Connect your own AI to your training log
Reppit runs a first-party MCP server. Point Claude, Cursor or any other MCP client at it and your assistant can read your training and propose changes — which you confirm in the app, not in the chat.
What is actually different
- An open protocol, not one assistant. MCP is a standard, so this is not a single ChatGPT integration. Other lifting apps expose a REST API and community servers wrap it; here the server is the product surface.
- First-party. The server is built and run by us against the same data the app uses, so it does not drift behind the product the way a third-party wrapper does.
- Writes are proposals. A token with write scope can propose a program change. Applying it needs you, in the app. An agent cannot rewrite your training while you are asleep.
What your client can read
- Recent workouts, and any single session in full
- The current program and what the next session is
- Personal records and estimated rep maxes
- Per-muscle recovery and weekly volume
- Strength level against sex-aware standards
- The rule behind any prescribed weight, and the paper behind the rule
- Plate breakdown and the warm-up ramp for a given weight
What it can propose
- Propose a program change
- Propose a goal, habit or profile change
- Propose an update to body stats
Body measurements and progress photos are owner-only. They are never exposed to an MCP client, and never to Coach.
Connecting
Two ways in. OAuth 2.1 with dynamic client registration, for clients that support it — the authorization server metadata is published at the standard well-known path. Or a personal access token from Settings, for a local client like Claude Desktop or Cursor. Tokens are scoped: read-only unless you explicitly grant write.
The connector is part of Pro. Everything it reads is your own data.
Start logging first — the connector is only interesting once there is a training history behind it.
Start free