It tells you which key it ran on
Every tool checks /v1/account/usage and prints the plan and window
before anything else. A table built on a free key looks identical to the real thing and
is not.
Labs
Every one of these runs against the same API you would use, with the same key. We publish the code, including the parts where the answer was no. A tool that only ever confirms what you hoped is worth nothing.
All 63 endpoints, in four languages, with no dependencies in any of them. A free key works with every one. How they work.
Every endpoint as a method, on Composer. cURL where available and the stream wrapper where not, so it runs on shared hosting.
Open on GitHub JS · Python · Go PublishedThe same methods in JavaScript, Python and Go. Generated from the API's own registry, with zero dependencies in any of them.
Open on GitHubCommand line tools you can run yourself, and studies you can reproduce. Python, nothing beyond the standard library, and each reads your key from one config file.
Measure draw bias at any course and distance, and find out that most of it is folklore.
Not published yet Python CLI In buildEvery popular staking system and selection method, run honestly against the archive with confidence intervals attached.
Not published yet Python CLI In buildWhat backing the market leader actually returns, per course, per field size, per going.
Not published yet Python CLI In buildDoes money moving before the off tell you anything? British and Irish racing only, because that is where the data exists.
Not published yet MCP server In buildPoint Claude, ChatGPT or any MCP client at the archive and ask it questions in English.
Not published yet Python CLI In buildOne command line across every one of our APIs, holding several keys at once.
Not published yetEverything above is a command line tool you run yourself. If you would rather not, the same data answers questions in plain English through Claude, ChatGPT or any Model Context Protocol client.
14 curated tools over the same API and the same key. No install, nothing running on your machine, and your plan applies exactly as it does everywhere else. Four of the tools work on a free key.
Set it up“Is there a draw bias at Chester over five
furlongs?”
“How has this trainer done at Ascot on soft
ground?”
“Have these two horses ever met?”
Because a number without these is not worth printing.
Every tool checks /v1/account/usage and prints the plan and window
before anything else. A table built on a free key looks identical to the real thing and
is not.
Every figure comes back with a confidence range and a sample size. If the interval spans 1.00 the tool says so in the same line, rather than leaving you to notice.
Downloaded data lands under the repo's own /data, so a range you have
already pulled is never pulled twice and a re-run costs nothing.
They will work, on a delayed window, and each one says so at the top of its output rather than quietly producing a smaller answer.