Workout Chat for Cursor

Log the session without leaving your editor.

Workout Chat is a workout logger with its own chat — and it runs a remote MCP server too. Add it to Cursor and the training you did this morning gets logged in the window you're already working in, while you still remember the numbers.

Start for free →

14 days free. Card not charged until the trial ends.

Why it fits

The log that doesn't happen is the expensive one

Training data is only worth anything in aggregate. One missed session doesn't matter; a habit of missing them turns a progression chart into noise, and by the time you want to know whether the last block worked, the answer isn't in there.

The fix isn't discipline, it's proximity. If you train early and code after, the cheapest place to record the session is the editor you open next — one sentence, in a window that's already focused, before the numbers fade.

Setup

One entry in your MCP config

Remote server, Streamable HTTP transport. Nothing installs, nothing runs locally, and there's no wrapper process to die quietly in the background.

  1. Add the server Drop it into Cursor's MCP configuration. { "mcpServers": { "workout-chat": { "url": "https://workout.chat/mcp" } } }
  2. Authorise First use opens a browser tab. OAuth 2.1 with PKCE and dynamic client registration — nothing to paste into a dotfile, nothing to leak.
  3. Log it Ask the agent to record what you did. It reports back what it wrote. log this morning: squat 5x3 at 275, pull-ups 4x8 bodyweight

Full setup docs →

For people who'd rather script it

MCP is one interface, not the only one

Everything the assistant can do is available over plain HTTP. The MCP server and the REST API share an auth scheme, so a token you already have works against both — write a shell alias, a cron job, or your own client and none of it is second-class.

  • /api — JSON REST over workouts, exercises, weigh-ins, injuries, goals
  • /mcp — Streamable HTTP, JSON-RPC 2.0, OAuth-gated
  • /llms.txt and /llms-full.txt — a curated index for agents
  • /.well-known/mcp/server-card.json — transport, capabilities, auth pointers
  • CSV and JSON export, because it's your data

Read the API reference →

What it looks like

"log this morning: squat 5x3 at 275, pull-ups 4x8 bodyweight"
Dated to this morning rather than to now, with bodyweight handled as a load rather than a missing number.
"what have I trained this week?"
Session list with dates and top sets, straight out of the log.
"am I still on track for the 405 deadlift goal?"
Compares the goal against the real progression and tells you what the current trajectory implies.
"export everything since January as CSV"
Or hit /export.csv yourself — the API isn't a second-class citizen.

Questions

Why would I log workouts from my editor?

Because that's where you already are. If you train before work or at lunch, the log usually happens hours later or not at all — and the session you don't record is the one missing from the chart six months from now.

How do I add it to Cursor?

Add the server to Cursor's MCP config with the URL https://workout.chat/mcp. It's a remote server, so there's no command to run and no local process to keep alive.

Does this work in Cline, Continue, Goose, or Zed?

Any client that speaks the Model Context Protocol over Streamable HTTP and can handle an OAuth flow. The config shape differs per client but the server URL is the same.

Is there a plain REST API too?

Yes — the full JSON API at /api covers every resource, using the same bearer tokens as MCP. If you'd rather write a script, a shell alias, or your own client, nothing is gated behind the AI interface.

How does it authenticate?

OAuth 2.1 with PKCE and dynamic client registration (RFC 7591), plus RFC 9728 and RFC 8414 discovery documents so a well-behaved client can find everything on its own. No key to paste.

Can I export what I've logged?

Yes, CSV and JSON from the export page, or straight from the REST API. It's your training history — it isn't locked in.

Also worth a look

ChatGPT
Log a session in the chat window you already have open.
Claude
A remote MCP connector, authorised with one OAuth click.
Powerlifting
Top singles, estimated maxes, and PRs on the big three.
Bodybuilding
Weekly volume and per-movement progression, without the data entry.
New lifters
No exercise database to learn before you can log anything.
Lifting + cardio
Conditioning in the same session as the squats.
Runners who lift
Lifting tracked properly, runs recorded alongside.
Strong
For when the tapping is the part you want gone.
Hevy
Same log, reached by describing the session instead of building it.
Just using ChatGPT
A conversation isn't a database, and it can't be exported.

See every use case →

Robot deadlifting

Log it before you forget it.

One line in your MCP config and you're done.

Create free account →

14 days free · card not charged until it ends · cancel in one click

Already have an account? Log in