages through it.
- **Nodes** (companion apps on iOS, Android, macOS) expose device capabilities, camera, canvas, screen recording, location, and voice, to agents running on the Gateway.
- **Channels** (WhatsApp, Telegram, Discord, Slack, etc.) are messaging transports that the Gateway connects to for sending and receiving messages.

All AI operations go through the Gateway → LLM Provider pipeline. The desktop app never calls an LLM directly.

## Key Capabilities

### Multi-Channel Messaging
Connect to 20+ messaging platforms simultaneously. Your agents can send and receive messages on WhatsApp, Telegram, Discord, Slack, Signal, iMessage, Google Chat, Microsoft Teams, Matrix, IRC, and more, all from a single Gateway instance. See [Channels](channels.md).

### Multi-Agent System
Run multiple isolated agents, each with their own workspace, personality, model configuration, and channel bindings. Route inbound messages from specific channels, servers, or contacts to the right agent automatically. See [Agents](agents.md).

### Device Capabilities via Nodes
Connect iOS, Android, and macOS companion apps as "nodes" that expose device hardware to your agents. Agents can take photos, record video, capture the screen, get GPS location, render interactive canvases, and run system commands, all through natural language. See [Companion Apps](companion-apps.md).

### Browser Automation
Agents can control a full Chromium browser, open pages, click elements, fill forms, take screenshots, extract content, upload files, and run JavaScript. Multiple browser profiles are supported for account isolation. See [Tools & Capabilities](tools-capabilities.md).

### Canvas & A2UI
The Agent-to-UI (A2UI) system lets agents render interactive visual content on connected devices. Agents can present web pages, push structured UI updates, execute JavaScript in the canvas context, and take snapshots of what's displayed. See [Tools & Capabilities](tools-capabilities.md).

### Flexible Model Support
Use any major LLM provider, Anthropic (Claude), OpenAI (GPT), Google (Gemini), Together AI, Moonshot, OpenRouter, and more. Configure model fallback chains, aliases, and per-agent model overrides. Run local models via Ollama or vLLM. See [Models & Providers](models-providers.md).

### Plugin & Skill Ecosystem
Extend Neotask with plugins that add new channels, tools, RPC methods, and capabilities. Browse available skills in the [skills catalog](/skills). Create your own skills to teach agents new workflows. See [Plugins & Skills](plugins-skills.md).

### Scheduling & Automation
Schedule agent wakeups with cron expressions, one-shot reminders, or interval-based triggers. Agents can run automated tasks, deliver results to channels, or post to webhooks. See [Automation](automation.md).

### Voice Interaction
Use wake words to activate your agent by voice. Talk mode enables continuous voice conversations with real-time speech-to-text and text-to-speech (ElevenLabs, OpenAI). The Swabble daemon on macOS provides always-on, on-device voice detection. See [Voice](voice.md).

### Session Intelligence
Sessions automatically reset daily, compact when approaching context limits, and persist across restarts. Memory search lets agents recall information from previous conversations using vector similarity. See [Sessions & Memory](sessions-memory.md).

### Self-Hosted & Private
Everything runs on your infrastructure. Your messages, API keys, and agent data never leave your machines unless you explicitly configure an external service. The Gateway binds to localhost by default, zero external network exposure.

### Sandboxed Execution
Run agent commands in isolated Docker containers with configurable resource limits, network policies, and filesystem confinement. Per-agent sandbox profiles let you control exactly what each agent can access. See [Security](security.md).

## How It All Fits Together

1. **You** interact with your agents through the Neotask desktop app, a mobile companion app, or directly via messaging platforms like WhatsApp or Telegram.
2. **The Gateway** receives your message, routes it to the right agent, and starts an AI turn.
3. **The agent** processes your request using the configured LLM (Claude, GPT, etc.), calls any tools it needs (browser, canvas, file operations, shell commands), and generates a response.
4. **The response** is delivered back to you through the same channel, or announced on a different channel if configured.
5. **State** (session transcripts, agent config, scheduled jobs) is persisted locally so everything survives restarts.

---

### [Plugins & Skills](https://neotask.ai/docs/neotask-gateway/plugins-skills)

## Install Or Review A Skill

1. Open **Skills**.
2. Search the catalog or open **My skills**.
3. Review the tools and configuration the skill needs.
4. Approve a learned skill only after checking its source and proposed behavior.

![Skills catalog with installed and available skills](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/skills-main-1280.webp)

## Plugins

### What Are Plugins?

Plugins are TypeScript modules that extend Neotask's functionality at runtime. They can add new messaging channels, agent tools, RPC methods, CLI commands, background services, and skills.

### Official Plugins

**Messaging Channels:**
- Microsoft Teams, Matrix, Nostr, Zalo, Feishu/Lark, LINE, Nextcloud Talk, Mattermost, Tlon, Twitch

**Features:**
- **Voice Call**, SIP telephony support for inbound/outbound voice calls
- **Memory (LanceDB)**, Long-term memory with vector similarity search
- **MCP Native**, Model Context Protocol server integration for bridging external tools

**Auth Providers:**
- Google Antigravity OAuth, Gemini CLI OAuth, Qwen OAuth, Minimax OAuth, VS Code Copilot Proxy

### Plugin Capabilities

Plugins can register:
- **RPC Methods**, Custom WebSocket methods for clients to call
- **HTTP Handlers**, Custom HTTP endpoints on the Gateway
- **Agent Tools**, New tools with typed schemas that agents can invoke
- **CLI Commands**, Additional command-line commands
- **Background Services**, Long-running processes managed by the Gateway
- **Skills**, Skill directories bundled with the plugin
- **Auto-Reply Commands**, Commands that execute without AI agent involvement

### Plugin Security

Neotask checks plugins in several layers before they run:

- **Signed and verified.** Packages in the public Construct catalog are published by Neotask, inspected automatically before release and signed with Neotask's release key. Before installing one, Neotask checks the signature against keys built into Neotask and verifies the download's SHA-256 hash.
- **No install scripts.** Construct packages ship with their dependencies bundled, so installing one never runs a package manager or an install script. When another plugin install uses npm, install scripts are turned off.
- **Scanned on install.** Plugins installed from outside Construct are scanned for dangerous code, and critical findings block the install.
- **Licensed per release.** A Construct plugin loads only while your account holds a current grant for that release. Revoked releases do not load.
- **Approvals for agent actions.** Agent tool calls, including tools that plugins add, are checked against your approval policy, and actions that need your sign-off wait for it. Software an agent wants to install outside Neotask's bundled install recipes always shows you the exact command and waits for your approval.
- **Optional sandbox.** You can run an agent's tools in an isolated Docker container instead of directly on your computer. Sandboxing is off by default; see [Security](security.md#sandboxing) to turn it on.

Native plugins run inside the Gateway process, so a plugin you install from outside Construct is code you choose to run. Use `plugins.allow` to list the plugins that may load.

---

## Skills

### What Are Skills?

Skills are modular capabilities that teach agents new workflows. Each skill is a markdown file (`SKILL.md`) with instructions that get injected into the agent's context when relevant.

### Skill Locations (Priority Order)

1. **Workspace skills**, `<workspace>/skills/` (highest priority)
2. **Managed skills**, `~/.neotask/skills/`
3. **Bundled skills**, Shipped with Neotask
4. **Plugin skills**, Skills packaged with installed plugins
5. **Extra directories**, Additional configured paths (lowest priority)

When multiple skills share the same name, the highest-priority location wins.

### Skill Features

- **User-invocable**, Trigger skills directly with slash commands (e.g., `/docker-build`)
- **Model-invocable**, Skills the agent can invoke on its own when relevant
- **Command dispatch**, Skills that map to specific tools
- **Requirements gating**, Skills only load when their requirements are met (required binaries, config keys, environment variables, plugins)

### Creating Skills

When an agent proposes a learned skill, review the instructions and requested capabilities before approving it.

![Learned skill proposal awaiting review](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/skill-proposal-1280.webp)

Create a `SKILL.md` file with YAML frontmatter defining the skill's metadata and instructions:

**Frontmatter fields:**
- `name`, Skill identifier
- `description`, What the skill does
- `homepage`, Link to documentation
- `user-invocable`, Whether users can trigger it directly
- `disable-model-invocation`, Prevent the agent from invoking it autonomously
- `command-dispatch`, Tool to dispatch to
- `metadata.neotask.requires`, Prerequisites (bins, config, env, plugins)

**Body:** Markdown instructions that teach the agent how to use the skill, including examples and guidelines.

### Skills Catalog

Browse available skills in the [skills catalog](/skills).

### Skill Categories

Skills cover a wide range of capabilities:

| Category | Examples |
|----------|---------|
| **Productivity** | Email management, calendar, document generation |
| **Development** | Git operations, CI/CD, Docker, code review |
| **Communication** | Messaging, notifications, announcements |
| **Media** | Image processing, video editing, audio transcription |
| **AI & Models** | Model switching, prompt management, RAG |
| **Smart Home** | IoT device control, automation triggers |
| **Utilities** | File conversion, data transformation, web scraping |

---

### [Security](https://neotask.ai/docs/neotask-gateway/security)

## Check The Controls That Affect Agent Actions

Open **Approvals** to review pending actions. Open **Settings → Permissions** to change the default approval policy or Safe Mode.

![Approval detail with the requested action and review controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-detail-1280.webp)

## Overview

Neotask is designed with a security-first architecture. The Gateway binds to localhost by default, all sensitive data is encrypted at rest, and agent execution can be sandboxed in isolated containers.

## Network Security

### Loopback by Default

The Gateway only listens on `127.0.0.1` (localhost) unless you explicitly change the binding mode. No external network exposure by default.

### Authentication

All WebSocket connections require authentication:

- **Token auth**, Bearer token (UUID or custom string)
- **Password auth**, Bcrypt-hashed password
- **Trusted proxy**, Pre-authenticated headers from reverse proxies
- **Local trust**, Loopback connections are implicitly trusted

### Remote Access Security

For remote access, the recommended approach is **Tailscale VPN** or **SSH tunneling**, never expose the Gateway directly to the internet.

## Device Pairing

### How Pairing Works

Every client that connects to the Gateway must be paired:

1. Device presents its identity (fingerprint + public key)
2. Gateway issues a pairing challenge (nonce)
3. Device signs the nonce with its private key
4. You approve the device through the UI
5. Gateway issues a device token for future connections

### Trust Model

- **Local devices** (loopback) are auto-approved for convenience
- **Non-local devices** require explicit approval
- Device tokens are stored locally and reused on reconnect
- Devices can be revoked at any time

## Encryption

### Data at Rest

- **AES-256-GCM**, All tokens, secrets, and API keys are encrypted at rest
- **Machine-derived keys**, Encryption keys derived via scrypt from device identity
- **No plaintext storage**, Tokens are never stored in plaintext

### Data in Transit

- **TLS**, HTTPS for API communication
- **WSS**, WebSocket Secure for Gateway connections (when using Tailscale Serve or reverse proxy)
- **HMAC signing**, Request integrity with nonce and timestamp

## Sandboxing

### Docker-Based Isolation

Agent command execution can be sandboxed in Docker containers:

- **Per-agent profiles**, Each agent can have its own sandbox configuration
- **Resource limits**, Configurable CPU, memory, and timeout limits
- **Network isolation**, Containers can be network-isolated
- **Filesystem confinement**, Agents only access their workspace
- **Custom images**, Use pre-built images with specific tools installed

### Sandbox Scopes

| Scope | Description |
|-------|-------------|
| **Per-session** | Fresh container for each session |
| **Per-agent** | Persistent container per agent |
| **Shared** | Shared container across agents |

### Workspace Access

Sandboxed agents can access their workspace directory (mounted into the container) but cannot access the host filesystem outside their workspace.

## Exec Approvals

Guardrail settings define which classes of agent actions pause for review. Check them before changing the policy for an individual request.

![Agent guardrails and approval controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/v7/2026-08-11-r1/ui/control-guardrails-1280.webp)

Control what commands agents can execute on nodes and the Gateway host:

### Modes

| Mode | Description |
|------|-------------|
| **Allowlist** | Only pre-approved commands execute |
| **Ask** | Unknown commands prompt for user approval |
| **Full** | No restrictions (use with caution) |

### Per-Node Configuration

Each node (macOS, headless host) has its own exec approval configuration, stored locally. You can allow specific binaries (e.g., `/usr/bin/docker`, `/usr/local/bin/terraform`) while blocking everything else.

## Elevated Mode

Some operations require running on the Gateway host directly (not in a sandbox). Elevated mode is:

- Gated by explicit permission in agent config
- Only available on the Gateway host
- Disabled by default
- Auditable through logging

## Security Audit

The built-in security audit checks:

- Configuration file permissions (should be 600)
- Session file permissions
- Node.js version (security patches)
- Plaintext secret detection
- Plugin allowlist validation
- Open network interface audit
- Auth configuration completeness

The audit can automatically fix many issues when given permission.

## Best Practices

1. **Keep the Gateway on loopback**, Use Tailscale or SSH for remote access
2. **Enable sandboxing**, Run agent commands in Docker containers
3. **Use exec allowlists**, Restrict what commands agents can run on nodes
4. **Set auth tokens**, Always configure token auth, even for loopback
5. **Review agent permissions**, Use minimal tool profiles where possible
6. **Keep Neotask updated**, Updates include security patches
7. **Audit regularly**, Run security audits to catch misconfigurations
8. **Restrict plugin loading**, Only install trusted plugins

---

### [Sessions & Memory](https://neotask.ai/docs/neotask-gateway/sessions-memory)

## Trace An Answer Back To Its Source

1. Open **Memory**.
2. Search for the decision, customer, project, or result.
3. Open a matching source to see the cited evidence.
4. Open **Sessions** when you need the full conversation and tool history.

![Memory answer with citations and source links](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r2/memory-cited-answer-1280.webp)

## Sessions

### How Sessions Work

Every conversation with an agent happens in a **session**. Sessions are identified by keys that encode the context, which agent, which channel, which chat.

Sessions maintain a complete transcript (append-only JSONL) of all messages, tool calls, and results. This transcript is what gives agents context about the ongoing conversation.

### Session Keys

Sessions are routed automatically based on context:

| Pattern | Meaning |
|---------|---------|
| `agent:main:main` | Main direct conversation |
| `agent:<id>:<channel>:group:<id>` | Group chat on a specific channel |
| `agent:<id>:<channel>:channel:<id>` | Channel/room conversation |
| `cron:<jobId>` | Scheduled job execution |
| `hook:<uuid>` | Webhook-triggered execution |

### Session Lifecycle

- **Daily reset**, Sessions automatically reset at a configurable time (default 4 AM local). This creates a fresh conversation context each day while preserving history.
- **Idle expiry**, Sessions that haven't been active for a configurable period can auto-reset.
- **Manual reset**, Start fresh anytime with a `/new` or `/reset` command.

### Token Tracking

Sessions track cumulative token usage (input, output, cache read, cache write) for cost monitoring and context window management.

---

## Compaction

### What Is Compaction?

When a conversation approaches the model's context window limit, Neotask **compacts** it, summarizing older messages into a condensed overview and starting a new session with that summary plus recent messages.

### How It Works

1. **Detection**, Monitor context usage (default trigger: 80% of context window)
2. **Summarization**, Older messages are summarized into a compaction entry
3. **New session**, A fresh session starts with the summary injected at the top
4. **Preservation**, Old transcripts are preserved for history but no longer loaded into context

### Configuration

- **Enabled/disabled**, Toggle auto-compaction
- **Threshold**, Percentage of context window that triggers compaction (default 80%)
- **Minimum messages**, Don't compact if there are fewer than N messages

### Manual Compaction

You can trigger compaction manually at any time with the `/compact` command. This is useful when you want to reset context without losing important information.

---

## Memory

### Agent Memory

Each agent has a memory system for persisting important information across sessions:

- **MEMORY.md**, A markdown file in the agent's workspace where important facts, preferences, and context are stored
- **Memory directory**, Additional files in `memory/` for structured knowledge

### Vector Search

The results table lists the evidence returned for a memory search. Open a result to inspect its source before relying on it in a session.

![Memory search results table with governed evidence](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/surfaces/2026-08-13-r2/memory-results-table-1280.webp)

Neotask supports vector similarity search over agent memory:

1. **Indexing**, Memory files are indexed using embeddings
2. **Search**, Agents query memory with natural language ("What did the user say about their API preferences?")
3. **Recall**, Relevant memories are returned ranked by similarity

### Memory Plugin (LanceDB)

The LanceDB memory plugin provides advanced vector storage:

- Persistent vector database
- Configurable chunk size and overlap
- Efficient similarity search
- Automatic re-indexing when memory files change

### Automatic Recall

At session start, relevant memories can be automatically recalled and injected into the agent's context. This gives agents continuity across sessions without loading entire conversation histories.

---

## Session Persistence

All session data is stored locally:

- **Session metadata**, `sessions.json` tracks active sessions, last activity, token counts, and config
- **Transcripts**, Individual session transcripts stored as `.jsonl` files
- **Recovery**, Sessions survive Gateway restarts without data loss

### Cleanup

- Sessions not accessed for a configurable period (default 30 days) are eligible for cleanup
- Manual session management is available through the UI or API
- Compacted transcripts are archived but can be reviewed

---

### [Tools & Capabilities](https://neotask.ai/docs/neotask-gateway/tools-capabilities)

## Review Tool Access

1. Open **Settings → Permissions**.
2. Review the tool groups available to the agent.
3. Keep Safe Mode enabled for commands that can change or delete data.
4. Inspect an approval request before allowing the action.

