Connecting Your Own AI Assistant (MCP)

The same tools that power the AI Console are also exposed through an MCP server. MCP (Model Context Protocol) is an open standard that lets AI assistants connect to external systems — so you can plug Full Fabric into Claude, or any other MCP-capable assistant, and work with your institution's data from the chat you already use, without opening Full Fabric.

What you get

The MCP server exposes the full Full Fabric tool set — the same areas listed in the overview, from profile lookups and statistics to creating campaigns and events. There is no separate or reduced catalogue: anything the AI Console can do, your own assistant can do through MCP.

The same safeguards apply:

Every tool call is checked against your own permissions. Your assistant can never see or change data you could not access yourself in Full Fabric.
Actions that change data follow the preview-and-confirm convention: the assistant first fetches a preview of the change and only carries it out after you approve it in your chat.

Setting it up

You need two things, both provided by an administrator:

MCP access for your role.
Your access token, which authenticates each request.

Then point your assistant at the endpoint:

https://<your institution domain>/ai/mcp

The token is sent with each request in the FF-Auth header. For clients configured with a JSON block (Claude Code, Claude Desktop, and most MCP-capable agents follow this shape):

{
  "mcpServers": {
    "fullfabric": {
      "type": "http",
      "url": "https://institution.fullfabric.com/ai/mcp",
      "headers": { "FF-Auth": "<your access token>" }
    }
  }
}

Once connected, the assistant discovers the available tools automatically — ask it something like "how many applications were submitted this month?" to confirm everything works.

Warning: Your token grants access to live institution data under your identity. Treat it like a password: store it in your client's secure configuration and never share it.

Using the AI Console

Overview