other
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. Instead of a developer hand-writing queries against a permission schema to debug why a user can or can't see a resource, Neotask can query and manage that schema directly: read the relationships between users, objects, and roles, trace why a specific permission check passed or failed, and help adjust the schema when access rules need to change. This is aimed at engineers working on fine-grained authorization, the kind of system where one misconfigured relation can silently lock someone out or leave something open that shouldn't be. Connected through Neotask, the agent becomes a second set of eyes on the permission graph, one that can walk through the relationships instead of you tracing them by hand in the SpiceDB schema language.
| Permission relationship query | Looks up whether a given subject has a given permission on a resource |
| Schema inspection | Reads the defined object types, relations, and permissions in a SpiceDB or AuthZed schema |
| Relationship management | Creates or updates relationship tuples between subjects and resources |
A developer reports that a specific user can't open a document they should have access to. They ask Neotask to check why, and the agent queries AuthZed for the relationship chain between that user, the team, and the document, then points out the missing relation that's causing the permission check to fail.
Before rolling out a new feature, a developer asks Neotask to walk through the updated SpiceDB schema and explain what access it actually grants. The agent inspects the schema, describes the resulting permission relationships in plain language, and flags a relation that would grant broader access than intended.
Yes, both are built on the same Zanzibar style model and the MCP server queries against either backend.
Both. The agent can query existing relationships to explain current access, and it can also create or update relationship tuples when you ask it to change who has access to what.
Developers and platform engineers who maintain a fine-grained authorization system and need to debug or evolve its permission schema.
It refers to a permission model where access is defined as relationships between subjects, objects, and roles rather than flat access-control lists, which is what SpiceDB and AuthZed implement.
The agent reads and writes exactly what you ask it to, so treat write operations on a production schema the same way you'd treat any direct schema change, with review before applying it.