![Approval queue with pending agent actions](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/approvals-main-1280.webp)

## Overview

Tools are the actions your agents can take during conversations. When an agent needs to browse the web, run a command, read a file, or control a device, it uses tools.

Neotask provides a rich set of built-in tools covering file management, code execution, browser automation, device control, messaging, web search, and more.

## File Operations

Agents can read, write, and edit files in their workspace:

- **Read**, Read file contents
- **Write**, Create or overwrite files
- **Edit**, Make targeted edits to specific parts of a file
- **Apply Patch**, Apply multi-hunk unified diffs for complex file modifications

## Code Execution

Agents can run shell commands with configurable isolation:

- **Gateway execution**, Run commands directly on the machine hosting the Gateway
- **Sandbox execution**, Run commands in isolated Docker containers with resource limits
- **Node execution**, Run commands on connected companion devices (macOS, Linux nodes)

### Execution Features
- Configurable timeouts (default 30 minutes)
- Background job management (list, poll, kill, clear)
- Auto-background for long-running commands
- TTY support for interactive programs
- Elevated mode for privileged operations (gateway host only)

### Approval Workflows

Destructive command protection explains why a command was stopped and what review is required. Use it to distinguish a safety block from a failed tool call.

![Destructive command guard explaining a blocked command](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/destructive-guard-1280.webp)
Sensitive commands can require user approval before execution:
- **Allowlist**, Only pre-approved commands execute
- **Ask**, Unknown commands prompt for approval
- **Full**, No restrictions

## Browser Automation

Agents control a full Chromium-based browser (Chrome, Brave, or Edge):

### Actions
- **Navigate**, Open URLs and navigate between pages
- **Interact**, Click, type, hover, drag, select, fill forms
- **Capture**, Take screenshots, export PDFs
- **Inspect**, Snapshot the DOM or accessibility tree
- **Execute**, Run JavaScript in the page context
- **Upload**, Upload files to web forms
- **Manage**, Handle dialogs (alerts, confirmations), read console output

### Browser Profiles
Multiple isolated browser profiles for separating accounts:
- Each profile has its own cookies, storage, and history
- Auto-assigned CDP ports for programmatic control
- Remote browser support (connect to browsers on other machines)

## Canvas & A2UI

The Agent-to-UI (A2UI) system lets agents render interactive visual content on connected devices:

- **Present**, Display web pages or custom content on a device's canvas
- **Navigate**, Control what's shown in the canvas window
- **Evaluate**, Execute JavaScript in the canvas context
- **Snapshot**, Capture what's currently displayed
- **A2UI Push**, Send structured UI updates (JSONL payloads) to the canvas
- **A2UI Reset**, Clear the canvas state

Canvas rendering is available on macOS (WebKit), iOS (SwiftUI), Android (Jetpack Compose), and web (WebChat).

## Node Device Capabilities

Through connected companion apps, agents can:

### Camera
- **Snap**, Take photos (front or back camera)
- **Clip**, Record video clips with configurable duration and FPS

### Screen
- **Record**, Capture screen recordings with configurable duration and quality

### Location
- **Get**, Retrieve GPS coordinates with accuracy selection

### Notifications
- **Notify**, Send native OS notifications with title, body, and priority

### System Commands
- **Run**, Execute shell commands on the node device
- **Which**, Check if a command is available on the node

## Messaging

Agents can send and manage messages across connected channels:

- **Send**, Send messages to any connected channel/recipient
- **Edit**, Edit previously sent messages
- **React**, Add reactions to messages
- **Thread**, Create threads and reply within them
- **Poll**, Create polls on supported platforms
- **Voice**, Check voice channel status (Discord)

## Web

- **Search**, Web search via Brave Search API
- **Fetch**, Retrieve web pages and convert HTML to markdown/text

## Memory & Context

- **Memory Search**, Vector similarity search over agent memory files
- **Memory Get**, Retrieve specific memory files by name

## Session Management

- **List Sessions**, View active sessions across agents
- **Session History**, Fetch transcript for any session
- **Send to Session**, Send messages to other agent sessions
- **Spawn Sub-Agent**, Start isolated sub-agent runs
- **List Agents**, View available agents

## Automation

- **Cron**, Create, edit, run, and manage scheduled jobs
- **Gateway**, Restart gateway, apply configuration changes, run updates

## Image Analysis

Agents can analyze images using configured vision models, describe content, extract text, identify objects, and answer questions about visual content.

## Tool Groups

Tools are organized into groups for easy permission management:

| Group | Tools |
|-------|-------|
| **runtime** | exec, bash, process |
| **fs** | read, write, edit, apply_patch |
| **sessions** | sessions_list, sessions_history, sessions_send, sessions_spawn, session_status |
| **memory** | memory_search, memory_get |
| **web** | web_search, web_fetch |
| **ui** | browser, canvas |
| **automation** | cron, gateway |
| **messaging** | message |
| **nodes** | nodes |

---

### [Troubleshooting](https://neotask.ai/docs/neotask-gateway/troubleshooting)

Use this guide when the Neotask Gateway will not start, disconnects repeatedly, or cannot complete agent work.

## Check Gateway Health In The Desktop App

### Connection

Open **Settings → Infrastructure → Connection**. This screen tests the full path through the desktop app, local Gateway, Neotask cloud, and connected services. It also provides **Start**, **Restart**, and **Stop** controls for the local Gateway.

![Gateway connection path and Gateway controls](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-connection-1280.webp)

### Runtime Health

Open **Settings → Infrastructure → Health** and click **Run check**. Review responsiveness, active work, queued work, channels observed, and recent stability events.

![Runtime Health with responsiveness, workload, channels, and stability checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-health-full-1280.webp)

### System Report

Open **Settings → Advanced → System report**. This combines the local Gateway, session access, runtime response, connected services, current work, and scheduled checks in one report.

![System report with live Gateway and service checks](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-system-report-1280.webp)

### Doctor And Repairs

1. Open **Settings → Advanced → Troubleshoot**.
2. Click **Check system**. The check is read-only.
3. Open **Show technical output** for the individual findings.
4. Click **Apply repairs** if Neotask offers it.
5. Wait for the repairs and automatic follow-up check to finish.

![Gateway Doctor controls in Neotask Troubleshoot settings](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r4/troubleshoot-system-check-1280.webp)

Retry the failed action after the follow-up check. If the problem remains, copy the technical output and include it in the support request.

### External Dependencies And Logs

Open **Settings → Infrastructure → External Dependencies** when an optional browser, document, media, local-search, or coding component is missing.

![External Dependencies with installed optional components](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/external-dependencies-1280.webp)

Open **Settings → Advanced → Logs** to search and copy the redacted local Gateway event stream.

![Redacted local Gateway Logs](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r5/gateway-logs-1280.webp)

---

## Gateway Issues

### Gateway won't start

- Fully quit Neotask and open it again once.
- Open **Settings → Infrastructure → Connection** and check the **Local gateway** stage.
- Use **Start** or **Restart** under **Gateway controls**.
- Review **Settings → Advanced → System report** for the first live system check that needs attention.
- Run **Check system** from **Settings → Advanced → Troubleshoot** if the Gateway still does not start.
- Open the technical output and read the first failed check.
- Click **Apply repairs** and wait for verification.
- If verification fails, copy the technical output for support. Do not delete Gateway files or edit local configuration at random.

### Gateway starts but no channels connect

- **Missing credentials**, Each channel needs its own auth (bot token, QR scan, API key).
- **Network issues**, Channels need internet access to connect to messaging platform APIs.
- **Rate limits**, Some platforms rate-limit new connections. Wait and retry.

### Can't connect from the desktop app

- **Wrong port**, Ensure the desktop app connects to the correct Gateway port.
- **Auth mismatch**, The Gateway token must match.
- **Firewall**, Ensure the port is accessible if the Gateway is on another machine.

---

## Channel Issues

### WhatsApp won't connect

- **QR expired**, QR codes expire after ~60 seconds. Re-scan quickly.
- **Multi-device limit**, WhatsApp limits linked devices.
- **Session corrupted**, Delete the WhatsApp session directory and re-pair.

### Telegram bot not receiving messages

- **Bot token invalid**, Verify your bot token with BotFather.
- **Privacy mode**, Bots only see messages when mentioned in groups by default.
- **Webhook conflict**, Another service may be consuming messages.

### Discord bot not responding

- **Missing intents**, Enable required Gateway Intents in the Discord Developer Portal.
- **Missing permissions**, The bot needs read and send permissions in target channels.

---

## Model Issues

### Auth errors

- **Key not configured**, Ensure the provider API key is set.
- **Key expired**, Some OAuth tokens expire. Re-authenticate.
- **Rate limit**, Key rotation will switch automatically if you have multiple keys.

### Slow responses

- **Model choice**, Larger models are slower. Try a faster model for quick tasks.
- **Context size**, Long conversations slow processing. Try `/compact`.
- **Network latency**, Check connectivity to your model provider.

---

## Node Issues

### Companion app can't find the Gateway

- **Binding mode**, The Gateway must be bound to LAN or Tailnet (not loopback) for external devices.
- **Same network**, For Bonjour discovery, both devices must be on the same network.
- **Manual entry**, Enter the Gateway host and port manually in app settings.

---

## Session Issues

### Context window exceeded

- **Compact**, Use `/compact` to summarize and reset context.
- **Enable auto-compaction**, Set a compaction threshold in config.
- **New session**, Start fresh with `/new`.

---

## Technical Output

The Health and System report screens show the current operating state. The Troubleshoot screen keeps the Doctor's detailed diagnostic output collapsed by default.

1. Run **Check system**.
2. Select **Show technical output**.
3. Read or copy the checks listed there.
4. Share the copied output with support if the automated repair and follow-up check do not resolve the issue.

The output stays on your device unless you copy and share it.

---

## Getting Help

1. Run **Check system** and **Apply repairs** from the Troubleshoot screen.
2. Copy the technical output if the follow-up check fails.
3. Contact support through the chat widget in the desktop app.

---

### [Voice](https://neotask.ai/docs/neotask-gateway/voice)

## Start A Voice Session

1. Open **Settings → Voice** and choose the input device.
2. Confirm that the microphone status is ready.
3. Start voice mode from the agent chat.
4. Review any action that pauses for approval on screen.

![Wake word and microphone status ready for voice control](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/docs/product/2026-08-13-r2/microphone-ready-1280.webp)

## Overview

Neotask supports voice interaction through multiple systems: wake word detection, continuous voice conversation (talk mode), and text-to-speech for spoken responses.

## Wake Words

### Swabble (macOS)

Swabble is a native macOS daemon that provides always-on, **on-device** voice wake word detection using Apple's Speech.framework.

**Features:**
- Local-only processing, no audio leaves your device during wake word detection
- Customizable wake words
- Continuous audio capture and transcription
- Hook execution, triggers shell commands when the wake word is detected
- File transcription, convert audio files to text (TXT or SRT format)
- Configurable cooldown, minimum character count, and timeout

**How it works:**
1. Swabble listens continuously using the system microphone
2. When it detects the wake word in spoken text, it captures the following speech
3. The captured text is sent to your agent via a configured hook command
4. The agent processes the voice command and responds

### Node Wake Words

On iOS and Android companion apps, voice wake is handled natively:

- Wake word configuration is owned by the Gateway
- Nodes receive wake word config on connect
- Detection uses platform-native speech recognition

## Talk Mode

Talk mode enables continuous voice conversations, speak naturally and hear your agent respond.

### How It Works

1. **Speech-to-Text**, Your voice is transcribed in real-time (Deepgram streaming or platform-native STT)
2. **Agent Processing**, The transcribed text is sent to your agent as a regular message
3. **Text-to-Speech**, The agent's response is spoken back to you

### Voice State Machine

Talk mode transitions between four states:

| State | Description |
|-------|-------------|
| **Idle** | Not actively listening |
| **Listening** | Capturing and transcribing your speech |
| **Thinking** | Agent is processing your request |
| **Speaking** | Agent response is being spoken |

### Text-to-Speech Providers

| Provider | Description |
|----------|-------------|
| **ElevenLabs** | High-quality voice synthesis with voice selection |
| **OpenAI TTS** | OpenAI's text-to-speech API |

### Voice Preferences

- **Voice selection**, Choose from available TTS voices
- **Custom system prompt**, Override the agent's personality for voice mode
- **Custom response format**, Control how the agent formats spoken responses
- **Language support**, Voice strings localized for 18+ languages

## Voice Commands

### Multi-Intent Detection

Agents can detect and execute multi-step voice commands:

> "Create a calendar event for tomorrow at 3 PM, then send an email to the team about it, and post a reminder in Slack"

This is automatically parsed into a sequence of commands, each executed in order with the results flowing to the next step.

### Tool Execution

The voice briefing keeps the current action, supporting context, and next steps visible while the session runs.

![Voice briefing session with governed next actions](https://neotask-marketing-assets-417007889150.s3.us-east-1.amazonaws.com/landing/agents/2026-08-12-r1/voice-briefing-1280.webp)

During voice conversations, agents can execute tools just like in text conversations, browse the web, run code, manage files, control devices, and more. Results are summarized and spoken back.

### Action Truth Enforcement

Voice mode includes validation that agent claims match actual tool outcomes. If an agent says "I've sent the email" but the email tool failed, the system catches the discrepancy and reports the actual result.

## Voice Calling (Plugin)

The Voice Call plugin adds SIP telephony support:

- Inbound call handling
- Outbound calls (provider-dependent)
- Real-time bidirectional audio (PCM streams)
- TTS synthesis injected into the call audio

## Quota Management

Voice services may have usage quotas:

- Monthly minute allocation for TTS + STT
- Per-session tracking
- Warning at 80% usage
- Automatic cutoff at quota limit

---

## Blog Articles



### [AI workflow automation: what changes when the workflow can think](https://neotask.ai/blog/ai-workflow-automation)

AI workflow automation covers three different setups that vendors all label the same way, and the difference matters more than the marketing suggests. A classic trigger-action pipeline moves data between apps on rails someone wired up in advance, with no model involved anywhere in the run. A pipeline with an AI step keeps those same rails but adds one node, usually an LLM, that summarizes an email, classifies a lead, or drafts a reply at a single fixed point in an otherwise scripted sequence. Genuinely agentic automation is a different kind of thing: the AI decides which steps to take at runtime, given a goal and the current state of your systems, so it can handle a case nobody scripted for. If your team already runs a stack of Zapier-style automations, the honest starting point is sorting what you have, not a full rebuild. Keep the reliable pipes exactly as they are, and hand only the ones that keep breaking to something that can use judgment.

## What is AI workflow automation, exactly?

A pipeline is a fixed sequence: a trigger fires (a new row in Google Sheets, a new email in Gmail, a new deal stage in HubSpot), and each subsequent step runs exactly as configured, no matter what the data looks like. Zapier, Make, and n8n are built around this model, and they are good at it: predictable data movement between apps, running quietly for years once set up correctly.

A pipeline with an AI step is the same fixed sequence with one node replaced by a call to a language model. The model might read an inbound message and pick a category, summarize a document into a few bullet points, or draft a reply for a human to approve. The rest of the pipeline does not change: the trigger fires the same way, the surrounding steps run in the same order, and the AI node has one job with no ability to change the plan around it.

Agentic automation removes the fixed sequence entirely. You give it a goal in plain language, for example "when a payment fails, try to recover it and keep the books straight," and the agent decides what to do given the current state of the customer and the tools available: check history, choose a different path for a first-time failure than a repeat one, retry, escalate, or close the loop, without anyone drawing that branching logic in advance.

![Diagram: three tiers of AI workflow automation stacked, from a plain pipeline to a pipeline with an AI step to fully agentic automation, each labeled with who decides the steps and what happens on a surprise](data:image/svg+xml;base64,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)

## How is a pipeline with an AI step different from a plain pipeline?

On the setup screen, not much, which is part of why vendors sell all three tiers under one name. The real difference shows up on a bad day. When step three of a plain pipeline gets a value it was not built to handle (an empty field, an unrecognized currency, a duplicate it should have caught) the pipeline errors out and someone gets paged to fix it. Adding an AI step does not change that. The model absorbs variance inside its own node, since it can summarize a message it has never seen the exact shape of before, but the rails around it stay fixed. If the step after the AI node expects a clean category and gets something else, the pipeline still breaks in the same place.

That is a real improvement for the one job the AI step does. It is a separate thing from a system that can change its own plan when the surprise is bigger than a single field.

## What does genuinely agentic automation look like running?

Take a concrete example: a support ticket comes in through a helpdesk tool, and the instruction is "triage it and resolve what you can." A plain pipeline routes the ticket by keyword and assigns it to a queue. A pipeline with an AI step can also draft a suggested reply for a human to send. An agent goes further: it reads the ticket, checks the customer's account and order history, and decides whether the issue is answerable from documentation, from account data, or needs a human. A straightforward billing question with a clear answer in the account record gets a reply and a closed ticket. An ambiguous complaint gets an assembled summary of the relevant account details, handed to a human so the first touch starts informed instead of from a blank ticket.

Nobody wrote a rule for every category of complaint in advance. The agent worked out the path from the goal and the data in front of it, which is the entire point of paying for judgment instead of just data movement.

## Which tier is the right choice for a given job?

Match the tool to how much the job actually varies, not to how advanced the label sounds.

