Help Center Back to FitKeep
Help Center

Connect an AI Assistant with MCP

FitKeep has a built-in MCP (Model Context Protocol) server, so you can point Claude or another MCP-compatible AI assistant at your own FitKeep data and ask it questions in plain English — “how many kilometers did I run last week?” or “what’s my current 10K PR?” — without leaving your chat app.

What is MCP?

MCP is an open standard that lets AI assistants securely connect to external tools and data sources. Connecting FitKeep over MCP gives your assistant read-only access to your own activities, best efforts, and goals — nothing is ever written back to your account, and no other user’s data is ever reachable.

Requirements

  • An MCP client that supports MCP protocol version 2026-07-28 or later (the current per-request MCP transport). FitKeep’s server only speaks this newer version and does not accept the older 2025-era MCP transport, so an outdated client will fail to connect — update your AI assistant to its latest release if the connection doesn’t work
  • An MCP client that can complete an OAuth sign-in (most current clients, including Claude, do this automatically via Dynamic Client Registration)
  • A FitKeep account to sign in with when your client prompts you

Connect FitKeep to your AI assistant

The exact steps vary a little by client, but the general flow is the same everywhere:

  1. In your AI assistant, find the option to add a custom MCP server (for example, “Connect apps” or “Add MCP server”)
  2. Enter FitKeep’s MCP endpoint: https://fitkeep.cloud/api/mcp
  3. Your client will redirect you to FitKeep to log in (if you aren’t already) and to approve the connection on a consent screen
  4. Once approved, your assistant can start calling FitKeep’s tools on your behalf

There is nothing to turn on in FitKeep itself — every authenticated FitKeep account can connect an MCP client for free.

What your assistant can do

Once connected, your assistant can:

  • List and filter your activities by type and date range
  • Look up the full detail of a specific activity
  • List your best efforts and personal records for standard distances
  • Get a summary of an activity or a rolling summary across a time period
  • List your active goals and their progress

All of this only ever covers your own FitKeep data, scoped to your account by the access token your client obtained during sign-in.

Revoking access

Access is granted through a normal OAuth consent screen, so you control it the same way you would for any connected app: decline or close the consent screen to stop a connection before it’s created, and stop using or remove the server from your AI client to end an existing one.

Troubleshooting

  • Connection fails immediately or the client reports an unsupported protocol version — your MCP client is likely too old. Update it to the latest version, since FitKeep requires the 2026-07-28 MCP protocol
  • You’re asked to log in every time — this is expected the first time you connect; after that, your client should reuse its access token until it expires
  • Your assistant can’t see a recent activity — make sure the activity has finished processing in FitKeep (check your activity list) before asking about it

Related