Access Elastic Docs in machine-readable formats

The Elastic Docs website is one way to access Elastic Docs. The same content is also available as per-page Markdown, bulk archives, through a remote Model Context Protocol (MCP) server, and from the command line. Use this page to select the access method that best fits your tool.

All channels on this page cover the current documentation set: Elastic Stack 9.0 and later, and Elastic Cloud Serverless. Documentation for earlier versions at elastic.co/guide is not included. Refer to Versioning and availability for details.

Every page on elastic.co/docs has a Markdown equivalent. This is a machine-readable rendition, not the MyST authoring source. The rendition includes:

• Generated frontmatter: title, description, url, products, applies_to
• Substitutions resolved (for example, {{stack}} becomes Elastic Stack)
• Directives flattened to prose or CommonMark
• All links rewritten to absolute https://www.elastic.co/docs/... HTML URLs

To find the Markdown source, use Edit this page on the page you're viewing.

Append .md to any documentation page URL. For example, this page is available at:

https://www.elastic.co/docs/get-started/machine-readable-docs.md
		

Two URL rules to keep in mind:

• Section landing pages sit one level up. The Markdown for https://www.elastic.co/docs/get-started is at https://www.elastic.co/docs/get-started.md, not https://www.elastic.co/docs/get-started/index.md (which returns a 404).
• The docs homepage has no Markdown equivalent at /docs.md. Use llms.txt to get an overview of the whole site instead.

The View as Markdown link on each documentation page opens the same .md URL.

You can also fetch the HTML page URL and request Markdown by adding an Accept header. Send Accept: text/markdown or Accept: text/plain. A wildcard (*/*) or browser-default Accept returns HTML, so an explicit header is required.

curl -H "Accept: text/markdown" https://www.elastic.co/docs/get-started
		

Responses include Vary: Accept, so caches handle the two representations correctly.

Three bulk artifacts cover the whole documentation site. Use the index (llms.txt) when you need a map of what exists. Use the Markdown bundle (llm.zip) to seed a local RAG index or process each page. Use the structured bundle (okf.zip) to import Elastic docs into a knowledge-management platform that supports the Open Knowledge Format.

The https://www.elastic.co/docs/llms.txt file follows the llmstxt.org convention: a prose introduction to Elastic Docs followed by per-section link lists with titles, descriptions, and absolute URLs.

The file is an index of where things are, not the content itself. It covers the top-level navigation groups and their first-level children. Use it to orient an agent or provide a high-level map of the documentation structure.

The https://www.elastic.co/docs/llm.zip file is a ZIP archive containing one .md file per documentation page, mirroring the URL tree. Each file uses the same frontmatter and link format as the per-page Markdown endpoint, so the two channels are interchangeable in content.

The archive is tens of megabytes in size. Two things to expect when processing it:

• It includes internal snippet fragments (files under _snippets/ paths) that are not independently addressable pages.
• It includes an llms.txt file at the root, but this copy might not match the version served at https://www.elastic.co/docs/llms.txt. Use the live URL for the authoritative index.

The https://www.elastic.co/docs/okf.zip file is a ZIP archive conforming to Open Knowledge Format v0.1. The bundle is tens of megabytes in size.

The OKF bundle differs from llm.zip in several ways:

There is no per-page OKF endpoint. The bundle is the only distribution format.

The Elastic Docs MCP server is available at:

https://www.elastic.co/docs/_mcp/
		

It uses the streamable HTTP transport and requires no authentication. Every request is an independent POST with no session header.

claude mcp add --transport http elastic-docs https://www.elastic.co/docs/_mcp/
		

{
  "mcpServers": {
    "elastic-docs": {
      "url": "https://www.elastic.co/docs/_mcp/"
    }
  }
}
		

{
  "servers": {
    "elastic-docs": {
      "type": "http",
      "url": "https://www.elastic.co/docs/_mcp/"
    }
  }
}
		

The Elastic plugin on the Cursor Marketplace bundles Elastic agent skills for Elasticsearch, Kibana, Observability, Security, and Cloud together with an elastic-docs MCP entry that points at the same server endpoint above. Installing the plugin is an alternative to manual Cursor configuration. The plugin source is available in the public elastic/cursor-plugins repository.

The docs commands in the Elastic CLI let you search, ask questions about, and read Elastic documentation directly from a terminal.

Install the CLI with npm:

npm install -g @elastic/cli
		

Or run without installing:

npx -y @elastic/cli docs --help
		

The docs commands don't require authentication.

Three commands are available:

• elastic docs search <query> — search the documentation and return matching pages with summaries.
• elastic docs ask <question> — get an AI-generated answer based on the documentation.
• elastic docs read <path | URL | free text> — fetch a page as Markdown. This command reads the same .md endpoint described in Get a single page as Markdown, making it the terminal equivalent of that channel. Passing free text reads the top search result.

Pass --json to any command for structured output suitable for scripts or agents. Pass --accept-experimental to search and ask to suppress the experimental warning.

Refer to Elastic CLI docs commands for the full command reference.