A plain pipeline is right for pure data movement with close to zero variance: syncing a new customer into your CRM, posting a Slack message when a form is submitted, copying rows between two sheets. There is no judgment to make, so paying for judgment is waste.

A pipeline with an AI step earns its place when you need one specific enrichment, classification, or draft inside an otherwise stable process, and the surrounding steps really are stable: auto-tagging inbound leads by intent before a fixed follow-up sequence, or summarizing a long thread before it lands in a ticket queue.

Agentic automation is worth the setup when the work recurs, requires a judgment call partway through, and the inputs genuinely vary run to run: usually the work a person still does by hand, precisely because a rigid tool could not handle the exceptions. If nobody on your team can describe "the normal case" for a piece of work in one sentence, the exceptions are the actual job.

## What does each tier cost you over a year?

Cost shows up differently for each, and the differences compound over months, not days.

A plain pipeline is cheap to build and run, but carries a quiet maintenance tax: every time a connected app changes a field name or a vendor updates its API, something breaks and nobody notices until a report is missing. Teams running dozens of these usually have one person whose real job, unofficially, is watching for broken zaps.

A pipeline with an AI step carries that same tax, plus a second one on the AI node itself: prompts drift as the underlying model changes, per-call cost accumulates with volume, and the output still needs occasional spot-checking, because it is real judgment inside a box built for fixed output.

Agentic automation front-loads cost differently. Setup takes longer, because scoping permissions and defining what the agent can do without approval takes real thought. After that, the ongoing cost shifts from someone fixing a broken step to someone reviewing an exception queue, and that queue should get shorter as trust and scope widen with a track record.

## How do you migrate a pile of existing zaps without a rebuild?

You do not need to touch anything that already works. The practical migration path starts with a short audit, not a rebuild plan.

List the automations that break most often: the ones with a support rotation, a Slack channel for "the zap is down again," or a step someone manually re-runs most weeks. Those are your first candidates for an agent, precisely because a fixed pipeline was never well suited to them. Everything that has run quietly for a year stays exactly as it is; moving it would only add risk for no benefit.

Move the worst offenders one at a time. Give the agent the same integrations the old zap used, state the outcome in plain language instead of redrawing the flowchart, and run it alongside the old pipeline for a stretch before retiring it. Widen from there as each migrated workflow proves out.

![Diagram: a decision path sorting existing automations into keep as pipes versus hand to an agent, with the sorting question of whether the work breaks or needs judgment in the middle](data:image/svg+xml;base64,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)

## How do you measure whether it actually worked?

Track the same three numbers before and after, on the same workflow, using your own system's real activity, never a number pulled from somewhere else.

Exception rate is the share of runs that needed a human to step in at all. A falling exception rate on a workflow that used to page someone weekly is the clearest signal the change is working.

Human touches per run counts how many separate manual actions a case required from trigger to close. A ticket that used to need someone to read it, check an account, draft a reply, and send it, now needing only an approval, is a real reduction you can count.

Time-to-resolution measures how long a case takes from the trigger firing to the work actually being done, not just acknowledged. That is the number a customer or a manager actually feels.

Pull these from your own logs before you change anything, then again after a real stretch of runs. If a number does not move, that tier was probably not the bottleneck, so look at a different workflow rather than declaring victory anyway.

## How would you actually start with an agent instead of another zap?

Neotask is built for the agentic tier specifically: a desktop app and gateway that connects to the real apps your team already uses through integrations and MCP, takes a goal in plain language, and plans and executes the work through real tool calls, the same ones a person would use by hand. Every action runs under permissions you set, sensitive steps can require approval first, and each run keeps a step-level log, so reviewing what the agent did takes minutes, not an investigation.

The candidates from your own audit (the zap that breaks most, the ticket queue triaged by hand every morning, the reconciliation someone still does manually) are a reasonable place to start. Connect the same apps, state the outcome instead of the steps, and leave the reliable pipes where they are.

## Frequently asked questions

**Is AI workflow automation the same thing as agentic automation?**
Not exactly. AI workflow automation is the broader label vendors use for anything from a plain pipeline with an AI step bolted on to a fully agentic setup. Agentic automation is the specific case where the AI decides the steps at runtime.

**Do I need to replace Zapier, Make, or n8n to use agentic automation?**
No. Most teams run both at once. Predictable data movement stays on pipes, because pipes are cheap to run and good at that job. The jobs that keep breaking or need a judgment call move to an agent.

**How do I tell if a workflow needs a pipeline with an AI step or full agentic automation?**
Check how many decision points the work has. One enrichment or classification step inside an otherwise stable process fits an AI step. Multiple judgment calls and a recurring pattern of exceptions fit an agent.

**What is the biggest risk in moving a workflow to an agent?**
Under-scoped permissions, more than the AI itself. Start with broad read access and narrow write access, gate irreversible actions like refunds or deletions behind approval, and widen scope as the step-level log shows good calls.

**How long does it take to see results after migrating a workflow?**
Enough real runs to compare against a baseline, not a fixed number of days. Track exception rate, human touches per run, and time-to-resolution before the change, then again on the same workflow after a real stretch of activity.

---

### [Human in the loop AI: how to stay in control of what an agent can do](https://neotask.ai/blog/human-in-the-loop-ai)

Human in the loop, for an AI agent, means a person still approves or reviews the actions that matter, while the agent decides and executes everything else on its own. In practice that means three boundaries: what the agent can read without asking, what it can write or execute on its own, and which actions always stop for a person no matter how much trust the agent has earned. Get those three right and an agent can run most of a workday unsupervised while nothing irreversible, external-facing, or financial happens without a human sign-off.

Most explanations of human in the loop come from machine learning, where it describes people labeling training data or grading model outputs. That is a real practice, but it is a different question from the one that matters once an agent runs inside your business systems, taking actions in your calendar, your CRM, your inbox. There, the term is an operational design question: which actions need a person to see them before or after they happen, and how that boundary moves as the agent proves itself.

## What does human in the loop actually mean for an AI agent?

An agent running on your systems reads the current state, decides what to do, and acts through real tools: sending a message, updating a record, moving a file. Human in the loop is the decision about which of those actions happen freely and which pause for a person. It is not one switch. A well-run setup is a spectrum: routine reads happen with zero friction, most writes happen on their own once the agent has a track record, and a small set of actions always waits for approval regardless of history.

The mistake teams make early is treating this as one toggle, either fully supervised, every action needs a click, or fully autonomous, nothing does. Neither survives real work. Full supervision turns the agent into an expensive way to generate approval requests. Full autonomy on day one means the first mistake happens with no one watching. The workable answer sits in between, and where it sits should be a decision you make on purpose, not a default you accept.

## What are the three real control points?

Three levers determine how much control you actually have, and each does a different job.

**Scope of tools.** Which systems and actions the agent can even reach. An agent connected to your calendar, CRM, and support inbox has a different blast radius than one also connected to your payment processor and production database. Scope is the cheapest control: an agent cannot do what it has no tool to do.

**Approval gates.** Within the tools it has, which specific actions require a yes before they run. Most of the day-to-day tuning happens here.

**Review of logs.** After the fact, can a person see exactly what the agent did and why. This one is easy to underrate. Logs are how you catch a bad pattern before it repeats, and how you build the evidence that lets you loosen the other two levers.

Most failures trace back to setting only one of these. A team that scopes tools carefully but never reviews logs has no real idea whether the agent's judgment is good, they are just hoping. A team that gates everything but never narrows scope ends up with a queue full of harmless reads mixed in with the one action that mattered.

![Diagram: three zones of agent permission, broad read access, narrow write access, and an inner zone of irreversible actions that always requires approval](data:image/svg+xml;base64,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)

## What's the read-broad, write-narrow rule?

The simplest working rule for scope: let the agent read broadly, and let it write narrowly. Reading is how an agent builds context, and context is what makes its plans good. An agent that can see your whole calendar, CRM history, and support queue makes better decisions than one working from a narrow slice, because it has fewer blind spots to reason around. Reading rarely destroys anything, so being generous here costs little.

Writing is the opposite case. Every write changes a real system, and that is what you have to walk back if the agent gets something wrong. Start narrow: a specific set of record types, fields, and destinations. Widen write scope only in categories where the agent has shown good judgment, not everywhere at once just because it has been fine in the areas you happened to check.

## Which actions should always require a human, no matter what?

Some categories of action stay gated permanently, no matter how long the agent has run cleanly. Three shapes of action belong on that list:

- **Anything irreversible.** Deleting a record, an account, or a file. If there is no undo, a person looks first.
- **Anything external-facing.** A message to a customer, a public post, an email that leaves the building. Internal mistakes are cheap to fix. External ones carry your name and cannot be pulled back once seen.
- **Anything touching money or deletion.** Refunds, payouts, price changes, and permission changes that affect who can access what.

The reasoning has nothing to do with how smart the agent is. It comes down to the cost of being wrong. A reversible internal mistake costs a few minutes to fix. An external message or a deleted record can cost a customer relationship or a day of recovery work. Gate by consequence, not by how often the agent tends to get it right.

## Why doesn't approving everything work?

The instinct to gate every action feels safe, and it fails within a week, for reasons that are mechanical rather than a discipline problem.

People rubber-stamp. When someone sees the fortieth approval request in a row that reads "read three emails and draft a reply," they stop reading it carefully and start clicking approve on reflex. The gate is still there, but it has stopped doing its job.

The agent becomes useless for the thing you built it for. Its value is acting without you driving every step. If every step needs a click, you have rebuilt a slower version of doing the work yourself.

The queue backs up. Approvals pile up while you are in a meeting or asleep, and the agent sits idle waiting on the one item that actually needed attention, buried under a dozen that never should have been gated.

The fix is not fewer gates in general. It is fewer categories gated, applied more precisely, so the requests that do show up genuinely deserve a look.

## How do you widen an agent's autonomy safely?

Widen scope on evidence, not on comfort level. A run of clean outcomes in a specific category earns that category more room, not the fact that a week passed without an incident.

A workable rhythm: start a category gated, and let it run long enough to build a real sample. Review the log for that category: did the agent's read of the situation match reality, did the action match what a competent person would have done, were there any near misses even if nothing broke. If the pattern holds, move that category to auto-approve and keep watching its logs at a lower frequency. If it does not hold, tighten the gate and figure out why before trying again.

This is close to how you would manage a new hire's authority: broad access to information from day one, a narrow set of things they can do without checking in, and more responsibility as their judgment proves out on what they have actually done.

![Diagram: a trust ramp over time, starting fully gated, widening as log evidence accumulates, with irreversible actions staying gated at every stage](data:image/svg+xml;base64,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)

## What does good step-level logging look like?

Logging is the control point that makes the other two sustainable, and it only works if the log is built for a fast read, not an investigation. A log entry worth trusting shows, for each step, what the agent was trying to do, what it read to decide that, the exact tool call it made, and the result, not a paraphrase written after the fact.

That level of detail sounds heavier than it is. A reviewer can open yesterday's run for a category, scan ten or fifteen entries in a minute, and answer the only question that matters: would a competent person have done the same thing, given what the agent saw. If yes across the board, that is the evidence for widening scope. If one entry looks off, the digging is fast, because the log points at exactly where to look.

Vague logging, "sent an email," "updated a record," forces a reviewer to check the actual system to find out what happened. Specific logging, the literal recipient, subject, and body of that email, means the review happens inside the log itself, which is what keeps review sustainable instead of something everyone quietly stops doing.

## How do you set this up with an AI agent like Neotask?

In Neotask, these three levers are how you would actually configure an agent for real work. You connect the specific apps involved, your inbox, CRM, or accounting software, through the platform's integrations and MCP connections, so scope starts as the specific list of systems in play rather than everything at once. Within those connections, you set which actions run on their own and which wait for approval, and the categories above, irreversible, external-facing, money-touching, are the ones worth gating from day one. Every step the agent takes is logged, so reviewing a category before widening it means reading an actual record instead of guessing. The setup is a starting position you adjust as the log gives you real evidence about where the agent has earned more room.

## Frequently asked questions

**What does human in the loop mean for AI agents specifically, as opposed to machine learning?**
In machine learning it usually means people labeling data or reviewing model outputs during training. For an agent running in your business systems, it means deciding which live actions need a person's approval or review, and which the agent can carry out on its own.

**Is human in the loop the same thing as requiring manual approval of everything an agent does?**
No, and treating it that way tends to backfire. Full approval on every action leads to rubber-stamping, a backed-up queue, and an agent slower than doing the work yourself. The useful version gates specific high-consequence categories and lets the rest run.

**What actions should never be automated, even for a trusted agent?**
Anything irreversible, like deletions, anything external-facing, like a message to a customer, and anything touching money or account permissions. These stay gated regardless of how long the agent has run cleanly, because the cost of a rare miss is high.

**How do you decide when to give an agent more autonomy?**
On evidence from its logs, not a feeling that enough time has passed. Review a category's history, confirm the pattern of good decisions holds over a real sample, then widen that category specifically.

**What should an agent's activity log actually include?**
The exact tool call, what it read before deciding, and the real result, not a summary written after the fact. That detail is what lets a review take a minute instead of a full investigation.

---

### [What is an AI agent? A plain-English definition with a real example](https://neotask.ai/blog/what-is-an-ai-agent)

An AI agent is software that takes a goal you give it in plain language, decides for itself what steps get there, and acts by calling real tools in your actual apps, then checks the result and adjusts if the first attempt does not land. That is the whole difference from a chatbot, which only answers inside its own window and never touches anything else, and from a regular automation tool, which runs a fixed sequence of steps someone wrote down in advance and breaks the moment reality strays from that script. For a business, the payoff is direct: an agent can read your calendar, pull a customer's order history, send the email, update the record, and report back that it is done, without a person wiring each of those steps together by hand first.

## How do you tell an agent apart from a chatbot or a fixed automation?

Run whatever you are looking at through one test: does it decide its own steps at runtime, and does it act through real tools? If the answer to both is yes, you are looking at an agent. If either answer is no, you are looking at something else wearing an agent-shaped label.

A chatbot passes neither half. It answers the question in front of it, inside the chat window, and the conversation ends there. Ask it to also update a spreadsheet and it cannot, because it has no way to reach outside the chat.

A traditional automation tool passes half. Zapier, Make, and n8n all act through real tools: a workflow can genuinely create a row in a spreadsheet or send a Slack message. What it cannot do is decide. A person designed the exact sequence of steps ahead of time, step one then step two then step three, and the tool executes that sequence faithfully every single run. Change the input in a way the designer did not plan for, a new field, a reply with no invoice number attached, and the workflow either fails or does the wrong thing, because there was never any judgment built into it, only a path.

An AI agent passes both halves. Given the goal "follow up on this customer's support ticket," it looks at the actual ticket, decides what the next right step is based on what it finds, maybe a clarifying question, maybe an order-status check, maybe an escalation, and then executes that step by calling a real tool: sending the email, updating the ticket, notifying a person. Next ticket, the steps might be completely different, because the agent decided them from the specific situation instead of following a diagram drawn last quarter.

![Diagram contrasting a chatbot and a fixed automation, both grayed out, against an AI agent lane that decides its own steps and acts through real tools, with a test box asking whether both conditions hold](data:image/svg+xml;base64,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)

That test also explains the mess in this space right now. Plenty of products that run one hardcoded path with a language model bolted onto step three for text generation now call themselves AI agents, because the label sells. Run the test on them and the answer to "does it decide its own steps" is no, it does a fixed thing that happens to include a language model. That is automation with a language model in it, a fine product, just not an agent.

## What does an AI agent actually do, step by step?

Abstract definitions only get you so far. Here is one agent doing one real job, from goal to finished.

The goal, stated in plain language: "When a new lead comes in from the website form, qualify it and get it into the pipeline properly." Nobody wrote out the individual steps. The agent works those out itself, every time it runs.

First it reads real data: the form submission, the lead's company domain, whatever history already exists for that email address in HubSpot, and any prior email thread in the inbox. Second, it decides: is this a real fit for what the business sells, is it a duplicate of an existing contact, does it look like a competitor or a student project rather than a buyer. That judgment call is different for every lead, which is exactly why a fixed script cannot make it well.

Third, it acts, through real tools in real apps: it drafts and sends a reply that references the lead's actual stated need instead of a generic template, creates or updates the HubSpot record with the right lifecycle stage, and adds a task for the right sales rep if the lead clears the bar. Fourth, it verifies: it confirms the CRM record actually saved with the fields it meant to set, and it logs exactly what it did and why, so a person reviewing the account later sees the reasoning, not just an outcome.

![Diagram showing the anatomy of one agent run: goal, then reads real data, then decides, then acts in real apps, then verifies, captioned with the product action of connecting HubSpot and Gmail and stating the goal](data:image/svg+xml;base64,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)

Nobody drew that decision tree in a flowchart tool beforehand. The agent produced it from the situation in front of it, which is the entire point of using one instead of scripting one.

## Why does nobody seem to agree on what an AI agent is?

Search around and you will find a popular complaint: nobody agrees what the term means, and vendors slap "agent" on everything from a chatbot with a plugin to a full autonomous system. The complaint is fair, and it comes from a real cause, not just marketing habit.

The term got adopted before the industry settled on a shared bar for what clears it. "Agent" sounds better in a sales deck than "scripted integration with a language model," so plenty of products picked up the word without picking up the behavior. A chatbot that can browse the web still just answers in a window. A workflow builder that added an AI step to summarize an email in the middle of an otherwise fixed pipeline is still a fixed pipeline. Both get called agents in marketing copy.

