MCP & AI Integrations
Give Claude, ChatGPT, Cursor, or any MCP-compatible client access to exactly the published knowledge you choose.
Primiso exposes your organization's approved knowledge to external AI tools through the Model Context Protocol (MCP) — an open standard supported by Claude, Cursor, and a growing list of other AI clients.
MCP Identities
An MCP Identity is a lightweight service-account concept, created once by an org owner/moderator at the organization level. It has a name, but no login of its own — it exists purely to represent "an AI tool that should be able to read our knowledge."
Granting project access
An identity gains access to a project's knowledge by being granted into that project. From a project's MCP tab, an org admin adds an identity and picks exactly which Resources and Agent Profiles it can see. Access can be edited any time — updating an identity's grants updates every key issued for it instantly, with no need to reissue anything.
Creating a key
Each identity can have one or more API keys — the actual bearer credential your AI client authenticates with. Keys are shown once at creation time; store them like any other secret.
What the connected AI tool can actually see
An MCP client can only ever see Resources and Agent Profiles that are both published and explicitly granted to that identity in that project. Draft content, and anything not granted, is never exposed — regardless of what else exists in the project.
Connecting a client
Point your MCP-compatible AI tool at Primiso's MCP endpoint for the project, authenticating with the API key you generated. From there it can list and fetch the granted resources and Agent Profiles directly — no more re-pasting your architecture docs into every new chat.