The whole API in one file

Everything on this site, assembled into one markdown file:

Download the complete reference →

Around 160KB, which is roughly forty thousand tokens — small enough to paste whole into Claude, ChatGPT or anything else with a reasonable context window, and still leave room for it to answer with.

What is in it

• Every endpoint, with its method, path, tier, parameters and a full sample response.
• Every guide on this site, in full — the conventions, the edge cases and the things that are easy to get wrong.
• Base URL, authentication, rate limits, prices and coverage.

Why it cannot be out of date

It is not written, it is generated. The endpoint list comes from the same definition the API routes on, the guides from the same source these pages render from, and the limits from the same configuration the billing uses. Add an endpoint and it appears here on the next request. There is no separate copy to forget to update, which is the usual reason a manual like this is wrong within a month.

The date at the top is the date the underlying definition last changed, not the date you downloaded it — so two downloads with no release between them are byte-for-byte identical, and a diff shows you only what actually moved.

For assistants

There is also llms.txt, a short index that points at the manual and lists every guide and endpoint with a one-line description. It follows the emerging convention for making a site legible to language models: read the index first, fetch the detail if needed.

If you would rather your assistant query the live API than read about it, the MCP server does that directly, and the OpenAPI description is available for tooling that expects it.

Licence and fair use

The document is free to read, copy and feed to whatever you like. The data behind it still needs a key, and the terms that apply to the API apply to anything you build from this. Nothing in it is a prediction, a rating or a tip — this API publishes facts about races that have happened and cards that are scheduled.