The fix is not a better definition, more adjectives usually make it murkier. The fix is the test from earlier in this piece: does it decide its own steps at runtime, and does it act through real tools. Apply that consistently and the confusion mostly clears, because the test does not care what a landing page calls the product.

## What is an AI agent not?

Being honest about the edges matters more than the marketing version of this answer. An AI agent is not a chatbot with extra steps: a support bot that answers FAQs and occasionally opens a ticket for a person is still mostly a chatbot unless it is actually deciding and acting across a real range of situations on its own. It is not a guarantee of correctness: an agent that decides its own steps can also decide the wrong one, so the ones worth trusting come with scoped permissions, an approval gate on anything irreversible like a refund or a deletion, and a log of every action taken. It is not something that needs zero setup: it still needs your real apps connected so it has something to act on, and it still needs a clearly stated goal, vague instructions produce vague results from a person too. And it is not always the right tool: work that is pure, unvarying data movement with no judgment involved is often served fine by a simple automation, and building an agent for that is over-engineering in the other direction.

## What can an AI agent actually do for your business?

The lead-qualification example above is one shape of a broader pattern: any recurring job that needs your real data read, a judgment call made, and an action taken in your actual tools is a candidate. Triaging support tickets, following up on overdue invoices, reconciling a payment failure, running a morning check across your project tools and flagging what needs attention, drafting and routing a first response to an inbound sales lead.

In Neotask, this runs the way the definition above describes it, on purpose. You connect the real apps involved: a CRM, an inbox, a support desk, a finance tool, whatever the job touches. You state the goal in plain language, the way you would hand it to a new hire. The agent plans the steps at runtime against your actual current data, acts through the connected tools, and shows its work at every step, with permissions scoped to what it should touch and an approval gate on anything irreversible. That is the agent half of the test from earlier, running against your business instead of a diagram.

## Frequently asked questions

**Is ChatGPT an AI agent?**
The base chat product is not: it answers inside its window and does not act on outside systems on its own. Give a language model tools it can call and let it decide when and how to use them across multiple steps, and that combination is an agent. Several products built on top of language models, including ones from OpenAI itself, do meet that bar.

**Is Zapier an AI agent?**
Zapier is a real, well-built automation tool: it acts through genuine integrations, but a standard workflow follows a fixed sequence a person configured in advance rather than deciding its own steps. Zapier has also been building agent-style products alongside its core automation, worth judging separately against the same test.

**Do I need to know how to code to use an AI agent?**
No. The setup is connecting the apps involved and stating the goal in plain language, the same way you would brief a person on a task. The planning and tool-calling happen at runtime, not in a diagram you build yourself.

**Can an AI agent make mistakes?**
Yes, because deciding its own steps means it can decide a wrong one. That is exactly why scoped permissions, approval gates on irreversible actions, and full step logs matter: they let you catch and correct a bad decision quickly instead of discovering it after the fact.

**What is the single fastest way to check if something is really an agent?**
Ask whether it decides its own steps at runtime and whether it acts through real tools in real apps. If both are true, it is an agent. If only one is true, it is a chatbot or an automation wearing the label.

---

### [n8n alternatives: what to use when you want the outcome, not the diagram](https://neotask.ai/blog/n8n-alternatives)

If you want AI agents that handle the work instead of a workflow diagram you keep building and patching, the field breaks into four real options: managed visual builders like Make and Zapier if you still want to draw the flow but skip the hosting, developer frameworks like LangGraph and CrewAI if your team wants to write the agent logic in code, agent platforms like Lindy, Relay.app, Gumloop, and Neotask if you want to state the outcome in plain language and let the agent handle the variance, or staying on n8n itself if the self-hosted, code-adjacent control was the actual reason you picked it. n8n is a strong tool with a real fan base, so the right alternative depends on which specific part of the n8n experience you are trying to get away from.

## Why do people go looking for an n8n alternative?

n8n earns its popularity honestly. It is self-hostable, so you can run it on your own infrastructure and keep the data path entirely yours. Its licensing is source-available in a way people describe as open-source-ish: you can see the code and self-host it, even though it is not a plain permissive license like MIT or Apache. It ships one of the largest node libraries of any workflow tool, plus a generic HTTP node and code nodes for anything that is not already built, and it has an active community that shares templates and answers questions fast. None of that is in question.

What sends people searching for an alternative is usually one specific friction, not a rejection of the whole idea. The node graph that felt manageable at ten steps turns into a maintenance job at eighty, with every new condition adding another branch and every branch adding another way the workflow can be wrong when the input looks slightly different. If you did not want to run infrastructure, self-hosting turns into a second job: updates, uptime, backups, and a server that is now your problem. The builder itself has a real learning curve; expressions and item structures make sense once you know them but are not obvious on day one. And a workflow built as a fixed set of branches has no way to use judgment on a case nobody drew a path for. It stalls, throws an error, or quietly does the wrong thing, and someone has to notice.

![Diagram comparing a workflow diagram you have to maintain, with branch explosion and an owner keeping it working, against a single stated outcome that an agent loop executes and adapts on its own](data:image/svg+xml;base64,PHN2ZyB4bWxucz0iaHR0cDovL3d3dy53My5vcmcvMjAwMC9zdmciIHZpZXdCb3g9IjAgMCA3NjAgMzgwIiBmb250LWZhbWlseT0idWktbW9ub3NwYWNlLCBTRk1vbm8tUmVndWxhciwgTWVubG8sIG1vbm9zcGFjZSI+CiAgPHJlY3Qgd2lkdGg9Ijc2MCIgaGVpZ2h0PSIzODAiIGZpbGw9IiNmN2YzZWMiLz4KICA8ZGVmcz4KICAgIDxtYXJrZXIgaWQ9Im4xYSIgdmlld0JveD0iMCAwIDEwIDEwIiByZWZYPSI5IiByZWZZPSI1IiBtYXJrZXJXaWR0aD0iNyIgbWFya2VySGVpZ2h0PSI3IiBvcmllbnQ9ImF1dG8iPgogICAgICA8cGF0aCBkPSJNMCAwTDEwIDVMMCAxMHoiIGZpbGw9IiNjMjQxMGMiLz4KICAgIDwvbWFya2VyPgogICAgPG1hcmtlciBpZD0ibjFnIiB2aWV3Qm94PSIwIDAgMTAgMTAiIHJlZlg9IjkiIHJlZlk9IjUiIG1hcmtlcldpZHRoPSI3IiBtYXJrZXJIZWlnaHQ9IjciIG9yaWVudD0iYXV0byI+CiAgICAgIDxwYXRoIGQ9Ik0wIDBMMTAgNUwwIDEweiIgZmlsbD0iI2E4YTI5ZSIvPgogICAgPC9tYXJrZXI+CiAgPC9kZWZzPgogIDx0ZXh0IHg9IjE2IiB5PSIyNiIgZm9udC1zaXplPSIxNCIgZm9udC13ZWlnaHQ9IjcwMCIgZmlsbD0iIzdjMmQxMiI+VEhFIERJQUdSQU0gWU9VIE1BSU5UQUlOPC90ZXh0PgogIDx0ZXh0IHg9IjQyMCIgeT0iMjYiIGZvbnQtc2l6ZT0iMTQiIGZvbnQtd2VpZ2h0PSI3MDAiIGZpbGw9IiM3YzJkMTIiPlRIRSBPVVRDT01FIFlPVSBTVEFURTwvdGV4dD4KCiAgPGcgZmlsbD0iI2YwZWNlNiIgc3Ryb2tlPSIjYThhMjllIiBzdHJva2Utd2lkdGg9IjEuNSI+CiAgICA8cmVjdCB4PSIxMCIgeT0iMTQwIiB3aWR0aD0iMTAwIiBoZWlnaHQ9IjQwIiByeD0iMTIiLz4KICAgIDxyZWN0IHg9IjE1MCIgeT0iNjAiIHdpZHRoPSIxMTAiIGhlaWdodD0iMzYiIHJ4PSIxMiIvPgogICAgPHJlY3QgeD0iMTUwIiB5PSIxNDIiIHdpZHRoPSIxMTAiIGhlaWdodD0iMzYiIHJ4PSIxMiIvPgogICAgPHJlY3QgeD0iMTUwIiB5PSIyMjQiIHdpZHRoPSIxMTAiIGhlaWdodD0iMzYiIHJ4PSIxMiIvPgogICAgPHJlY3QgeD0iMzAwIiB5PSI0MCIgd2lkdGg9IjEwMCIgaGVpZ2h0PSIzMiIgcng9IjEyIi8+CiAgICA8cmVjdCB4PSIzMDAiIHk9IjgwIiB3aWR0aD0iMTAwIiBoZWlnaHQ9IjMyIiByeD0iMTIiLz4KICA8L2c+CiAgPGcgZmlsbD0ibm9uZSIgc3Ryb2tlPSIjYThhMjllIiBzdHJva2Utd2lkdGg9IjEuNSI+CiAgICA8cGF0aCBkPSJNMTEwIDE2MCBMMTUwIDc4IiBtYXJrZXItZW5kPSJ1cmwoI24xZykiLz4KICAgIDxwYXRoIGQ9Ik0xMTAgMTYwIEwxNTAgMTYwIiBtYXJrZXItZW5kPSJ1cmwoI24xZykiLz4KICAgIDxwYXRoIGQ9Ik0xMTAgMTYwIEwxNTAgMjQyIiBtYXJrZXItZW5kPSJ1cmwoI24xZykiLz4KICAgIDxwYXRoIGQ9Ik0yNjAgNzggTDMwMCA1NiIgbWFya2VyLWVuZD0idXJsKCNuMWcpIi8+CiAgICA8cGF0aCBkPSJNMjYwIDc4IEwzMDAgOTYiIG1hcmtlci1lbmQ9InVybCgjbjFnKSIvPgogIDwvZz4KICA8ZyBmaWxsPSIjNzg3MTZjIiBmb250LXNpemU9IjExIiBmb250LXdlaWdodD0iNzAwIiB0ZXh0LWFuY2hvcj0ibWlkZGxlIj4KICAgIDx0ZXh0IHg9IjYwIiB5PSIxNjQiPlRSSUdHRVI8L3RleHQ+CiAgICA8dGV4dCB4PSIyMDUiIHk9IjgyIj5GSUVMRCBBPC90ZXh0PgogICAgPHRleHQgeD0iMjA1IiB5PSIxNjQiPkZJRUxEIEI8L3RleHQ+CiAgICA8dGV4dCB4PSIyMDUiIHk9IjI0NiI+RUxTRTwvdGV4dD4KICAgIDx0ZXh0IHg9IjM1MCIgeT0iNjAiPlNURVAgMUE8L3RleHQ+CiAgICA8dGV4dCB4PSIzNTAiIHk9IjEwMCI+U1RFUCAxQjwvdGV4dD4KICA8L2c+CiAgPHRleHQgeD0iMTUwIiB5PSIyOTAiIGZvbnQtc2l6ZT0iMTEiIGZpbGw9IiM3ODcxNmMiPmV2ZXJ5IG5ldyBmaWVsZCBpcyBhbm90aGVyIGJyYW5jaDwvdGV4dD4KICA8dGV4dCB4PSIxNTAiIHk9IjMwNiIgZm9udC1zaXplPSIxMSIgZmlsbD0iIzc4NzE2YyIgZm9udC13ZWlnaHQ9IjcwMCI+c29tZW9uZSBvd25zIHRoaXM8L3RleHQ+CgogIDxnIGZpbGw9IiNmYWY2ZjAiIHN0cm9rZT0iI2MyNDEwYyIgc3Ryb2tlLXdpZHRoPSIxLjUiPgogICAgPHJlY3QgeD0iNDcwIiB5PSI2MCIgd2lkdGg9IjE4MCIgaGVpZ2h0PSI1MCIgcng9IjEyIi8+CiAgICA8cmVjdCB4PSI0NzAiIHk9IjE1MCIgd2lkdGg9IjE4MCIgaGVpZ2h0PSI2MCIgcng9IjEyIi8+CiAgICA8cmVjdCB4PSI0NzAiIHk9IjI1MCIgd2lkdGg9IjE4MCIgaGVpZ2h0PSI0NCIgcng9IjEyIi8+CiAgPC9nPgogIDxnIGZpbGw9Im5vbmUiIHN0cm9rZT0iI2MyNDEwYyIgc3Ryb2tlLXdpZHRoPSIyIj4KICAgIDxwYXRoIGQ9Ik01NjAgMTEwIEw1NjAgMTUwIiBtYXJrZXItZW5kPSJ1cmwoI24xYSkiLz4KICAgIDxwYXRoIGQ9Ik01NjAgMjEwIEw1NjAgMjUwIiBtYXJrZXItZW5kPSJ1cmwoI24xYSkiLz4KICAgIDxwYXRoIGQ9Ik02NTAgMTYwIEMgNzAwIDE2OCwgNzAwIDIwMiwgNjUwIDIxMCIgbWFya2VyLWVuZD0idXJsKCNuMWEpIiBzdHJva2UtZGFzaGFycmF5PSI0IDMiLz4KICA8L2c+CiAgPGcgZmlsbD0iIzdjMmQxMiIgZm9udC1zaXplPSIxMyIgZm9udC13ZWlnaHQ9IjcwMCIgdGV4dC1hbmNob3I9Im1pZGRsZSI+CiAgICA8dGV4dCB4PSI1NjAiIHk9IjkwIj5PTkUgSU5TVFJVQ1RJT048L3RleHQ+CiAgICA8dGV4dCB4PSI1NjAiIHk9IjE4NCI+QUdFTlQgTE9PUDwvdGV4dD4KICAgIDx0ZXh0IHg9IjU2MCIgeT0iMjc2Ij5ET05FPC90ZXh0PgogIDwvZz4KICA8ZyBmaWxsPSIjNzg3MTZjIiBmb250LXNpemU9IjEwIiB0ZXh0LWFuY2hvcj0ibWlkZGxlIj4KICAgIDx0ZXh0IHg9IjU2MCIgeT0iMTA0Ij5wbGFpbiBsYW5ndWFnZSwgb25jZTwvdGV4dD4KICAgIDx0ZXh0IHg9IjU2MCIgeT0iMTk4Ij5wbGFucywgYWN0cywgY2hlY2tzPC90ZXh0PgogICAgPHRleHQgeD0iNzA4IiB5PSIxODgiPmhhbmRsZXMgdmFyaWFuY2U8L3RleHQ+CiAgICA8dGV4dCB4PSI1NjAiIHk9IjI5MCI+b3V0Y29tZSArIGV2aWRlbmNlPC90ZXh0PgogIDwvZz4KICA8dGV4dCB4PSIzODAiIHk9IjM0NSIgdGV4dC1hbmNob3I9Im1pZGRsZSIgZm9udC1zaXplPSIxMSIgZm9udC13ZWlnaHQ9IjcwMCIgZmlsbD0iIzlhMzQxMiI+U1RBVEUgVEhFIE9VVENPTUUgT05DRS4gVEhFIEFHRU5UIEhBTkRMRVMgV0hBVCBDSEFOR0VTIEVBQ0ggVElNRS48L3RleHQ+Cjwvc3ZnPgo=)

The loop on the right is the part worth noticing. A node graph has no arrow back into itself: when a branch meets a case nobody drew, it fails and a person gets pulled in to add the missing path. An agent loop treats that same surprise as information, reasons about it, and keeps going. The difference was never about which tool has more nodes. It is about whether the steps are decided in advance or at runtime.

## What if you want the same builder model, just managed for you?

Make and Zapier are the natural move if what you actually want is to keep drawing the flow yourself, just without owning a server. Both are fully managed: no host to patch, no uptime to watch, and a visual canvas that plenty of non-technical users find approachable. Zapier has the largest app directory in the category and the simplest mental model, trigger then action, which makes it an easy on-ramp for a first automation. Make's scenario builder gives you more visual control over branching and data shaping than a simple Zap, closer to n8n's flexibility, while staying fully hosted.

The tradeoff moving from n8n to either one is the same tradeoff you started with, just relocated. You are still drawing a diagram, and it still needs updating when the underlying apps change. You are trading self-hosting overhead for a managed bill and someone else's infrastructure, not trading away the branch-maintenance problem. If your actual complaint was hosting, this solves it. If your actual complaint was the diagram, it does not.

## What if you want to write the agent logic yourself?

LangGraph and CrewAI fit if your team includes engineers who want to build agent behavior in code rather than draw it or state it in plain language. LangGraph, from the LangChain team, lets you define an agent's logic as an explicit state graph in Python or JavaScript: you write the nodes, the edges, and the conditions, and you get full control over how state moves through the loop. CrewAI takes a different shape, an open source Python framework for putting together a team of agents, each with a defined role and goal, that collaborate on a shared task.

Both give you a level of precision a no-code builder cannot: you can inspect and version every decision path, write custom logic for edge cases as you discover them, and integrate directly with your existing codebase. The cost is that you are now maintaining code, not clicking together a canvas, and the people who can extend it are the people who can read Python. If your team already ships code and wants an agent system living inside that codebase, this is the honest choice. If you would rather not build and maintain the agent logic yourself, keep reading.

## What if you want an agent platform that takes the outcome, not the diagram?

Lindy, Relay.app, Gumloop, and Neotask sit in a different category from everything above: you are not drawing a flow and you are not writing code, you are describing what you want done and letting the platform work out the steps.

