<!-- headlesslaw {"release":{"id":"a4f2a0495fe55b47323883f82977a9b34ab66171ec9623e39925d3a7045b6c81","published":"2026-11-01T16:41:30Z","pinned":false},"limits":{"coverage":{"history":"current_only","held_since":null,"gaps":[],"covered":[],"not_covered":["national guidance","court decisions"]},"freshness":{"release_id":"a4f2a0495fe55b47323883f82977a9b34ab66171ec9623e39925d3a7045b6c81","release_published":"2026-11-01T16:41:30Z","checked_at":null,"checked_against":null},"truncation":null,"failures":[],"incomplete":[]},"page":{"path":"/mcp","url":"https://www.headlesslaw.com/mcp","title":"MCP server","family":"home","work_id":null,"status":"in_force","force_status":null,"date_state":"current","requested_date":null,"as_of":"2026-11-01","applies_from":null,"application":null,"version_label":null,"check":null,"check_result":null,"official_url":null,"valid_from":null,"valid_until":null,"revision":null,"content_id":null,"change_date":null,"nearest_versions":[],"first_text_date":null,"correction_applies":false,"versions":[],"version_url":null,"revision_url":null,"versions_url":null,"relations_url":null,"source_line":null,"attribution":null}} -->
---
title: "MCP server"
description: "The site's Model Context Protocol endpoint: its address, how to add it, what it serves."
type: section_index
---

The site has a Model Context Protocol (MCP) server. An AI assistant that supports MCP can search the site and read its pages through it.

## Address

`https://mcp.headlesslaw.com/mcp`

Transport: Streamable HTTP. No sign-in and no key.

The earlier address, `https://www.headlesslaw.com/~gitbook/mcp`, still works and serves the same tools.

## How to add it

Claude Code:

```
claude mcp add headlesslaw --transport http https://mcp.headlesslaw.com/mcp
```

Other MCP clients: add a remote (HTTP) server with the address above, for example:

```json
{
  "mcpServers": {
    "headlesslaw": { "type": "http", "url": "https://mcp.headlesslaw.com/mcp" }
  }
}
```

## What it serves

The server answers with the pages of this site: the text of each act, its Level 2 and Level 3 material, and the definitions. Its tools (tool contract version 1.1.0):

| Tool | Input | Returns |
| --- | --- | --- |
| `searchDocumentation` | `query` | Matching passages from the site's pages, each with its page title and page address. |
| `getPage` | `url` | The full Markdown of one page, given its address (for example `https://www.headlesslaw.com/mica/articles/3`). |
| `sendFeedback` | `content`, `pageUrl`, optional `goal` | Sends a report about a problem on a page to the site's maintainers. Reports are not published. |

Tools are only added, never changed in place. A tool is removed only after a notice on this page and in the tool's description.

## What the server logs

The text of a request is not kept: what an assistant searches for, the page address it asks for and the content of a report are read to answer the request and then dropped. For each request the server logs the time, the tool, the outcome, the size and duration of the answer, the name and version the MCP client gives, and a few features worked out from the request before it is dropped: the length of the query in bands, whether its letters are Latin, recognised citations as references to this site's acts (for example [Article 75 of MiCA](https://www.headlesslaw.com/mica/articles/75)), the number of results and the acts they came from. It logs no IP address, cookie, session or client identifier and no full User-Agent. Each request's log entry is deleted after 90 days; daily totals are kept.

The same pages are available without MCP: add `.md` to any page address for its Markdown, or read [llms.txt](https://www.headlesslaw.com/llms.txt) for an index of the site.