Skip to main content
The MCP server is in public beta. It is open to use and still being refined, so tools and behavior may change.
The GC AI MCP server lives at one endpoint:
You do not need an API key. The server authenticates with your GC AI login through a standard OAuth flow: when you add it, your host opens a GC AI sign-in page, you approve access, and the host stores the connection. Every request then runs as you.

Before you connect

Two things need to be true for the connection to work:
  1. You have a GC AI account. The assistant acts as you, so you sign in with your own credentials.
  2. Your organization has the MCP server turned on. It is off by default. An org admin turns it on under Settings, API.
If your organization has not turned MCP on yet, you can still add the server and sign in. The tools appear in your host, and calling any of them returns a message telling you to ask your admin. Once the admin turns it on, the same tools start working with no need to reconnect.

Add the server to your host

The steps differ slightly by host, but the shape is the same everywhere: point the host at the endpoint above, then complete the GC AI sign-in when prompted.
In a ChatGPT plan that supports connectors, add a new connector using the URL https://app.gc.ai/api/mcp. ChatGPT walks you through the GC AI sign-in, then lists the GC AI tools. Enable the ones you want the assistant to use.
The exact menu labels vary by host and change over time. If you do not see a place to add a remote MCP server or connector, check that your plan supports custom connectors.

Verify the connection

Once connected, ask the assistant to run a quick read, for example: “List my GC AI projects.” That calls get_projects and confirms the tools are wired up and scoped to your account. From there the assistant can upload files, run playbooks, and ask questions grounded in your knowledge base.