---
title: "Docs for agents"
description: "How a coding agent reads these docs without parsing HTML. Every page as markdown, the llms.txt index, llms-full.txt and the sitemap."
url: "https://docs.alisx.com/reference/docs-for-agents/"
---
# Documentation for agents

Every page on this site is also served as plain markdown, so a coding agent can read it without parsing HTML. The markdown is the page's source, so it uses the syntax the site is written in: headings, tables, fenced code blocks, and MkDocs Material's tabs (`=== "macOS"`), admonitions (`!!! note`) and key names (`++ctrl+c++`).

## Index files

| URL | Contents |
| --- | --- |
| `https://docs.alisx.com/llms.txt` | Every page as a markdown link with a one-line summary, grouped as the site's navigation is. Start here. Follows the [llms.txt convention](https://llmstxt.org). |
| `https://docs.alisx.com/llms-full.txt` | Every page's markdown in one file, in the order of the index. |
| `https://docs.alisx.com/sitemap.xml` | Every page's HTML URL, for crawlers. `robots.txt` allows everything and names it. |

## One page as markdown

There are two ways to ask for a page as markdown.

**Add `.md` to the page URL.** `https://docs.alisx.com/tutorials/install-the-cli/` becomes `https://docs.alisx.com/tutorials/install-the-cli.md`. A section index such as `https://docs.alisx.com/how-to/cli/` becomes `https://docs.alisx.com/how-to/cli.md` or `https://docs.alisx.com/how-to/cli/index.md`; both work.

**Ask for markdown.** Request the page URL with the header `Accept: text/markdown`. The response is the markdown, with `Content-Type: text/markdown` and a `Content-Location` header naming the `.md` URL. A browser, which asks for `text/html` first, keeps getting the page.

```bash title="Terminal"
curl -s https://docs.alisx.com/reference/cli/build.md
curl -s -H 'Accept: text/markdown' https://docs.alisx.com/reference/cli/build/
```

Each markdown page starts with a front matter block: `title`, `description` where the page declares one, and `url`, the page's HTML address. Links inside the page are absolute and point at the markdown version of the linked page, so an agent can follow them without keeping track of a base URL. Every HTML page also names its markdown twin with `<link rel="alternate" type="text/markdown">`.

## The CLI's own manual

These pages describe the platform for people and agents alike. The `alis` command line carries a separate operating manual written for agents, with the output contract, the exit codes and the recovery steps for each error: run `alis docs` (see [alis docs](https://docs.alisx.com/reference/cli/docs.md)). An agent working in a terminal reads that first and comes here for the broader picture.