Skip to content

AI assistants and coding agents read documentation better as Markdown than as HTML. The portal publishes it for them, with nothing to configure:

URL Content
/docs/<page>.md Each guide as Markdown. The home page is /index.md.
/llms.txt An index (llmstxt.org): the site, then the pages of each documentation tab, then every API with its OpenAPI file.
/llms-full.txt Every guide’s Markdown in one file.

Try it on this site: /llms.txt, or this page as Markdown.

Next to the title of each guide:

  • Copy page copies the page as Markdown, ready to paste into a chat.
  • View as Markdown opens the .md file.
  • Open in ChatGPT and Open in Claude start a conversation that reads the page.
rapidoc.config.yaml
llms:
description: Build on the Acme payments platform. # top of /llms.txt (default site.description)
pageActions: [copy, markdown, chatgpt, claude] # default all four; false hides the buttons

llms: false turns all of it off: no .md files, no llms.txt, no buttons.

Guides are converted from their source, so the Markdown stays close to what you wrote:

  • Frontmatter and MDX import lines are removed; the title and description become the first lines.
  • Callouts (:::note, {% callout %}, <Aside>) become quotes: > **Note:** ....
  • {% api-operation %} becomes the method and path, with a link to the API page.
  • Cards become linked titles, tabs and steps keep their labels, and other tags keep their text.
  • Code blocks and inline code are left exactly as written.
  • /llms.txt and /llms-full.txt list public pages only, and only public APIs.
  • A protected page’s .md is protected like the page: an access rule for /docs/internal/** covers /docs/internal/runbook.md.
  • On protected pages, Open in ChatGPT and Open in Claude are hidden: an assistant can’t sign in to read the page. Copy page and View as Markdown still work for the signed-in reader.