Documentation formats

The user guide is written for people, but it doesn’t have to be read by them. The same content is available in a range of formats, so you can read it in your browser, copy it into your favourite notes app, or have an AI agent use it to answer your questions. This page describes each of those formats, and how to use them.

Agent-friendly documentation

AI agents work best with content they can request and read programmatically, without having to guess how a web page is laid out. We provide several ways for them to access the guide, each suited to different situations:

• A Markdown version of every page, for agents that can fetch a specific page.
• llms.txt, an index of the whole guide, for agents that need an overview first.
• An agent skill, for agents that support the Agent Skills format and need instructions on how to help with the guide.
• A public Model Context Protocol server, for agents that can connect to MCP servers.
• WebMCP tools, for agents that browse the site in a browser.
• A read-only API, for scripts and applications that need the guide’s contents as JSON.

All of these are open and don’t need authentication. They’re covered by the site’s Content Signals policy, and they’re registered in the site’s Agentic Resource Discovery entries, so agents can find them.

Markdown

Every page of the guide has a Markdown version. Use it when you want the page’s contents as plain text, without the site’s navigation, layout, and styling. There are two ways to get it:

• Page URL: add markdown/ to the end of any page URL. For example, guide.wagtail.org/en/about/markdown/.
• Content negotiation: request the page URL with an Accept: text/markdown header. Browsers still get the HTML page, so the same URL works for both.

Each Markdown page starts with the page’s title, description, and date last modified, followed by its full URL. The Markdown uses the same headings and links as the page, so tables of contents, section links, and cross-references all work as you’d expect. Pages that only exist in English are also available from the other languages.

llms.txt

llms.txt follows the llms.txt specification, a predictable way to give AI tools an overview of a website. It’s the best place for an agent to start when it doesn’t know what the guide contains.

It includes the guide’s title and description, guidance on when to use the guide and when not to, a table of contents linking to the Markdown version of every page, and a list of optional resources such as this page. For agents that need the whole guide at once, llms-full.txt contains every page in a single file. It’s much larger, so it’s better suited to broad questions that span many topics than to a single task.

Both files are cached, and only change when the guide’s pages change, so it’s safe to fetch them regularly.

Agent skill

An agent skill is a small set of instructions that tells an AI tool how to help with a specific task, in the Agent Skills format. Our Wagtail guide support skill tells an agent to use the guide as its authoritative documentation for questions about using the Wagtail admin interface, and to refer developer questions to the developer documentation.

You can copy-paste the skill into your preferred AI tool, or point the tool at the skill’s URL. The skill is also listed in the site’s agent skills index, so tools can discover it automatically.

MCP server

The Model Context Protocol (MCP) is an open standard that lets AI tools connect to outside services. Our public MCP server lets an agent search the guide and read its pages, without having to fetch and parse the site itself. The server provides two read-only tools: one to search the guide, and one to fetch a page as Markdown, including its title, description, and canonical URL.

Add https://guide.wagtail.org/mcp to your MCP client as a remote server, with no authentication. Some clients need to know which transport to use, in which case choose the Streamable HTTP one.

WebMCP

WebMCP lets a website offer tools to AI agents that browse it in a browser. On the guide, that means agents can search the guide and submit page feedback while you’re using the site. WebMCP is still experimental, and only works in browsers that support it, so it complements the other formats rather than replacing them.

API

The guide’s contents are also available through a Wagtail v3 API, which returns pages as JSON. Use it when you need the guide’s content in a structured format, for example to build a custom guide, or to automate part of your own documentation workflow.

The API is read-only and public. Its OpenAPI description lists every endpoint, so tools that support OpenAPI can discover them automatically.

Historical copies

Historical copies of the site contents are available in our Git repository. You can browse or compare how the documentation has changed over time, in the content directory in the wagtail/guide repository. Each sync stores every page in the Markdown format described in Markdown.

Feedback

We’d love to hear how you use these formats. Let us know what works, and what could be better, in our llms.txt feedback discussion thread.