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. That gives you access control, observability, and a record of what happened on top of whatever underlying MCP servers you're actually calling — useful when you want visibility into agent tool usage, not just the results.

• Routes MCP tool calls through Portkey's gateway rather than direct connections
• Applies access control between agents and upstream MCP providers
• Adds observability and logging over tool calls

What you can automate

• Portkey MCP Gateway - Route and manage MCP tool calls through Portkey's gateway, applying access control, observability, and logging between your AI agents and upstream MCP providers.

Real workflows

Get visibility into agent tool usage

With Portkey in the path, ask Neotask what tool calls have gone through recently and the gateway's logging gives you a real record instead of relying on memory of what happened.

Restrict which providers an agent can reach

Use Portkey's access control between Neotask and upstream MCP providers so certain tool calls are limited to what's explicitly permitted.

Frequently asked questions

Does Portkey replace my MCP providers, or sit in front of them?

It sits in front of them — it routes and manages MCP tool calls through its gateway between your agents and the upstream MCP providers.

What do I get beyond just routing?

Access control, observability, and logging are applied on top of the routing between your agents and upstream providers.

Plans

Related integrations

• Fintual - Retirement math is the kind of thing people put off because running the numbers feels like a project.
• Exa - Neotask uses Exa to research the web at depth - company intelligence, people profiles, LinkedIn data, and deep research reports through conversation.
• Algolia - Neotask manages your Algolia search infrastructure - index data, tune relevance, run A/B tests, and monitor performance across 154 available actions.
• 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…
• 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.
• 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.

Explore: Skills · Glossary · Solutions · Use cases · Examples · Comparisons · Templates · Blog · Docs