LLMs.txt
Three ways to read these docs as text, for an agent that would rather not parse the HTML.
This design system is consumed by coding agents as much as by people, so every page is available as plain Markdown.
The index
/llms.txt is the llmstxt.org index: every page, nested by
section, with its description.
It is generated from the page tree — the same meta.json files that order the sidebar — so it
cannot drift from the navigation. A page appears there when it exists, not when someone remembers
to list it.
A single page
Append .mdx to any docs URL for that page's Markdown source:
/docs/actions/button → the page
/docs/actions/button.mdx → its MarkdownThat is the source, frontmatter included, not a rendering of it. The MDX components stay as tags:
a <ComponentPreview componentName="button" /> tells a reader that a live example sits there and
which example file it is, which is more useful than the alternatives — dropping it, or inlining
the source of the example twice.
From the page
Every page carries Copy Markdown, which puts the same text on the clipboard, and an Open menu beside it: view as Markdown, or open the page in Claude, ChatGPT, Cursor or Scira.
There is no llms-full.txt. One request per page is the shape the .mdx routes already give, and
a single concatenated file would be stale the moment any page changed.