Rush SR Lap Analyzer

Rush SR Lap Analyzer

A remote MCP server that analyzes AiM RaceStudio lap data. It works as a connector in Claude and as a plugin in ChatGPT. Built by Rush Auto Works.

https://laps.mcp.rush.sr/mcp        Streamable HTTP, no sign-in

Tools

Tool	What it does
analyze_session	Lap times, best lap, theoretical best, consistency, and which laps were excluded and why.
compare_laps	Time delta, braking points and corner minimum speeds for two laps in one session.
find_time_loss	The sectors where a lap loses the most time against your best lap.
compare_sessions	Two sessions on their best valid laps: delta, sectors gained and lost, braking points, corner minimum speeds, top speed. Use it for two drivers or two cars on the same track.

Every tool is read-only. Every result ends with one line, Built by Rush Auto Works, with a link to rushautoworks.com.

Giving it data

Send an AiM RaceStudio CSV export as csv_text, or attach the file where the client supports uploads. compare_sessions takes csv_text_a and csv_text_b (or file_a and file_b) and optional lap_a and lap_b. Files are limited to 8 MB, so a very long session may be rejected with a message naming the limit.

Excluded laps: the first lap (out-lap), the last segment (in-lap, because it ends with the data, not on a beacon crossing), and any lap more than 7% slower than the best.

AiM .xrk logs are binary and are not read directly. In Claude, install the Rush SR plugin, which converts an .xrk with libxrk and runs the same engine in the sandbox.

Try it

Two synthetic sessions, not from a real driver, are in examples/:

• synthetic-session-a.csv: best lap 4 in 45.793 s, theoretical best 45.547 s, 4 valid laps.
• synthetic-session-b.csv: the same track about 3% slower. compare_sessions on the pair gives a delta of -1.416 s.

What it does with your data

It analyzes the data in memory and stores nothing. It makes no calls to a language model and needs no sign-in. A file link is downloaded once, over HTTPS, with limits on size, redirects and time. Requests pass through Amazon CloudFront and Cloudflare as infrastructure providers. Workers Logs are switched off in wrangler.jsonc, and a test fails if that changes.

Privacy policy: https://rushautoworks.com/privacy-policy/

Support

[email protected], or an issue at https://github.com/Rush-Auto-Works/llm-plugins/issues.

Deploy

npx wrangler deploy

ChatGPT's plugin submission checks a token at /.well-known/openai-apps-challenge on the MCP host. The dashboard issues the token when a submission draft is started. Add it to vars in wrangler.jsonc and redeploy:

"vars": { "OPENAI_APPS_CHALLENGE": "<token from the ChatGPT plugins dashboard>" }

The route returns the token as plain text. Until the variable is set it returns 404, so nothing is served by default. Check it with curl https://laps.mcp.rush.sr/.well-known/openai-apps-challenge.

Run it locally

npm ci
npm test         # boots wrangler dev and runs the E2E suite
npm run dev      # http://127.0.0.1:8787/mcp

MIT licensed.