Agent-ready docs

How your published site serves AI agents and assistants, so they answer from your docs instead of guessing.

More and more readers never open your docs themselves. They ask ChatGPT, Claude, Cursor, or a coding agent, and that agent fetches your site. Every build gives those agents a clean, complete, Markdown version of the docs alongside the pages people read.

What every build includes

Built in	What it does
llms.txt	An index of the site for agents, following the llms.txt convention: the site name, one line per page with its title, description, and a link to its Markdown version.
llms-full.txt	The whole site as one Markdown file, so an agent can read everything in a single fetch.
A Markdown twin of every page	Each page is also published at the same address with a .md extension, as clean Markdown with no layout, navigation, or script.
Content negotiation	A client that requests a page with Accept: text/markdown gets the Markdown twin at the page’s normal address, without knowing the .md convention.
Discovery hints	Each page links to its Markdown version with rel="alternate" in its head, and carries a short note pointing agents at the twin and at llms-full.txt.

The Markdown versions follow the same rules as the pages: private pages are never built, and pages marked noindex are left out of llms.txt and llms-full.txt. When search engines are turned off, robots.txt keeps the Markdown mirrors out of their results too.

For readers who use an assistant

The page action menu at the top of every page is built for the same readers. Copy copies the page as Markdown to paste into any assistant, View as Markdown opens the twin, and Ask ChatGPT and Ask Claude open a conversation about the page in those tools. You choose which of these appear. See Context menu.

For readers who stay on the site, Ask the docs answers from the published pages with citations.

Write for agents

Agents read the same Markdown your readers see, so good docs for people are good docs for agents. A few habits help both:

• Give every page a description in its frontmatter. It’s what llms.txt shows next to the page.
• Use one topic per page and a clear title, so an agent reading the index can pick the right page.
• Prefer plain Markdown and the supported components, which degrade to their text in the Markdown twin.