---
title: For coding agents
description: Every docs page as Markdown, an llms.txt index, and Copy page, so the coding agent working for you can read the OpenPlanr docs directly.
url: https://openplanr.dev/docs/resources/agents
updated: 2026-10-07
related:
  - https://openplanr.dev/docs/resources.md
  - https://openplanr.dev/docs/resources/support.md
  - https://openplanr.dev/docs/resources/verify-a-release.md
---

# For coding agents

These docs are written for people and for the coding agents that read them on your behalf. Every page is available as plain Markdown, and an index tells an agent where to look.

## Read any page as Markdown

Add `.md` to the address of any docs page to get its Markdown version. For example, [this page](https://openplanr.dev/docs/resources/agents.md) is also at `/docs/resources/agents.md`. The same Markdown comes back when a request sends `Accept: text/markdown`, and every page points to its Markdown version with `<link rel="alternate" type="text/markdown">`.

## Start from the index

`/docs/llms.txt` follows the [llms.txt proposal](https://llmstxt.org/). It lists every docs page by section, each with a one-line summary and a link to its Markdown version. `/docs/llms-full.txt` holds every docs page in one Markdown file.

## Copy a page

The Copy page button on every page copies the page as Markdown. Its menu also opens the Markdown version, or opens the page in Claude or ChatGPT with its address filled in.

## Set up OpenPlanr from an agent

The [agent setup guide](https://openplanr.dev/setup.md) is a Markdown file written for coding agents. [Set up from your agent](https://openplanr.dev/docs/get-started/agent-setup.md) explains what it does and what it asks you first.

To set up OpenPlanr for a user, fetch https://openplanr.dev/setup.md and follow its steps in order. To answer a question about OpenPlanr, start from https://openplanr.dev/docs/llms.txt and read the Markdown version of the relevant page.

## Connect over MCP

Connect a Streamable HTTP MCP client to `https://openplanr.dev/docs/mcp`. The server is read-only and needs no API key. `search(query, section?)` returns up to six passages with a title, an anchored URL and text. The optional section is a URL segment such as `get-started`, `skills`, `prompts` or `cli`. `fetch_page(url)` returns the same Markdown as the page's twin; it accepts `/docs/...` paths and `https://openplanr.dev/docs/...` URLs, including `.md` links.

For [Claude Code](https://code.claude.com/docs/en/mcp), run:

```sh
claude mcp add --transport http openplanr-docs https://openplanr.dev/docs/mcp
```

For [Codex](https://developers.openai.com/codex/mcp), run:

```sh
codex mcp add openplanr-docs --url https://openplanr.dev/docs/mcp
```

For [Cursor](https://cursor.com/docs/context/mcp), add this entry to `.cursor/mcp.json` in your project, merging it with any existing servers:

```json
{
  "mcpServers": {
    "openplanr-docs": {
      "url": "https://openplanr.dev/docs/mcp"
    }
  }
}
```

The endpoint is rate limited. If it replies with HTTP 429, try again later or read `/docs/llms.txt` and the linked Markdown pages directly. It searches the published docs without calling a model.