Lindy leans into building agents for recurring personal and team work, things like triaging an inbox, scheduling meetings, and following up on threads, configured mostly in plain language rather than a node graph. Relay.app keeps a visual workflow layer but adds AI steps alongside it and leans on human-in-the-loop checkpoints, so a person approves the parts that matter before the agent moves on, a middle position between full automation and full autonomy. Gumloop is built for non-engineers who want to assemble AI-powered automations without writing code, closer in spirit to a no-code builder but with AI reasoning steps doing more of the interpretation work than fixed logic.

Neotask fits here too: it is a desktop app and gateway that connects to the apps you already use through integrations and MCP, takes a goal in plain language, and plans and executes the multi-step work through real tool calls, with permission and approval gates and a step-level log of what it actually did. The shared idea across all four is that runtime judgment replaces pre-drawn branches. When an input looks different than expected, the agent reasons about it instead of falling through to an error, which a node graph cannot do no matter how many branches you add to it.

## What if self-hosting is the actual requirement?

If keeping everything on infrastructure you control is the non-negotiable part, for compliance, cost at scale, or simply not wanting a vendor between you and your data, the honest answer is that n8n's self-hosted deployment is still the strongest fit on this list. Lindy, Relay.app, Gumloop, and Neotask all run as managed cloud products, and LangGraph and CrewAI still typically call out to hosted model APIs even when you run the orchestration code yourself. These are managed-first designs by nature, a different deployment model than fully self-hosted infrastructure, and that is a fair description of where the category stands today. If self-hosting is the requirement and not just a preference, staying on n8n and reducing the branch-maintenance load with better workflow hygiene is the more realistic path right now than switching platforms.

![Decision flow for choosing an n8n alternative: managed versus self-hosted, building versus stating the outcome, and whether you need code-level control, ending at the matching category](data:image/svg+xml;base64,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)

## What would switching one workflow actually look like?

Take a single n8n workflow you already have: it watches a form submission, checks a few conditions, and routes the lead to a rep with a Slack message. In n8n, every new condition, a lead from a region you had not planned for, a field that is sometimes blank, a company size the branches did not anticipate, means opening the canvas and adding a path.

Moved to an agent platform like Neotask, the same job becomes an instruction: connect the form tool, the CRM, and Slack once, then tell the agent to qualify incoming leads and route each one to the right rep with context, and to flag anything it is unsure about instead of guessing. The agent reads the actual submission, applies judgment the way a person routing leads would, executes the CRM update and the Slack message as real tool calls, and logs what it did so you can check its work. The condition nobody anticipated does not break anything, because nobody had to anticipate it in the diagram. It becomes an input the agent reasons about instead of a missing branch.

## When does it make more sense to stay on n8n?

Plenty of real situations point back to n8n rather than away from it. Stay if you already have a working library of workflows and the switching cost is real. Stay if your team includes engineers who like precise, inspectable branching logic and do not mind maintaining it. Stay if a specific integration or transformation only works cleanly through an n8n node or its HTTP and code flexibility. Stay if self-hosting is a genuine requirement rather than a preference. These are not edge cases; they are the normal reasons a genuinely good tool keeps its users, and no alternative on this list should pretend otherwise.

## Frequently asked questions

**Is n8n a good product, or does looking at alternatives mean something is wrong with it?**
n8n is a solid, well-built tool with a large node library, a strong community, and a genuine self-hosting option. Looking at alternatives usually means one specific part of the experience, hosting overhead, the learning curve, or wanting runtime judgment instead of fixed branches, does not fit your situation, not that the product is bad.

**What is the closest replacement if I just want an easier visual builder?**
Make and Zapier. Both are fully managed visual builders, so you get a drag-and-drop model similar to n8n's canvas without running your own server. Zapier has the larger app directory and the simplest setup; Make gives you more visual control over branching.

**What should I use if I want to build custom AI agents with code?**
LangGraph if you want an explicit state graph in Python or JavaScript, CrewAI if you want a role-based team of agents collaborating in Python. Both expect an engineering team comfortable maintaining code rather than a visual canvas.

**Can I keep n8n and add an agent platform alongside it, or do I have to switch entirely?**
You can run both. Many teams keep n8n for straightforward, high-volume data movement between apps and move the judgment-heavy, variable-input work, ticket triage, lead qualification, and similar tasks, to an agent platform like Lindy, Relay.app, Gumloop, or Neotask.

**Is there a self-hosted alternative to n8n if I want agents instead of workflows?**
Not with the same maturity today. Most agent platforms, including the ones in this guide, run as managed cloud products. If self-hosting is a hard requirement, staying on n8n and cutting the branch-maintenance burden with better workflow hygiene is currently more realistic than switching platforms.

---

### [Best AI agent platforms in 2026: an honest comparison](https://neotask.ai/blog/best-ai-agent-platforms)

The best AI agent platforms for a business that wants to automate real work without writing code are built around a goal stated in plain language, not a path drawn in advance. Neotask, Lindy, Gumloop, Relay.app, and Microsoft Copilot Studio all work this way: you describe an outcome, and the platform decides what to do next as it runs. Zapier, Make, and n8n take a different approach: you build a fixed sequence of triggers and actions, and it repeats the same path every time. CrewAI and LangGraph sit in a third category, as developer frameworks for engineers who want to write the agent logic themselves in code. None of that shows up on a feature checklist. The real difference comes down to three things: whether the tool decides the next step at runtime or a person decided it weeks earlier, how much setup an outcome actually takes, and what happens the first time reality does not match the plan.

Search for this topic and most of what ranks is a listicle written by one of the vendors being compared, with that vendor's product listed first. That is worth knowing going in. This piece works differently: a short evaluation framework you can apply to any AI agent platform, then an honest look at the real field, Neotask included, held to the same standard as everyone else.

## What actually decides which AI agent platform is right for you?

Before comparing names, it helps to ask the same five questions of every option, because the marketing pages rarely volunteer the answers.

- **Does it decide the next step at runtime, or does it run a fixed path someone built in advance?** This is the split between an agent platform and a workflow pipeline. An agent replans when something changes. A pipeline follows the diagram, whatever the diagram missed.
- **What does setup actually require?** Some tools want you to wire triggers, filters, and branches by hand. Others want a sentence describing the outcome and the apps involved, then they handle the rest.
- **How does it reach your real apps?** Native integrations, a general API or MCP connection layer, or screen-level automation that clicks through a UI like a person would. Each has different reliability and maintenance costs.
- **What happens when something unexpected shows up mid-run?** A bad reply, a missing field, a record that does not match the assumption behind the setup. The honest answer varies a lot: some tools stop and page a human, some replan on the spot using whatever new information just arrived, and some fail with no clear explanation at all.
- **What control do you have over what it is allowed to do?** Look for scoped permissions, approval gates on anything irreversible, and a log of every step it took, not just the final result.

Answer those five honestly for any platform on this list, including ours, and the rest of the comparison gets much easier to read.

![Diagram: a decision flow that asks whether the steps are always the same, then whether you want to build the logic or state the outcome, routing to pipeline tools, developer frameworks, or agent platforms](data:image/svg+xml;base64,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)

## How do workflow builders like Zapier, Make, and n8n actually work?

Zapier, Make, and n8n are workflow builders, sometimes called iPaaS tools, not agent platforms, and that is not a knock against them. Each one connects a trigger in one app to a chain of actions in others, following the exact order you configured. That predictability is the whole point: the same input produces the same output every time, which is exactly what you want for invoice routing, lead capture, or moving a form submission into a spreadsheet.

Zapier's advantage is reach. It connects to an enormous catalog of apps, so if a tool exists, there is a good chance a Zapier trigger or action already exists for it too. Make, formerly Integromat, leans toward a visual, scenario-based canvas that suits people who want to see the whole data flow at a glance, including branches and routers. n8n stands apart because it is open source and can be self-hosted, which appeals to technical teams that want to own their automation infrastructure, inspect every node, and avoid per-task pricing on high-volume workflows.

All three ask you to anticipate the branches in advance. When a form field changes shape or an API returns something the workflow was not built for, the workflow stalls, and someone has to open it up and fix the path. That is the tradeoff for the predictability: dependable on the paths you planned for, and silent or broken on the ones you did not.

## How do agent platforms like Lindy, Relay.app, and Gumloop differ from pipeline builders?

This middle group sits between workflow builders and full developer frameworks. They are built for people who describe an outcome rather than assemble a flowchart, but they still give you a visual surface to configure and watch the work happen.

Lindy positions itself around the idea of an AI employee: you describe a role or a recurring job, such as handling inbound email, scheduling, or following up on leads, and the platform builds an agent around it, generally with approval steps built into how it operates. Relay.app leans hard into human-in-the-loop design, pairing automated steps with review points where a person approves or edits before the agent continues, which suits teams that want AI involved but not fully unattended. Gumloop offers a visual canvas for chaining AI steps together, aimed at ops and technical teams who want to build AI-powered workflows without writing a full application around them.

All three are worth taking seriously for the same reason: they were built agent-first, so the running logic, deciding what to check, what to skip, what counts as done, lives inside the platform rather than in a diagram you drew. The tradeoff is that you are trusting their judgment layer, so the approval and audit features from the framework above matter more here than they do for a fixed pipeline.

## Where does Microsoft Copilot Studio fit for enterprise teams?

Copilot Studio is Microsoft's platform for building agents and copilots that live inside the Microsoft 365 and Power Platform ecosystem. Its main strength is depth of integration with tools an enterprise likely already runs: Teams, SharePoint, Dynamics, and the broader Power Platform, governed through the same Entra ID identity and access controls IT already manages. For an organization standardized on Microsoft, that shared identity and governance layer removes a real barrier that other platforms have to solve from scratch.

The tradeoff is the flip side of the same strength. Copilot Studio makes the most sense inside a Microsoft-centric stack, and less sense for a business running mostly outside it. It fits IT-governed enterprise rollouts more naturally than it fits a small team trying to automate one specific process this week.

## What about CrewAI and LangGraph, the developer frameworks?

CrewAI and LangGraph are not no-code products, and they are worth naming anyway because they define the other end of the spectrum from everything above. Both are open-source frameworks, Python-first, with LangGraph also offering a JavaScript version, for engineers who want to write the orchestration logic themselves: how agents hand off tasks, what memory persists between steps, and exactly which tools each agent can call.

If your actual goal is automating this without writing any code, these are not the answer. But understanding them clarifies the rest of the field: every no-code and low-code agent platform on this list is, underneath, solving the same orchestration problem that CrewAI and LangGraph hand directly to a developer. Teams that need something a UI cannot express yet, or that are building an agent into their own product rather than automating internal ops, are the ones who reach for a framework instead of a platform.

![Diagram: a category map comparing developer frameworks, workflow builders, and agent platforms, showing what each one optimizes for, its examples, and who it fits](data:image/svg+xml;base64,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)

## Where does Neotask fit, and how would you actually set one of these up?

Neotask is an agent platform in the same category as Lindy, Relay.app, and Gumloop: you connect the real apps involved through integrations and MCP, state the outcome you want in plain language, and the agent plans the steps, calls the real tools, and works through the task rather than following a path you drew first. Setup looks like the framework above in practice: connect the apps, describe the job, and the agent takes it from there.

The parts that matter most from the evaluation framework carry through directly. Every step the agent takes is logged, so you can see what it checked and what it did, not just the final result. Permissions are scoped, and anything irreversible can require an approval before it runs, the same guardrail pattern this whole category depends on. None of that makes Neotask the right fit for every job on this list. It makes the most sense for the same kind of work the other agent platforms target: recurring tasks with real variation, where a fixed pipeline would need constant babysitting to keep up.

## Frequently asked questions

**What is the difference between an AI agent platform and a tool like Zapier?**
A workflow tool like Zapier, Make, or n8n runs a fixed sequence of steps you configure in advance. An AI agent platform is given a goal and decides the steps itself while it runs, which is why it can handle the version of the task that nobody anticipated when it was set up.

**Do I need to know how to code to use an AI agent platform?**
No, for the platforms in this comparison. Neotask, Lindy, Relay.app, Gumloop, and Copilot Studio are all built to be set up by describing the outcome and connecting the apps involved, not by writing code. CrewAI and LangGraph are the exception: those are developer frameworks meant for engineers.

**Can I switch from Zapier or Make to an AI agent platform without starting over?**
Usually you keep both for a while. Predictable, high-volume data movement tends to stay on the pipeline tool that already handles it reliably, while work that involves judgment calls or handling exceptions moves to an agent platform. Most teams migrate task by task rather than replacing everything at once.

**What is MCP, and why do some agent platforms mention it?**
MCP, the Model Context Protocol, is a standard way for an AI agent to connect to outside tools and data sources. Platforms that support it can reach a wider range of apps and services without needing a custom-built integration for each one, which is one reason it comes up often in how agent platforms describe connecting to your real apps.

**How do I keep an AI agent platform from doing something destructive?**
Look for scoped permissions, meaning it can read broadly but write only where you have allowed it, approval gates before anything irreversible runs, and a full log of every step taken. That combination is the actual answer, regardless of which platform you pick.

---

### [What is an AI employee, and what work can it actually take over?](https://neotask.ai/blog/ai-employee)

An AI employee is the marketing name for an AI agent that owns one recurring job start to finish, rather than a chatbot you prompt one message at a time. It reads the situation, works inside your actual apps, makes the small judgment calls the job requires, and reports back what it did, the same standard you would hold a person to on that job. It does not do everything a real hire does. It has no accountability in the legal sense, no relationships built over time, and no judgment for a situation nobody has ever seen before. What it can do, honestly, is take over a specific, bounded slice of recurring work: ticket triage and routing, failed-payment recovery, inbox-to-CRM logging, lead qualification, scheduling, reporting rollups, and infrastructure health checks, running today, with a person reviewing the parts that matter.

## What does "AI employee" actually mean?

The label is a positioning choice, not a new technical category. Under the hood, an "AI employee" is the same thing as an AI agent: a system that takes a goal, plans the steps, calls real tools such as your CRM or your ticketing system, and checks its own result. What earns the word "employee" is scope and duration. A one-off assistant answers the question in front of it and forgets the context tomorrow. An AI employee is assigned a job, the same job, repeatedly, and it keeps doing that job across weeks without someone re-explaining it every time. That is closer to how you would describe a role than a tool.

Vendors reach for the word because it is intuitive. Say "AI employee" and a buyer pictures a person doing the job, with a manager, a scope, and expectations. That mental model is mostly useful. It also invites a comparison worth checking carefully before you believe it.

## Where does the label get oversold?

The overselling shows up in three places. First, in the implication that one agent replaces one full-time person, dollar for dollar, when in practice most teams hand over a slice of a role and keep the person for the rest of it. Second, in demos that show the easy path and quietly skip the exceptions, so the pitch looks like a finished hire when the real coverage is closer to "works fine when nothing unusual happens." Third, in language that borrows the weight of employment, hire, onboard, manage, without the parts of employment that make a hire trustworthy on day one: legal accountability, a track record you can actually check, and judgment built over years on the job. An agent starts with none of that. It earns trust the way any automated system earns trust, through a visible record on real work, not through a title.

## What is the honest test for work you can hand over?

Skip the philosophy and ask four concrete questions about the specific job in front of you.

- **Does it recur?** A one-time project is a poor candidate no matter how well an agent could do it once. The setup cost only pays off across repetition.
- **Does it require judgment mid-stream?** If the job is pure data movement, trigger straight to output with nothing to decide along the way, a simple pipeline already handles it more cheaply. The interesting cases are the ones where a person currently has to stop and think partway through.
- **Do the inputs vary?** A job where every case looks a little different is exactly where scripted automation breaks and where an agent's ability to read the specific situation earns its keep.
- **Are the actions reversible or approvable?** If a mistake can be caught before it matters, through an approval step, an undo, or a review window, the downside stays bounded. If a wrong move ships and cannot be pulled back, a refund, an email to a customer, a legal filing, that step needs a person in the loop no matter how well the agent has performed so far.

Four yes answers make the work a strong candidate. One no, and the honest move is to keep that piece with a person, or hand over only the part of the job that clears the bar.

![Diagram: the handover test, four questions, does it recur, does it need judgment mid-stream, do inputs vary, are actions reversible or approvable, leading to good candidate or keep it human](data:image/svg+xml;base64,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)

## What jobs actually work today?

- **Ticket triage and routing.** The agent reads an incoming support ticket, checks account history and prior tickets, assigns priority and the right queue, and drafts a first response for the straightforward cases. Anything ambiguous gets escalated with the history attached instead of a blank slate.
- **Failed-payment recovery.** When a charge fails in Stripe, the agent checks the customer's payment history before doing anything. A long-standing account gets a retry on a sensible schedule and a personal note referencing their actual plan; a repeat failure gets flagged to a person with the full record attached.
- **Inbox-to-CRM logging.** Relevant emails, call notes, and meeting summaries get logged against the right contact and deal automatically, so the sales or success team stops doing manual data entry between real conversations.
- **Lead qualification.** New leads get checked against your actual criteria, company size, stated need, source, and get routed to the right rep with the qualifying detail already attached, instead of a generic handoff with no context.
- **Scheduling.** Meeting requests get matched against calendars, time zones, and stated preferences, with the back-and-forth handled by the agent instead of five email replies to land one meeting.
- **Reporting rollups.** The weekly or monthly numbers get pulled from the systems that actually hold them, assembled into the format your team reads, and delivered on schedule, so nobody spends an afternoon copying figures between spreadsheets.
- **Infrastructure health checks.** A recurring pass over logs, uptime, and error rates, with a plain-language report and an escalation when something crosses a threshold, run on a schedule nobody has to remember to run.

