# For AI agents

How an assistant, a crawler or a script reads this documentation as plain Markdown, with llms.txt, one Markdown file per page, and content negotiation.

Everything on this site is also served as Markdown, so an assistant can read a page without parsing the HTML around it. Nothing needs a key, a login or a browser.

## The index

`https://docs.copilka.app/llms.txt` lists every page in reading order, each with its one-line description and the address of its Markdown. It follows the [llms.txt](https://llmstxt.org) convention: a heading, a summary in a blockquote, a few paragraphs on how to read the site, then the pages by section. Start there.

## Everything in one file

`https://docs.copilka.app/llms-full.txt` is the whole documentation, every page one after another, each beginning with its title, its section and its address. It is the file to give a model that should know all of Copilka at once.

## One page

Every page has a Markdown twin at the same address with `.md` in place of the trailing slash:

| Page | Markdown |
| --- | --- |
| `https://docs.copilka.app/start/sign-in/` | `https://docs.copilka.app/start/sign-in.md` |
| `https://docs.copilka.app/saving/` | `https://docs.copilka.app/saving.md` |
| `https://docs.copilka.app/` | `https://docs.copilka.app/index.md` |

Or ask the page itself for Markdown. A request that prefers `text/markdown` over `text/html` gets the Markdown at the page's own address, with a `Content-Location` header saying where it came from:

```bash
curl -H 'Accept: text/markdown' https://docs.copilka.app/start/sign-in/
```

A request that accepts anything, or names HTML, gets the page. Every page carries a `Link` header and a `<link rel="alternate" type="text/markdown">` in its head pointing at its Markdown, and under each title there is a **Markdown** link and a **Copy for an assistant** button, which puts the page's Markdown on the clipboard for pasting into a chat.

The Markdown is the page's source: the same text, with links and images made absolute. Pictures are referenced by address, not embedded.

## Crawling

`robots.txt` allows everything, and `https://docs.copilka.app/sitemap-index.xml` lists every page. There is no rate limit. The site is small, some fifty pages, so reading all of it is a few hundred kilobytes.

## Your library is somewhere else

These addresses are for reading *about* Copilka. An assistant that should read or change what you have *saved* connects to Copilka's own MCP server at `https://mcp.copilka.app/mcp`, with your permission and your account. See [Connect Claude](https://docs.copilka.app/assistants/connect-claude/) and the [Tools reference](https://docs.copilka.app/assistants/tools/).

---

This page: https://docs.copilka.app/reference/for-agents/ · As Markdown: https://docs.copilka.app/reference/for-agents.md · Every page: https://docs.copilka.app/llms.txt
