Skip to main content
The Model Context Protocol (MCP) is an open standard that lets an AI assistant connect to an external system and use its tools. Archie speaks MCP, so assistants like Claude, Cursor, and ChatGPT can read and act on your data through a secure connection — no glue code to write. Archie exposes two MCP servers, for two different jobs:

Archie MCP

A platform-wide assistant for managing Archie itself — browse your projects and organizations and work across them. You sign in with your Archie account.

MCP Server (per project)

A per-project connector you enable for a single project. Your project’s users and agents query that project’s data — governed by your project roles, data-only by design.

Which one do I want?

The per-project MCP Server is data-only and separate from the Archie MCP. It never exposes platform or admin operations — a connected user can only do what their project role allows, over that project’s data. Turning it on for one project has no effect on any other project.

Key concepts

  • Data-only surface. The MCP Server exposes tools to explore your API, run GraphQL operations, and search your data — not to administer the platform. See Tools.
  • Permissions are inherited. A user who connects the MCP Server acts with their own project role. The MCP does not grant new access; it uses Role-Based Access and any row-level filters exactly as the GraphQL API does.
  • Two ways to authenticate. Depending on the authentication providers your project has enabled, users either sign in with their project identity (a browser login, no token to paste) or authenticate with a project API key.
  • You control the identity and the tools. From the panel you set the server’s name, icon, and description, and curate which tools it exposes.

Get started

1

Enable the MCP Server for your project

Go to Backend → App Services → MCP Server and turn it on. See MCP Server.
2

Connect your assistant

Pick your tool (Claude, Cursor, ChatGPT, or another MCP client) and copy the connection details the panel shows you. See Connecting an assistant.
3

Ask about your data

Your assistant can now explore the API and query your project’s data, within the connecting user’s permissions.