## What doesn't hand over well?

Some work fails the test on purpose, and pretending otherwise is how a rollout loses everyone's trust fast.

Anything that requires **accountability** in the legal or professional sense, signing a filing, approving a budget, representing the company externally, needs a named person whose judgment and liability are actually on the line. **Relationships** work the same way: a customer who has built trust with a specific account manager over two years will not transfer that trust to an agent just because the agent can technically send the same email. **Novel judgment**, a situation nobody has handled before with no precedent to check against, is exactly where an agent's pattern-matching runs out and a person's actual reasoning is the only thing that works. And anything requiring **physical presence**, showing up, a handshake, fixing a physical machine, is out of scope by definition. The honest version of this pitch says all of that out loud instead of hoping the prospect never asks.

## How do you actually onboard one?

Treat it like a new hire with a narrow first assignment, not a rollout across the whole department.

Start with a **narrow scope**: one job, clearly defined, such as "triage and route incoming tickets" rather than a vague mandate to "handle support." Give it **read-broad, write-narrow permissions**: let it see the context it needs across your systems, but limit what it can actually change or send until you have watched it work for real. Put **approval gates on anything irreversible**, refunds, external emails, account changes, so a wrong call gets caught before it ships. Then **widen the scope with the track record**: after real weeks of logged decisions you can review, expand what it can do without asking first, the same way you would extend more autonomy to a person once they have proven the pattern holds.

![Diagram: onboarding ramp from week one narrow scope with approvals on, to week two widening the write scope, to ongoing autonomy on proven work, with trust rising](data:image/svg+xml;base64,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)

In Neotask, this onboarding pattern is a literal permission setting, not a figure of speech. You connect the apps involved, state the outcome in plain language, and set the boundary yourself: read access wherever the agent needs context to do the job well, write access only where you have decided it has earned it, with every step logged so a review takes a few minutes instead of an investigation. Widening the scope later is a permissions change backed by the log, not a fresh round of trust-me.

## Frequently asked questions

**Is an AI employee a real employee?**
No. It has no legal status, no accountability, and no employment relationship. The word describes the shape of the work, one agent owning one recurring job over time, not a legal category.

**Can an AI employee replace a full-time hire?**
For a specific slice of recurring, judgment-light work, often yes. For the full scope of a role, especially the parts involving relationships, novel judgment, or accountability, no. Most real setups keep the person and hand over a defined piece of the job.

**How is an AI employee different from a chatbot?**
A chatbot answers the message in front of it and starts over next time. An AI employee is assigned a standing job, keeps doing it across weeks, and reports on the outcome, closer to a role than a single interaction.

**What happens when it gets something wrong?**
That is what the approval gates and the reversibility test are for. A well-scoped setup catches the mistake before it matters, through a review step or an undo, and the miss becomes part of the record you use to decide whether to widen or narrow its scope.

**How long before you can trust it with more?**
There is no fixed number. It depends on the job's stakes and how often it runs. The real marker is a visible log of decisions you have actually reviewed, not a date on a calendar.

---

### [AI agent examples: 9 real jobs an agent can run end to end](https://neotask.ai/blog/ai-agent-examples)

Real AI agents already run specific jobs end to end: recovering a failed Stripe payment, triaging a Zendesk ticket, logging a Gmail thread into HubSpot, qualifying a new lead, reconciling an invoice in QuickBooks, turning a meeting into tracked action items, checking Datadog every morning, watching for churn risk, and chasing a stalled recruiting pipeline. The ones worth setting up first are the ones with a clear trigger and a narrow judgment call: failed-payment recovery, ticket triage, and inbox-to-CRM logging. The harder ones (churn-risk watch, recruiting-pipeline chase) need more data sources and more nuanced judgment, so they are worth doing once you trust the setup.

What separates each of these from a plain automation is the middle step. A script runs the same steps no matter what it finds. An agent reads the situation first, makes a call about what it means, and only then decides what to do. That judgment call is the whole difference, so every example below is written the same way: the trigger, what it reads before acting, the call it makes, the real actions in real apps, and how it closes the loop.

## What does an AI agent example actually look like, step by step?

Take the clearest case: a payment fails in Stripe. The trigger is obvious, a webhook fires the moment the charge declines. What happens next is where "agent" earns the word.

**Trigger:** Stripe reports a failed charge on a subscription renewal.
**Reads:** the customer's billing history in Stripe (how long they have paid, any past declines), their existing invoices in QuickBooks, and prior email threads about billing.
**Judgment call:** is this a first card decline from a customer who has paid on time for two years, or the third failure this quarter from an account that already looks shaky? Same trigger, different customer, different read.
**Acts:** for the first case, it retries the charge on a sensible schedule, drafts a short personal note in Gmail that references their actual plan and renewal date, and once the retry clears, reconciles the payment in QuickBooks. For the second case, it skips the automatic retry, writes up the pattern, and hands the account to a human with the full history attached instead of guessing.
**Closes the loop:** the invoice status updates, the exchange is logged, and nobody has to go dig through three systems to find out what happened.

![Diagram tracing a failed-payment recovery agent from trigger through judgment call to two different closing actions](data:image/svg+xml;base64,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)

Nobody wrote an if-statement for "customer has paid for two years." The agent worked that out from the record, which is the point of using one here instead of a fixed retry rule.

## Which of these are the best AI agent jobs to set up first?

Three jobs share the same shape as the payment example: one clear trigger, a short read, and a judgment call that only needs a couple of signals. These are the ones to start with.

**Support ticket triage and routing (Zendesk, Jira, Slack).** The trigger is a new ticket landing in Zendesk. The agent reads the ticket text, the customer's plan tier, and their past ticket history. The judgment call is whether this is a billing question, a confirmed bug, or something urgent enough to need a faster response than the queue would normally give it. It tags and routes the ticket, opens a Jira issue for a real bug and links it, posts to Slack for anything flagged urgent, and drafts a first reply for questions it has seen before. It closes the loop with an internal note explaining why it routed the ticket the way it did.

**Inbox-to-CRM logging (Gmail, HubSpot or Salesforce).** The trigger is a new email thread with someone already in the CRM. The agent reads the email against the existing deal record. The judgment call is whether this is a real sales-relevant exchange or noise, like a newsletter reply, and whether it signals a stage change: pricing asked, a demo requested, or the thread gone quiet after weeks of back and forth. It logs the email to the contact's timeline, moves the deal stage when the signal is real, and creates a follow-up task when needed. It closes the loop by never logging the same thread twice.

These three are good first picks for the same reason the payment example is: the judgment call is narrow, the apps involved are usually already connected, and a wrong call is easy to spot and correct.

## Which AI agent jobs take more setup to get right?

The next tier still fits the same pattern, but the read is bigger and the judgment call weighs more signals.

**Lead qualification and enrichment (Google Sheets or HubSpot, plus web lookups).** Trigger: a new lead row appears. Reads: the submitted info, the company's domain, and size signals from a lookup. Judgment call: does this fit your customer profile, and is the email a real business address or a personal one that never converts? Acts: enriches the record with company data, scores the lead, and routes qualified leads into a rep's queue while archiving the rest with a reason attached. Closes the loop by writing the score and reasoning back to the record.

**Invoice reconciliation (QuickBooks, vendor emails, Google Sheets).** Trigger: a vendor invoice arrives or a new transaction posts in QuickBooks. Reads: the invoice amount, the matching purchase order, and the existing ledger. Judgment call: does this match what was expected, or is it a duplicate, a discrepancy, or a new vendor needing approval first? Acts: matches the invoice to its PO and logs it, or flags a discrepancy to accounts payable with both documents side by side. Closes the loop by marking the invoice reconciled or escalated, paper trail attached either way.

**Meeting-to-action-items (a meeting transcript, Notion, Jira, Slack).** Trigger: a meeting ends and a transcript is available. Reads: the transcript and the project's existing docs in Notion. Judgment call: which sentences are actual commitments versus just discussion, who owns each one, and whether a deadline was stated or only implied. Acts: writes structured action items to the Notion page, opens Jira tickets for the engineering items, and posts a summary to Slack tagging each owner. Closes the loop by keeping the action list attached to the meeting record instead of scattered across three tools.

## What is the hardest tier of AI agent job to run well?

The last three need the most data and the most judgment, because the right call depends on context that isn't in any single system.

**Morning infrastructure health check (Datadog, Slack, Jira).** Trigger: a scheduled morning run, or an alert crossing a threshold overnight. Reads: overnight error rates, latency dashboards, and any already-open incidents. Judgment call: is this an emerging problem or the normal spike from a nightly batch job, and does it need someone paged now or just noted in a summary? Acts: posts a morning digest to Slack with anomalies highlighted, opens a Jira ticket for anything crossing threshold with no existing ticket, and pages on-call only for the case it judges genuinely severe. Closes the loop by linking straight to the relevant dashboard and skipping anything already tracked.

**Churn-risk watch (Salesforce or HubSpot, support tickets, product usage data).** Trigger: a scheduled sweep or a usage-drop signal. Reads: the account's usage trend, the tone of recent support tickets, and the renewal date on the contract. Judgment call: is a usage dip a real warning sign or an expected seasonal lull, and which accounts need a human's attention now versus a spot on a watch list? Acts: flags at-risk accounts in the CRM with the reasoning attached, drafts a check-in for the account owner, and adds accounts nearing renewal to a Slack channel built for that. Closes the loop by keeping a running risk score current instead of re-alerting on every small fluctuation.

**Recruiting-pipeline chase (an applicant tracker, Gmail, Slack).** Trigger: a candidate's stage hasn't moved past its expected window. Reads: the candidate's stage, the date of last contact, and any interview feedback already submitted. Judgment call: has the candidate gone quiet because they lost interest, or has the delay been on the hiring team's side? Acts: sends a friendly nudge to the candidate if they went silent, pings the interviewer in Slack if feedback is overdue, and updates the tracker either way. Closes the loop by logging the nudge so the same candidate doesn't get pinged twice.

![Grid placing all nine AI agent examples by setup difficulty and impact, with the easiest and highest-impact jobs marked as the starting corner](data:image/svg+xml;base64,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)

Laid out this way, the pattern is obvious: start where the judgment call is narrow and the payoff is immediate, and treat the wider-context jobs as a second wave once you trust how the first ones behave.

## How do you actually set one of these up yourself?

None of the nine examples above needed a custom integration built from scratch. Each one needed the real apps connected and a goal stated in plain language. That is the actual setup in Neotask: connect Stripe, QuickBooks, Gmail, Zendesk, HubSpot, Salesforce, Notion, Google Sheets, Datadog, or Jira as integrations, describe the outcome you want ("when a payment fails, try to recover it and keep the books straight"), and the agent plans the steps, reads what it needs for the judgment call, and executes through real tool calls in those apps. You set permissions on what it can read versus write, gate anything irreversible like a refund behind approval, and every step gets logged so you can see why it made the call it made. Start with one easy-tier job, watch the log for a week, then widen from there.

## Frequently asked questions

**What's the difference between an AI agent and a regular automation like a Zapier zap?**
A zap runs the same fixed steps every time, so it breaks the moment something unexpected shows up. An agent reads the actual situation before acting, so the same trigger can lead to different actions depending on what it finds, the way the payment example above retries for one customer and flags another.

**Do I need to be a developer to set up something like the failed-payment recovery example?**
No. The setup is connecting the apps involved and describing the outcome you want in plain language. The agent handles the planning and the tool calls itself.

**How do I know an agent won't do something wrong, like refund the wrong customer?**
Scope what it can read versus write, and put an approval gate on anything irreversible, like refunds or external sends. Combined with full step-level logging, a wrong call gets caught and corrected quickly instead of silently repeating.

**Which of the nine examples is the best first project for a small team?**
Failed-payment recovery or ticket triage. Both have one obvious trigger, a short read, and a judgment call with only two or three real outcomes, so they are easy to watch and easy to trust quickly.

**Can one agent handle more than one of these jobs?**
Yes. A single agent can run several of these as separate standing instructions, as long as its permissions are scoped per job. Most teams start with one job, confirm it behaves, then add the next.

---

### [Multi-agent systems: when one AI agent is not enough](https://neotask.ai/blog/multi-agent-systems)

A multi-agent system is a setup where several AI agents, each with a distinct role, tool access, and area of focus, work toward a shared goal instead of one agent doing everything. You need more than one agent when the work genuinely requires different tool permissions, different expertise, or real parallelism on independent pieces. Feeling complicated is not the same test as needing multiple agents. When splitting is warranted, agents coordinate through one of a few patterns: an orchestrator that delegates to specialists, a pipeline that hands work from stage to stage, or parallel workers whose output a synthesizer merges back into one answer.

Most explanations stop at the definition. The more useful question, and the one that actually changes what you build, is when multiplicity earns its complexity and what happens when teams reach for it too soon.

## What is a multi-agent system, exactly?

A multi-agent system means more than one AI agent operating together on a goal, where each agent has its own role, its own tools, and often its own context, with something coordinating how their work fits together. The word "system" is doing real work here. Several copies of the same agent running side by side would just be parallel execution on identical jobs. A multi-agent system means agents with distinct jobs: one agent might have write access to a CRM and nothing else, another might have read-only access to a codebase and a search tool, a third might only see the outputs of the first two and decide what happens next.

That is the honest definition, and it covers a lot of shapes on purpose: a research agent paired with a writing agent, a support agent paired with a billing agent, a planner directing three specialist workers. The shape varies. The reason for splitting stays the same: distinct role, distinct tools, coordinated toward one outcome.

## When do you actually need more than one agent?

Most tasks pitched as "multi-agent" are actually one well-scoped agent with a good tool set. Splitting earns its complexity in three situations worth naming precisely, because "it felt complicated" is not one of them.

The first is genuinely different tool permissions. If one part of a job needs write access to production billing and another part only needs to read support tickets, keeping those as separate agents means a bad output from the ticket-reading agent can never touch billing. That is a real security boundary, not a convenience.

The second is genuinely different expertise. A legal-review step and a code-generation step call for different instructions, different context, and sometimes different models. Cramming both into one agent's prompt makes the instructions long, occasionally contradictory, and harder to improve without breaking the other half.

The third is parallelism on work that is actually independent. If an agent needs to research five unrelated competitors, five agents running at once beat one agent working through the list in sequence, because there is nothing for them to coordinate on until the results come back.

A fourth reason, related to the first two: keeping any one agent's context focused. An agent juggling a support ticket, a calendar invite, and a spreadsheet reconciliation at the same time tends to make worse decisions on all three than three agents each holding one job with full attention.

If none of those four apply, one agent with the right tools and a clear goal will outperform a multi-agent setup, because a multi-agent setup adds coordination cost that a single agent never has to pay.

## How do multiple agents coordinate work without stepping on each other?

![Diagram of three multi-agent coordination patterns: an orchestrator delegating to specialists, a pipeline handing work stage to stage, and parallel workers feeding a synthesizer](data:image/svg+xml;base64,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)

Once splitting is justified, coordination happens through one of three patterns, and most real systems are one of these rather than something novel.

An orchestrator delegating to specialists is the most common shape. One agent holds the goal and the running state, and calls out to specialist agents for pieces of the work, folding their answers back into its own plan. A support agent that calls a billing specialist and a shipping specialist before answering a customer is this pattern.

A pipeline hands work from stage to stage, each agent doing its piece and passing a defined output to the next. A research agent produces a brief, a writing agent turns the brief into a draft, an editing agent checks the draft against the brief. No agent needs to see what the others' internal reasoning looked like, only the handoff contract between stages.

Parallel workers with a synthesizer split independent work across agents that never talk to each other, then merge the outputs in one place. Five agents each summarize one document; a sixth agent combines the five summaries into one answer. The parallelism is the whole point, and the synthesizer is the only place synchronization happens.

The common thread across all three: exactly one place decides how conflicting or overlapping outputs get resolved. When that ownership is unclear, or missing, is when multi-agent systems start failing in specific, predictable ways.

## What goes wrong when teams split agents too early?

The failure modes are consistent enough to name.

Agents duplicate work when two agents both treat a task as theirs and nobody owns the boundary between them. This usually traces back to a role split that looked clean on a whiteboard but overlaps in practice.

Conflicting writes to the same record happen when two agents can both update the same row, file, or ticket without either knowing about the other's change. The fix is almost always a tool-permission fix, giving only one agent write access to that record, rather than a smarter agent.

Coordination overhead can exceed the benefit. Every handoff between agents costs a round trip, a serialization step, and a chance for context to get lost in translation. A system with six agents passing small pieces back and forth can end up slower and less reliable than one agent doing the whole thing directly, especially when the pieces were never actually independent.

Debugging gets harder when a wrong answer came from a chain. If agent three gives a bad answer because agent one handed it a subtly wrong brief, tracing the fault means reading every handoff in the chain, not just the last agent's output. Step-level logging on every agent in the chain stops being optional once there is more than one.

None of this makes multi-agent systems a bad idea. The complexity has a real cost, and that cost needs to buy something specific: a permission boundary, a genuine expertise split, real parallelism, or focus. Without one of those, the extra agents mostly add overhead and more places for something to go wrong.

