> ## Documentation Index
> Fetch the complete documentation index at: https://docs.get-exo.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Read these docs from your agent

> Add this documentation site as an MCP server, or read any page as plain Markdown, so your coding agent can look things up instead of guessing.

You are probably reading this inside a coding agent, or about to paste it into one. This site is built to be read that way.

<Note>
  This page is about the **documentation** MCP server, which serves these pages. It is not the Exo API MCP server, which serves your knowledge graph. Different URL, different purpose. For that one, see [Integrations](/integrations).
</Note>

## Add the docs as an MCP server

```bash theme={"theme":{"light":"github-light-default","dark":"github-dark-default"}}
claude mcp add --transport http exo-docs https://docs.get-exo.com/mcp
```

Your agent can then search these docs and navigate them directly, rather than working from whatever it remembers about an API it has never seen. There is no key and no account: the docs are public.

## Read any page as Markdown

Append `.md` to any URL on this site and you get the source, with no navigation, no styling and no HTML to strip:

```bash theme={"theme":{"light":"github-light-default","dark":"github-dark-default"}}
curl https://docs.get-exo.com/guides/retrieving.md
```

That is the cheapest way to hand an agent one specific page. It is also the right way to pin a single page into a prompt, because the Markdown is stable in a way scraped HTML is not.

## llms.txt

The site is hand-authored for a model rather than auto-generated: at 80 operations, Mintlify's automatic `llms-full.txt` would silently truncate past its 100,000-character cap, so Exo ships its own index plus one file per tab instead of a single dump.

| File                                            | Contents                                                                                                          |
| ----------------------------------------------- | ----------------------------------------------------------------------------------------------------------------- |
| `https://docs.get-exo.com/llms.txt`             | An index: every page on the site with its description.                                                            |
| `https://docs.get-exo.com/llms-guides.md`       | Full text of the Guides tab: quickstart, authentication, task guides, platform reference.                         |
| `https://docs.get-exo.com/llms-concepts.md`     | Full text of the Concepts tab: the knowledge graph and the seven named response objects.                          |
| `https://docs.get-exo.com/llms-api.md`          | Full text of the API reference tab: all 80 operations with their request and response fields.                     |
| `https://docs.get-exo.com/llms-integrations.md` | Full text of the Integrations tab: Claude Code, the Claude app, Codex, ChatGPT, imports and the GitHub connector. |

Start from `llms.txt` to let a model decide what it needs, then either fetch individual pages as `.md` or pull the one shard that covers what you are doing. None of these five files can silently truncate: they are committed, not generated on request.

## Copy this page into an assistant

Every page carries a menu in its header with **Copy page**, **View as Markdown**, and options to open the page directly in ChatGPT or Claude. When you are stuck on one endpoint, that is faster than any of the above.

## Why this exists

Exo's whole argument is that an agent works better when it can read the thing it needs instead of reconstructing it from memory. Documentation an agent can ingest cleanly is the smallest honest version of that claim, so we ship it.

## Next

<Columns cols={2}>
  <Card title="Integrations" icon="plug" href="/integrations">
    Connect your agent to your own knowledge graph, not just to these docs.
  </Card>

  <Card title="MCP tool reference" icon="wrench" href="/integrations/mcp-tools">
    What the Exo tools do once your agent is connected.
  </Card>
</Columns>