Skip to main content
Governance covers three decisions: who may build and edit an Agent, what the Agent may use, and what a run can reach. A workspace role decides who builds Agents: Admin and Developer create, edit, and delete them, and Researcher has view access. Roles lists the permission for every role.

​
What an Agent may use

Each Agent carries its own capability scope, chosen when it is built. Add tools, MCP servers, and skills, and connect the knowledge bases it may read. A run reaches the tools and servers on the Agent.

​
What a run can reach

  • MCP gateways: a run reaches the gateways attached to its Agent, and each gateway is shared with projects. A key minted for an Agent session carries only that Agent’s gateways. See Project access.
  • API keys calling the Agent: the key’s permission domains, its project scope, its mcp_access allowlist, and any Budget scoped to it. See API keys.

​
What an Agent runs as

A run acts as whoever invokes it: a member’s session carries that member’s role, and an API key carries the key’s grants. Two Agents built by the same person run with that person’s permissions.