## How do you decide between one agent and several?

![Decision flow diagram: does the work need different permissions or expertise, or can parts run in parallel, if yes split into multiple agents, if no use one agent](data:image/svg+xml;base64,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)

Ask it in order. Does the work need different tool permissions for different parts? Does it need genuinely different expertise per part? Can meaningful chunks of it run at the same time with no dependency between them? If the answer to all three is no, one agent handles it, and adding more will only add coordination cost with nothing to show for it. If the answer to any one is yes, pick the coordination pattern that matches: orchestrator and specialists for permission splits, a pipeline for expertise splits, parallel workers and a synthesizer for independent parallel work.

Most business problems, in practice, land on "no." A single agent with the right integrations and a clearly stated goal handles invoice reconciliation, ticket triage, and lead qualification without needing a second agent in the loop. Splitting too early is one of the most common mistakes teams make when they first build with agents. It reads like the mature architecture, but it usually just adds latency and failure surface to a job one agent was already handling fine.

## Where Neotask fits

In Neotask, you start with one agent connected to the tools a task needs, because that covers most real work. When a task genuinely calls for a permission split, a specialist step, or real parallel work, you add a second agent for that specific piece instead of rebuilding the whole workflow around a multi-agent design up front. The coordination, who owns which tool, what gets handed off, where results get merged, gets defined the same way you define any agent's job: connect the real apps, state the outcome, set the permissions. You do not have to guess whether a task needs one agent or five before you start. You can run it with one and split only the piece that earns it.

## Frequently asked questions

**Is a multi-agent system just several copies of the same AI agent?**
No. Running the same agent five times is parallel execution, not a multi-agent system. A multi-agent system requires agents with distinct roles, tools, or expertise, coordinated toward one goal.

**Do multi-agent systems need a special framework to build?**
Not necessarily. The three coordination patterns, orchestrator-and-specialists, pipeline, and parallel-and-synthesizer, can be built with plain code, an orchestration framework such as CrewAI, or a platform like Neotask. The pattern matters more than the framework.

**How many agents is too many?**
There is no fixed number. The right test is whether each additional agent buys a real permission boundary, an expertise split, or a genuine parallelism gain. Past that point, more agents just add handoffs and failure points.

**What is the most common mistake teams make with multi-agent systems?**
Splitting before the work justifies it. Most tasks that get architected as five coordinating agents would run just as well, and more reliably, as one agent with a clear goal and the right tool access.

**Who decides what happens when two agents disagree or produce conflicting outputs?**
Exactly one place should: the orchestrator in an orchestrator pattern, the next stage's checks in a pipeline, or the synthesizer in a parallel pattern. If no single place owns that decision, conflicting outputs are a design gap, not a rare edge case.

---

### [What is MCP? The Model Context Protocol, explained for people connecting real apps](https://neotask.ai/blog/what-is-mcp-model-context-protocol)

MCP, the Model Context Protocol, is an open standard that Anthropic introduced in November 2024 so AI applications can connect to external tools and data through one common interface instead of a custom connector for every pairing. In practice, it lets an agent reach your email, your CRM, your database, or your project tracker the same way every time, through a small program called an MCP server that describes what that app can do. You need to care about it if you want an agent to act on your actual accounts rather than a demo: MCP is the wiring that makes a real connection possible. It does not decide what the agent does with that connection, and it does not make the agent smart. That part is still on the model doing the reasoning, and on you.

## What problem does MCP actually solve?

Before MCP, every connection between an AI application and an external tool was a one-off. If you wanted an agent from one vendor to read your Notion pages, somebody built a Notion connector for that vendor specifically: its own login flow, its own error handling, its own way of describing what a "page" even is. If you then wanted an agent from a different vendor to do the same thing, that work started over, because the second vendor's agent framework has no idea the first connector exists. Multiply that by every tool your business uses and every agent platform you might ever try, and the number of custom integrations needed grows roughly by tools times platforms. That is the integration explosion: N tools and M agent platforms turn into something close to N times M pieces of bespoke code, each one somebody has to build, secure, and keep working when an API changes underneath it.

MCP turns that multiplication into addition. A tool maker builds one MCP server that describes what the tool can do, in a standard shape. An agent platform builds one MCP client that knows how to talk to any MCP server. Instead of N times M integrations, you need roughly N servers plus M clients, and any client can use any server it has permission to reach. That is the entire reason the protocol exists: not a smarter agent, a smaller integration bill.

![Diagram: before MCP, three apps each need a separate custom connector to each of three agents, nine one-off integrations; after MCP, the same three apps connect through one MCP server that any MCP-aware agent can use](data:image/svg+xml;base64,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)

## What is an MCP server, in practical terms?

An MCP server is not the app itself. It is a small program, running either on your own machine or hosted somewhere, that sits in front of an app's real API and describes a chosen set of that API's actions in the shape MCP expects. It exposes two kinds of things to whatever agent connects to it: tools, meaning actions the agent can trigger, like "create a calendar event" or "search these tickets," and resources, meaning data the agent can read, like the contents of a document or a list of records. When the agent calls one of those tools, the MCP server is the piece that actually authenticates and makes the underlying API call on your behalf, then hands the result back in a consistent format.

Official and community-built MCP servers already exist for a range of common tools, and the specification itself is open and documented, so if your app does not have one yet, a developer, your own team, or the app's vendor can build one without needing anyone's permission to define a new proprietary format. That openness is the point: MCP is a shared shape for "here is what this tool can do," not a product any single company controls.

## What changes for me once my agent speaks MCP?

For a non-developer, the useful question stops being "does this agent platform have a native integration with my accounting software" and becomes "does my accounting software have an MCP server, and does my agent platform know how to connect to one." Once both sides speak the same protocol, connecting them is closer to a setup step than an engineering project: you point the agent at the server, grant it access to your account, and the tools that server exposes become available to the agent without anyone writing integration code for that specific pairing.

The other real change is portability. Because the standard is shared rather than owned by one vendor, the same MCP server can, in principle, serve more than one agent platform, so a connection you set up once is not automatically thrown away the moment you add a second agent tool or switch vendors.

![Diagram: a single MCP request path, the agent sends a tool call to the MCP server, the MCP server makes an authenticated, permission-scoped request to your app, and the result travels back the same path](data:image/svg+xml;base64,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)

## What are the honest limits of MCP?

MCP is plumbing, not intelligence. It standardizes how an agent reaches a tool. It says nothing about whether the agent decides to use that tool well, whether it picks the right document out of a folder full of similar ones, or whether the message it drafts is one you would actually want sent. That judgment lives in the model doing the reasoning and in the guardrails wrapped around it, not in the protocol.

Auth and permissions are still your job. MCP defines how a client and server exchange requests, but somebody still has to decide which accounts get connected, what scope those credentials carry, and what the agent is allowed to do once it is connected. A protocol that makes it easy to connect an agent to your production database also makes it easy to connect an agent to your production database with far more access than the job needed, if nobody sets narrower permissions on purpose.

A standard does not vouch for the quality or safety of any specific server. Anyone can write an MCP server, which is the entire point of an open specification, but it also means a new MCP server deserves the same scrutiny as any other piece of software you are about to grant real account access to: know who built it, know exactly what it can do, and hand it no more scope than the task requires.

MCP does not tell you which of your processes are safe to hand to an agent in the first place. It explains how an agent reaches your apps once you decide to connect them. Deciding what should run with a human checking every step, and what has earned the right to run unsupervised, is still a judgment call you make case by case, not something the protocol resolves for you.

## How do you actually connect your apps to an agent through MCP?

This is where the idea stops being abstract. In Neotask, connecting an app sets up exactly the kind of link this guide describes: the connection carries its own scoped permissions, and once it exists, the agent can use the tools and data that connection exposes to carry out a goal you describe in plain language. You are not asked to configure a protocol by hand. You connect the app, the agent gets access to what that connection allows, and every tool call the agent actually makes is visible, with approval gates on the actions that matter, which is the same auth-and-permissions discipline covered above, just handled for you instead of something your team has to build and maintain on its own.

## Frequently asked questions

**Is MCP a product I buy, or something already built into an app?**
Neither by itself. MCP is an open specification, similar in spirit to how HTTP or an API format is a shared way of doing something rather than a product. Some apps ship with an MCP server already included, and some agent platforms build MCP client support in directly, but MCP itself is not a purchase. It is the shared language that lets those two pieces talk once each side implements it.

**Do I need engineers to use MCP?**
To build a brand new MCP server for a custom internal system, yes, someone with development skill does that work, once. To use an app that already has an MCP server, through an agent platform that already supports MCP, no: that part looks more like connecting an existing integration than writing a new one.

**Is MCP the same thing as an API?**
No, but the two work together closely. An API is how software already talks to an app. An MCP server typically sits in front of an app's own API and exposes a chosen set of that API's actions in the shape MCP expects, so an AI agent can call them the same consistent way instead of a developer wiring up bespoke calls for every new agent that wants to reach that app.

**Does using MCP make it safe to give an agent access to my accounts?**
No. MCP standardizes how permission and request exchanges happen, but the actual scope of access, and whether a given action needs a human approval step first, is set by whoever configures the connection. Treat an MCP connection the same way you would treat any account access: grant only what the job needs, and gate the actions that would hurt if they went wrong.

**Who created MCP and when?**
Anthropic introduced the Model Context Protocol as an open standard in November 2024, specifically to give AI applications a common way to connect to external tools and data sources instead of relying on separate custom integrations for every pairing.

---

### [How to build an AI agent: the practical path for a working business](https://neotask.ai/blog/how-to-build-an-ai-agent)

You do not need to know how to code to build an AI agent for a business. The setup most companies actually use is connecting the apps the job touches (email, the CRM, the payment processor, whatever holds the data), stating the outcome you want in plain language, and setting permissions for what the agent can read and write. That takes an afternoon, not weeks, because there is no flowchart to draw and maintain. Building an agent from scratch is the other real path, and it needs a developer: a model, tool definitions, an execution loop, memory, error handling, and evaluations, plus the ongoing work of keeping all of that current. Which path fits depends on what you are actually trying to get done, and the honest answer is that most businesses want the first one.

## How do you build an AI agent without being a developer?

Connecting an agent to a business runs through three steps, and none of them touch code. First, you connect the apps involved through integrations or MCP (Model Context Protocol, the open standard that lets an agent call a tool the same way regardless of who built it): your email, your CRM, your support desk, whatever the job needs to read or write. Second, you state the outcome in plain language: "When a support ticket comes in, categorize it, route it to the right queue, and draft a first reply." Third, you set permissions: what the agent can read freely, what it can write on its own, and what needs a human's approval before it goes out. From there, the agent plans its own steps at runtime instead of following a script somebody wrote in advance.

The work here is account connections and one sentence of instruction, not software development. If you can describe what you want a new hire to do on their first day, you can describe it to an agent.

## What does it actually take to build an agent from scratch?

Building your own agent means writing the pieces a configured platform hands you for free. You pick a model and write the system instructions that define its role. You write tool definitions: the functions the agent can call, with a schema describing what each one takes as input and returns as output, so the model knows a "send_email" tool exists and what arguments it needs. You implement the loop that runs at every step: the agent plans what to do next, acts by calling a tool, checks whether the result matches what it expected, and either finishes or replans with the new information.

![Diagram: the plan-act-check loop, with the check step feeding back into plan when the result does not match expectations, and forward to done when it does](data:image/svg+xml;base64,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)

You add memory, because without it the agent forgets the conversation the moment the session ends and repeats work or asks the same question twice. You write error handling for the tool calls that fail, time out, or return something the model did not expect, because a live API will eventually do all three. And you build evals: a test set of real scenarios you run every time you change a prompt or swap a tool, so you find out you broke something before a customer does.

The part people underestimate is what happens after launch. A model update changes how the agent responds to edge cases you had already handled. A tool's API changes its schema and your integration breaks quietly. New situations show up in production that your eval set never covered, and someone has to notice, diagnose, and add a case. None of this is a one-time cost. It is a standing engineering commitment, the same way any production service is.

## Which path is right for your business?

![Diagram: build it yourself, meaning code, tool definitions, the plan-act-check loop, evals, and ongoing upkeep, compared with configure it, meaning connect apps, state the outcome, set limits, and review runs](data:image/svg+xml;base64,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)

Configuring is right when the job is something an app connection and a plain-language goal can cover: reading and writing in tools you already use, judgment calls that a described policy can handle, work that recurs and would otherwise sit on a person's desk. This covers most of what a business actually wants automated.

Building from scratch is right in narrower cases: you need fine control over model behavior for a compliance or product reason, you are embedding an agent inside your own product for customers instead of your own operations, or the job touches a system with no existing integration and no MCP server, so someone has to write the tool definition regardless. If you are already staffed to maintain production software, the ongoing cost of a custom build is a smaller lift for you than it is for a team without that capacity.

## What should your first agent actually do?

Pick a first job with a bounded scope and a clear way to tell if it worked. Four that fit well for most businesses:

- Ticket triage: read incoming support tickets, categorize them, route to the right queue, draft a first response for review.
- Failed-payment recovery: when a charge fails, check the customer's history, retry on a sensible schedule, draft a personal note, and reconcile the result once it clears.
- Inbox-to-CRM logging: read email threads and log the contacts, deals, and follow-ups into the CRM so nothing depends on someone remembering to type it in.
- A morning health check: review the metrics or dashboards that matter each morning and flag anything that looks off before it becomes a bigger problem.

Each one is a real job someone is already doing by hand, has a small blast radius if the agent gets a step wrong, and gives you a clean way to judge whether it is working: did the ticket land in the right queue, did the payment recover, did the CRM entry show up, did the flag catch something real.

## What does a realistic first week look like?

A first week that works tends to follow the same shape, whichever path you took. Early on, you connect the two or three apps the job actually touches and state the outcome in one sentence, and you leave every action gated behind a human review so nothing goes out unchecked. For the next stretch of days, you watch the agent run on real cases, read every action it took before approving it, and correct the instruction when it misreads a situation instead of rewriting the whole thing. Once the review record looks right across a run of cases, you widen the write permissions on the parts that have proven out, and switch from reviewing every action to sampling a portion of them. By the end of the week you should have one agent doing one job reliably, and a decision in front of you: widen that agent's scope, or connect a second job to it.

## How do you set this up in Neotask?

Neotask is built for the configure path described above: you connect the apps involved, state the outcome in plain language, set permissions on what the agent can read and write, and it runs the plan-act-check loop itself, with every step logged so you can review it. The ticket triage, failed-payment recovery, inbox-to-CRM logging, and morning health check examples above are all outcomes you can state to an agent in Neotask today, using the CRM, email, and payment tools you already connect.

If your real need is embedding an agent inside your own product for external customers, or you need direct control over the model and tool layer for a compliance reason specific to your business, that is a different project, and the build-it-yourself path from earlier in this guide is the honest answer for it. Most internal operations work does not need that, which is why most businesses land on connecting apps and stating an outcome instead.

## Frequently asked questions

**Do I need to know how to code to build an AI agent for my business?**
No, if you are configuring an agent on a platform that already handles the model, the tool connections, and the execution loop. You connect your apps, state the outcome, and set permissions. Coding becomes necessary only if you are building the agent's tools and execution loop yourself.

**What is a tool definition, and do I need to write one?**
A tool definition is the schema that tells a model a function exists and what arguments it takes, like a "send_email" tool that needs a recipient, subject, and body. On a configured platform, the tool definitions for your connected apps already exist. You only write your own if you are building an agent from scratch against a system with no existing integration.

**How much does it cost to maintain a custom-built agent versus a configured one?**
A custom build carries an ongoing cost: model updates change behavior, tool APIs change their schemas, and the eval set needs upkeep as new situations show up in production. A configured agent shifts most of that maintenance onto the platform, so the ongoing cost becomes reviewing its actions and adjusting the instruction you gave it, not maintaining code.

**What is the difference between a plan-act-check loop and a simple automation script?**
A script follows steps someone decided in advance and breaks when reality does not match the diagram. A plan-act-check loop decides its next step at runtime: it acts through a real tool call, checks whether the result matches what it expected, and replans if it does not, which is what lets it handle a case nobody wrote down ahead of time.

**Can a business switch from a configured agent to a custom-built one later?**
Yes. Many businesses start by configuring an agent to prove the job is worth automating, then move to a custom build later if they need tighter model control or are shipping the agent inside their own product. The outcome defined and the review record built along the way carry over as the spec for that build.

---

### [Agentic workflows, explained with a real running example](https://neotask.ai/blog/agentic-workflows-explained)

An agentic workflow is a multi-step process where an AI agent decides the steps at runtime: given a goal, it plans, acts through real tools, checks the result, and adjusts until the work is done. That runtime judgment is the entire difference from classic automation. A traditional workflow is a script somebody drew in advance, and it breaks the moment reality stops matching the diagram. An agentic workflow starts from the outcome and figures out the path, which is why the same setup that answers a support ticket cleanly on Monday can handle the weird edge case on Tuesday without anyone rebuilding it.

Most explanations stop at that definition. The more useful question is what one actually looks like while it runs, so this guide walks through the loop, a real example on a live stack, and the honest limits.

## How does an agentic workflow actually work?

Every agentic workflow, whatever the vendor diagram looks like, runs the same four-beat cycle. It receives a goal in plain language. It plans the concrete steps that goal requires right now, against the current state of your systems. It acts through real tools, meaning actual reads and writes in the apps where the work lives. Then it checks its own result and either finishes with evidence or adjusts the plan and goes again.

