Skip to content
Platform Codeالعربية

Docs for agents

Every page as Markdown, plus llms.txt, llms-full.txt and llms.json for each language.

Version 0.4.1 ·

The docs are written for people and published for agents. Every page has a Markdown twin, and the whole site is indexed per language, so an agent never has to scrape HTML.

Endpoints

WhatURL
One page/{lang}/docs/<slug>.md
Index of every page/{lang}/llms.txt
Every page in one file/{lang}/llms-full.txt
Every page as JSON/{lang}/llms.json
Agent skills/.well-known/agent-skills/index.json

{lang} is en or ar. For example, the English Button page is https://dga.omqkhafi.dev/en/docs/components/button.md. The address without a language, /llms.txt, is English.

Every Markdown page ends with a link to the llms.txt of its language.

Ask for Markdown

Request a docs page with Accept: text/markdown and you get the Markdown twin at the same URL.

Terminal
bash
curl -H "Accept: text/markdown" https://dga.omqkhafi.dev/en/docs/installation

The response carries Vary: Accept, so caches keep the HTML and the Markdown apart.

Discovery

Pages also announce these files themselves.

  • An HTTP Link header with describedby for llms.txt, sitemap for the sitemap, agent-skills for the skills index, and an alternate for the page's Markdown twin.
  • The same relations as <link> tags in the page head.
  • /webmcp.js, which registers search_docs, get_page and list_pages in browsers that support WebMCP.

Use it

Give an agent the index first, then let it open the pages it needs.

Prompt
text
Read https://dga.omqkhafi.dev/en/llms.txt, then read the Button and Field pages
and build a citizen service request form.

Each page's Copy page menu does the same: copy the Markdown, open it, or hand it to an assistant.