Give the docs to your AI

An AI helps you better when it reads the real documentation. Here are three ways to give it the docs, from the simplest to the most complete.

The first two need no key. For the third, you need a key that can read, with a read scope such as Students › Read (students:read): create it in the app, under Settings › Integrations › API keys. Create it for the AI alone, with only what it needs.

1. Paste the llms.txt link

Paste one of these links into your conversation with Claude or ChatGPT:

• https://fitdistance.io/documentation/api/llms.txt: the short index, with a link and a sentence for each page and each family of routes.
• https://fitdistance.io/documentation/api/llms-full.txt: the whole documentation in a single file, longer.

Both files are in English, like the API contract. They can be read without a key: do not give it to the AI for that.

2. Use the buttons on each page

Every guide page has a Copy page menu, with five choices:

• Copy as Markdown: copies the text of the page, to paste into the conversation if your AI cannot open a link.
• View as Markdown: opens the page as Markdown in a new tab.
• Open in Claude and Open in ChatGPT: open the assistant with a message that asks it to read the page and help you use it. The message holds the address of the page, not its text.
• Copy the llms.txt link: copies the link from step 1.

Also give the AI the link to the page of the family concerned, for example https://fitdistance.io/en/documentation/api/reference/students for students. It finds every route of the family there, with its parameters and its responses.

3. Connect the MCP server with your key

The links give the AI the documentation, not your data. For it to read your students, connect the MCP server with the same key as the API. If your AI tool accepts an MCP server with a custom header, give it:

• the server address: https://mcp.fitdistance.io;
• the header: Authorization: Bearer fd_live_xxxxxxxx.

Replace fd_live_xxxxxxxx with your key. You get the same tools as with an account connection. The scopes of the key hide none of them, except in the tool search (fitdistance_search_tools), which shows the tools whose family scope the key holds. A tool called outside the scopes answers an error that names the missing scope, and costs no credit. Then create a new key that has it.

4. Check that it works

Ask the AI to call the whoami tool. It costs no credit and answers with the connected account. If it fails, check the address, the word Bearer and that the key is not revoked.

5. Know what it costs

A tool that touches your data costs one credit, as on the API: they are the same credits, explained in Keys and credits. Orientation tools cost no credit and answer with any key, such as:

• fitdistance_learn: explains to the AI what it can do, domain by domain.
• fitdistance_mcp_info: gives the version of the server and a diagnostic.
• fitdistance_search_tools: finds a tool by resource and action.

A preview is free too: a tool that asks for confirmation, called without giving it, costs nothing, even at zero credit: only the run is refused. Exception: nine tools whose confirmation depends on the content debit before the preview and refund when nothing is written. So you see what a tool would do before you give your confirmation.

What next

The details of the files and of the server are in llms.txt and llms-full.txt and the MCP server. Your AI can also help you work through a use case: an email for each new student, a reminder for students without a check-in, a reminder for inactive students or another use case from the home page. The use cases that write say so at the top of the page: their key must be able to write.