# AI assistants, API and MCP

## Live today

Every published library is easy for AI assistants to find, read and quote:

- **Markdown versions.** Every published page is also available as Markdown: add `.md` to its address, as in `https://your-team.docsahoy.eu/handbook/billing.md`.
- **llms.txt.** Every library has an `llms.txt` at `https://your-team.docsahoy.eu/handbook/llms.txt`: the library's pages in a form AI tools read.
- **AI crawlers are welcome.** The `robots.txt` of your sites doesn't block the established AI crawlers.
- **Counted separately.** [Analytics](https://docsahoy.docsahoy.eu/docs/analytics) shows AI crawlers and Markdown readers as their own source.

One switch controls all of it: **Open to AI assistants**, under **Visibility** in the library's settings. When it is off, `robots.txt` asks AI crawlers to stay away and the Markdown versions and `llms.txt` are switched off. It is separate from the switch for search engines.

## Coming

These are not available yet. We describe what they will do, not when.

- **A public MCP server per library.** MCP (Model Context Protocol) lets an AI assistant search and read a library directly, with the library's own search. It will follow the **Open to AI assistants** switch, and be included in every plan.
- **A developer API.** Read and write your libraries from your own code, with API keys you create for your team. Changes made through the API will go through [change requests](https://docsahoy.docsahoy.eu/docs/change-requests), like any other edit.
- **An MCP server for writing.** Let an AI assistant you trust propose changes to your documentation, as change requests that a writer reviews and merges.

This page will say so when they are live.