![Diagram: the agentic workflow loop, goal to plan to act to done, with check and adjust feeding back into the plan](data:image/svg+xml;base64,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)

The loop back from "check" to "plan" is the part worth staring at. A static pipeline has no such arrow. When step B of a script gets a surprise, the script fails, a human gets paged, and the automation earns a reputation for being fragile. When an agent gets a surprise, the surprise becomes input: it replans with the new information and keeps going. Fragility was never about the tools; it was about deciding every step before seeing the data.

## What does an agentic workflow look like in practice?

Here is an agentic workflow running on an ordinary revenue stack, stated the way you would actually state it: "When a payment fails in Stripe, try to recover it, and keep the books straight."

The agent sees the failed charge. It checks the customer's history before doing anything, because a first-time failure from a four-year customer and a third strike from a trial abuser deserve different treatment. For the good customer it retries on a sensible schedule, drafts a personal note referencing their actual plan, and logs the exchange. If the retry clears, it reconciles the payment in QuickBooks and closes the loop. If it does not, it flags the account to a human with the full history attached, so the first human touch starts informed instead of from zero.

Nobody drew that branching in a flowchart. The branches came from the agent reading the situation, which is exactly what you hired it for.

## How is this different from the automation you already have?

![Diagram: a static trigger-step pipeline that breaks on surprises, versus an agentic loop that takes an outcome and handles the unexpected itself](data:image/svg+xml;base64,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)

The comparison that matters is not capability lists, it is what happens on the bad day. Zapier-style pipelines are excellent at moving predictable data between systems, and they earned their place. But every team running fifty of them knows the maintenance tax: fields change, formats drift, and each surprise becomes a support rotation for the person who owns the zaps. The agentic version replaces the pile of brittle paths with one standing instruction and judgment at runtime. Setup stops being flowchart assembly and becomes a sentence.

## Where should you use an agentic workflow first?

The pattern pays off fastest on work with three properties: it recurs, it requires judgment mid-stream, and its inputs vary. Ticket triage, failed-payment recovery, lead qualification, inbox-to-CRM logging, morning infrastructure health checks. Work that is pure data movement with zero variance can stay on simple pipes. Work that a person currently does precisely because "the automation could not handle the exceptions" is the agentic sweet spot, because the exceptions were the point.

## What can go wrong with agentic workflows?

Runtime judgment means the agent can judge wrong, and pretending otherwise is how teams get burned. The mitigations are boring and effective. Scope the tools: an agent that can read everything but write only within defined boundaries has a small blast radius. Gate the irreversible: refunds, deletions, and external sends go through approval until trust is earned. Log everything: a good agentic workflow shows its work, every step, every tool call, so review takes a minute instead of an investigation. Start on low-stakes work and widen as the record proves out. Teams that skip the trust-building step tend to blame the concept for what was really a permissions decision.

## How do you set up an agentic workflow for your own business?

The fastest way to understand agentic workflows is to run one against apps you already use. In Neotask, the setup is the sentence itself: connect the integrations involved, state the outcome in plain language, and the agent handles the plan-act-check loop with the guardrails above. The failed-payment example, the ticket triage, and the morning health check from this guide are all outcomes you can state to the agent today.

## Frequently asked questions

**Is an agentic workflow the same thing as an AI agent?**
No. The agent is the worker; the agentic workflow is the recurring job it runs. One agent can run many workflows, and a complex workflow can hand off between several specialized agents.

**Do agentic workflows replace tools like Zapier?**
For predictable data movement, pipes remain fine. Agentic workflows take over the jobs pipes were never good at: multi-step work with judgment calls and variable inputs. Most real stacks run both during a transition, and the pipes shrink over time.

**How much setup does one actually need?**
Connecting the apps involved and stating the outcome. The plan is generated at runtime, so there is no flowchart to assemble and, more importantly, none to maintain.

**How do you keep an agent from doing something destructive?**
Scoped write permissions, approval gates on irreversible actions, and full step-level logging. Treat it like a capable new hire: broad read, narrow write, widen with the track record.

---

### [The best Salesforce integrations for teams that want the CRM to run itself](https://neotask.ai/blog/best-salesforce-integrations)

Most "best Salesforce integrations" lists are really lists of pipes: tools that copy a record from one system to another when something changes. The copying was never the hard part, though. The hard part is everything around it: deciding which records matter, writing the follow-up, noticing the deal that went quiet. This list is built for that layer. Neotask connects Salesforce to an agent that can query any object with SOQL, create and update records, pull reports, and deploy metadata. So the integrations below are ranked by how much real selling work they remove from your week.

## What makes a Salesforce integration worth having

A Salesforce integration earns its place when it removes a recurring job. The test we applied to every pairing below: after you connect it, is there a task nobody on the team does by hand anymore?

That standard rules out most of the long tail. It also changes what "integration" means here. Because Neotask's agent works from plain language and can act across both systems in one pass, a pairing like Salesforce plus Slack turns into a family of automations: deal alerts, pipeline digests, meeting-note logging, renewal reminders. Each pick below names the specific jobs it covers.

![Diagram: Slack, Gmail, Sheets, HubSpot, and QuickBooks feeding one Salesforce agent, producing a clean CRM, live reports, and faster invoicing](data:image/svg+xml;base64,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)

## Slack: the alert layer your pipeline actually reads

Salesforce dashboards are where deals go to be ignored. Slack is where your team already lives, which is why this pairing beats every other on daily usefulness.

The obvious version posts a message to the right channel when an opportunity changes stage. The version that changes behavior is the Monday-morning pipeline digest: every open opportunity by stage, what moved last week, and anything with a close date in the next 14 days that has no activity logged. Reps stop being surprised by their own pipeline, and sales managers stop building the same report by hand.

Because the agent can run SOQL on any object, digests go far beyond canned fields. "Show me enterprise deals over $50K stuck in negotiation for more than three weeks" is a sentence you say once, and then it becomes a scheduled report in a channel.

## Gmail: from inbox to clean CRM records

The dirty secret of every CRM is that the real customer history lives in email. The Salesforce and Gmail pairing closes that gap in both directions.

Inbound, the agent reads a prospect's reply, logs the exchange against the right contact and opportunity, and updates the stage if the message warrants it. That is the note-taking a rep was supposed to do and did not. Outbound, it drafts the follow-up from what the record shows, including the last touchpoint, the deal size, and the objection noted in the previous call, so a rep reviews and sends instead of writing from scratch.

The compounding effect is data quality. Salesforce records maintained by an agent that reads the actual correspondence stay current. Records maintained by busy reps drift out of date within a quarter.

## Google Sheets: reporting without the export ritual

Every sales team has a spreadsheet that someone rebuilds weekly from a Salesforce export. This pairing retires that ritual. The agent queries live data on a schedule and writes it into the sheet: pipeline by rep, win rates by segment, whatever the sheet existed to answer.

It works in reverse, too. Territory lists, quota tables, and lead imports that start life in a spreadsheet can flow into Salesforce as proper records, with the agent checking for duplicates before it creates anything rather than after you notice them.

## HubSpot: marketing and sales telling the same story

Teams running HubSpot for marketing and Salesforce for sales know the seam well: leads qualify in one system and go stale before anyone sees them in the other. The agent moves qualified leads into Salesforce with their engagement history attached, then sends deal outcomes back so marketing can see which campaigns produce revenue instead of form fills.

A static sync forwards every lead that crosses a score threshold. The agent can apply the rules the way you would state them out loud (right industry, real company, activity in the last two weeks) and route everything else back to nurture.

## QuickBooks: closed-won to cash without the gap

The distance between closed-won in Salesforce and invoiced in QuickBooks is where revenue sits and waits. This pairing closes it. When a deal is marked closed-won, the agent drafts the invoice in QuickBooks with amounts and terms pulled from the opportunity record, and finance reviews instead of re-keying.

It also answers the question sales operations asks at every month end: what closed but has not been billed? The agent checks both systems in one pass and reports the gap, so nobody reconciles two exports by hand.

## Honorable mentions, briefly

Google Calendar logs meeting outcomes to the opportunity and pulls prep notes from the record before the call. BoldSign keeps contract status visible on the deal without anyone chasing legal. Gong attaches call summaries where the deal history lives. Each covers one job well; none reshapes the week the way the five above do.

## How to actually roll these out

Resist connecting everything at once. Teams that get the most from this start with one pairing, one job, one week. The usual best first move is the Slack digest, since it requires nothing from reps and makes the value visible to the whole team immediately. Email logging comes second because it fixes data quality at the source. Reporting and invoicing follow once the underlying records are trustworthy.

One practical note on permissions: the agent acts within the Salesforce access you give it. Start it with the same profile you would give a trusted sales-ops hire, meaning broad read and scoped write, and widen it as the work proves out.

---

### [How to automate Datadog with AI agents](https://neotask.ai/blog/how-to-automate-datadog-with-ai-agents)

Datadog collects more telemetry than any one person can read. Metrics, logs, traces, and CI pipeline events pile up across every service, and the actual bottleneck was never data volume. It was the twenty minutes an engineer spends jumping between dashboards to connect a spike in latency to the log line that explains it. An agent that can query Datadog directly removes that twenty minutes, because it can pull metrics, search logs, and trace requests in one pass instead of one tab at a time.

## What automating Datadog with an agent actually changes

Connect Neotask to Datadog and the agent gains the same access an engineer has: query any metric, search logs by service or error pattern, trace a request across your mesh, and read feature-flag state. The difference is what happens with that access. Instead of a person deciding which dashboard to open next, the agent correlates sources on its own and gives you the answer, not the raw feed.

That reframes what "automation" means here. This is not a rule that fires when a threshold crosses. It is a colleague who already knows where to look.

## Incident investigation without the dashboard tour

When a service degrades, the standard move is to open the service dashboard, notice something odd, open the logs dashboard, filter by time window, then open the trace explorer to see if a downstream dependency is the real cause. Each step depends on remembering the last one correctly.

Ask the agent instead: "Why did checkout latency spike at 2pm?" It queries the relevant metrics, searches logs for errors in that window, and traces requests through the services checkout depends on, then reports what it found and where. The investigation that used to take twenty minutes of dashboard navigation compresses into one exchange, because the agent is not choosing where to look next. It already has all the access at once.

![Diagram: metrics, logs, and traces flowing into one agent pass that returns a correlated answer](data:image/svg+xml;base64,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)

## Daily health checks that replace the morning dashboard ritual

Most ops teams start the day with a lap around the same three or four dashboards, checking whether anything drifted overnight. That lap is a scheduled query dressed up as a habit, which means it can run itself.

Set the agent to check key metrics and search for anomalies every morning, then summarize what it finds in a message. Nothing wrong, it says so in one line. Something drifted, it names the service and the metric before anyone opens a laptop. The team starts the day knowing what needs attention instead of discovering it by clicking around.

## CI and pipeline analytics without the spreadsheet

Flaky tests waste more engineering time than almost anything else in a pipeline, and finding them by hand means scrolling through failed runs looking for a pattern that repeats. The agent can search pipeline events directly and aggregate which tests fail intermittently rather than consistently, which is the actual signal that separates a flaky test from a broken one.

Ask for a weekly reliability report and the agent pulls it from the same telemetry Datadog already has. No export, no spreadsheet someone maintains between sprints.

## Feature flags and security scanning from the same conversation

Datadog's feature flag and security scanning surfaces usually live in a different tab than the incident work, which means they get checked less often than they should. Since the agent already has access to both, a question like "did the last flag rollout correlate with the error spike" does not require switching context. It is the same conversation, querying a different part of the same system.

## Getting started without breaking anything

Start narrow. Give the agent read access to metrics and logs first, and let it run the morning health check for a week before asking it to touch anything else. That single habit, an automatic summary each morning, is usually enough to show the team what the agent notices that a person would have missed anyway.

Once that trust is established, expand into incident investigation and pipeline analytics. Feature flag management is the one area worth holding back until the team has seen the agent's judgment on the lower-stakes work first, since a wrong call there has a bigger blast radius than a missed anomaly in a log search.

---

### [How to automate GitHub with AI agents](https://neotask.ai/blog/how-to-automate-github-with-ai-agents)

You automate GitHub with AI agents by connecting an agent to your repositories through natural language, then letting it handle the recurring work around your code: triaging issues, drafting pull requests, watching Actions runs, and reading through repos to answer questions. Instead of writing a custom script for each task, you describe what you want in plain English and the agent executes it against the GitHub API on your behalf, with approval gates where you want a human in the loop. The result is less time spent on the project management layer that sits around code and more time spent writing it. This is different from a CI bot that only fires on a fixed trigger. An agent can hold a conversation about your repos, chain multiple GitHub actions together, and combine GitHub with other tools like Slack in one workflow. The rest of this post walks through what that actually looks like day to day, and where the limits still are.

## Automating issue triage

Most engineering teams lose real hours every week to the mechanical parts of issue management: reading a bug report, deciding what label it needs, checking if it duplicates something already open, and routing it to the right person or repo. An agent can do this pass continuously instead of in a weekly cleanup sprint. You might ask something like "create an issue in the payments repo for the failed webhook retries bug reported this morning, tag it P1" and the agent files it with a clean title and description pulled from the context you gave it.

The bigger win shows up at scale. Point an agent at your whole GitHub org and ask "list every open issue labeled P0 across our repositories" and it queries across repos in one pass rather than you clicking through each one. Teams that schedule this as a recurring job get a prioritized queue waiting for them each morning instead of starting the day by hunting for what broke overnight. The labeling and routing logic that used to live in someone's head, or in a brittle GitHub Action, now lives in a conversation you can adjust on the fly.

## Watching CI/CD without babysitting Actions

Nobody enjoys refreshing the Actions tab waiting for a build to go green. An agent that's connected to your repo can be asked directly: "what's the status of the latest run on main for the API repo, and did it fail?" It reads the run data and, if something broke, pulls the failure logs and gives you a plain-language read on what likely went wrong, rather than a wall of raw CI output you have to parse yourself.

This matters more once you're shipping multiple times a day across several repos. Instead of a person owning "watch the pipeline" as an implicit job, the agent can check build and deployment status on a schedule and flag failures the moment they happen. Pairing this with a chat tool is where it gets genuinely useful: the agent posts failed-build alerts straight into your engineering channel, so the first person to notice a broken build isn't whoever happened to be looking at GitHub.

## Turning commits into pull requests automatically

Writing a good PR description is one of those tasks everyone agrees matters and nobody wants to do at 6pm on a Friday. An agent that can read your commit history can generate a pull request from a feature branch into main, with a description that actually summarizes what changed rather than "misc fixes." Ask it to open a PR from `feature/payments` to `main` and it can pull the relevant commits and reference the issue number the work closes, so reviewers get context without digging through the branch themselves.

This connects naturally to the issue-triage workflow above: an agent that read the original bug report can carry that context into the PR it later creates for the fix, closing the loop between "someone reported this" and "here's the change that resolves it" without a human re-typing the summary in the middle. Some teams run this as a small pipeline: one agent checks Actions status, a second reviews open PRs for staleness, and a third keeps a changelog updated, each handling one slice of the release process instead of one person tracking all three at once.

## Reading and exploring unfamiliar repos

Not every use of an agent on GitHub is about writing to a repo. A large part of the daily value is read-only: asking an agent to read a README and the main config file from a repo you've never opened before, or to summarize what changed in a set of recent commits. This is especially useful for onboarding. A new engineer, or an agent acting on their behalf, can ask "walk me through what this microservice actually does" and get an answer grounded in the actual code rather than out-of-date documentation.

The same capability works for spot-checking work before it merges. You can ask an agent to review a pull request's diff and flag anything unusual, or to explain why a particular file exists in a repo's structure. None of this requires the agent to write anything back to GitHub, which makes it a low-risk starting point if you're cautious about giving an agent write access on day one.

## What agents can't safely do on GitHub yet

It's worth being honest about the limits here, because overselling automation is how teams end up with an agent merging something nobody reviewed. An agent reading commit history and generating a PR description is not the same as an agent making a judgment call about whether code is actually correct or safe to ship. Automated PR creation should still go through a human review before merging into main, and most teams keep an explicit approval gate there rather than letting an agent merge on its own.

Agents also don't have institutional memory the way a long-tenured engineer does. They can read what's in the repo and the conversation history, but they don't independently know that a certain workaround exists because of a production incident three years ago unless that context is written down somewhere they can read it. Treat an agent's PR descriptions and triage decisions as a strong first draft, not a final call, especially for anything touching production infrastructure or security-sensitive code paths. The approval-gate model, where the agent proposes and a person confirms before anything destructive happens, is the pattern that holds up in practice.

## FAQ

**Do I need to write custom scripts to connect an agent to GitHub?**
No. Once the connection is set up, you interact with GitHub through plain conversation rather than writing integration code for each task. The agent translates your request into the right API calls.

**Can an agent merge a pull request on its own?**
It can, but most teams don't let it. The safer and more common pattern is an approval gate: the agent prepares the PR or flags it as ready, and a person makes the actual merge decision.

**Will an agent replace my CI/CD pipeline?**
No. It doesn't replace your Actions workflows, it sits alongside them, watching runs, surfacing failures, and answering questions about build status so people spend less time manually checking the Actions tab.

**How does this help with a repo I've never worked in before?**
You can ask the agent to read the README, explore the file structure, and summarize what a service does before you touch any code yourself, which shortens the ramp-up time for new repos and new team members.

**Can one agent handle multiple repos across an organization?**
Yes. A single request like listing all P0 issues across every repo in your org runs as one query rather than requiring you to check each repository individually, which is where a lot of the time savings comes from at scale.

---