LLM-friendly documentation (llms.txt)

Konflux publishes an llms.txt file — a structured index that lets Large Language Models discover and consume the documentation at inference time. The file is automatically regenerated on every documentation build, so it always reflects the current state of the docs.

What is llms.txt?

llms.txt is a lightweight specification for providing LLM-readable content on websites. It is a Markdown file containing a title, a short description, and a categorised list of links to documentation pages. AI coding assistants and MCP-enabled IDEs can read this file to find relevant documentation without scraping HTML.

Using with an MCP server (mcpdoc)

mcpdoc is an open-source MCP server purpose-built for serving llms.txt files. It exposes two tools to MCP host applications (Cursor, Windsurf, Claude Code/Desktop):

  • list_doc_sources — lists available documentation sources.

  • fetch_docs — fetches content from URLs found in the llms.txt file.

Konflux llms.txt URLs

Use the following names and URLs when configuring mcpdoc. You can add one or both depending on the documentation you need:

Name URL

Konflux Documentation

https://konflux-ci.dev/docs/llms.txt

Konflux Operator And Administrator Documentation

https://konflux-ci.dev/konflux-ci/llms.txt

Quick start (Cursor example)

Open Cursor Settings > MCP to edit your ~/.cursor/mcp.json, then add:

{
  "mcpServers": {
    "konflux-docs": {
      "command": "uvx",
      "args": [
        "--from",
        "mcpdoc",
        "mcpdoc",
        "--urls",
        "Konflux Documentation:https://konflux-ci.dev/docs/llms.txt",
        "Konflux Operator And Administrator Documentation:https://konflux-ci.dev/konflux-ci/llms.txt",
        "--transport",
        "stdio"
      ]
    }
  }
}

For best results, add a rule to your Cursor User Rules (Settings > Rules):

For ANY question about Konflux — builds, testing, releases, compliance, or
installation — use the konflux-docs MCP server:
1. Call list_doc_sources to see available llms.txt files
2. Call fetch_docs to read the llms.txt index
3. Identify URLs relevant to the question
4. Call fetch_docs on those URLs
5. Use the retrieved content to answer

Other IDEs and clients

mcpdoc supports Cursor, Windsurf, Claude Code, and Claude Desktop. For setup instructions for each client, see the mcpdoc Quickstart guide. The configuration is the same across clients — only the --urls value is Konflux-specific:

Konflux Documentation:https://konflux-ci.dev/docs/llms.txt
Konflux Operator And Administrator Documentation:https://konflux-ci.dev/konflux-ci/llms.txt

Testing with the MCP Inspector

Start mcpdoc in SSE mode:

uvx --from mcpdoc mcpdoc \
    --urls "Konflux Documentation:https://konflux-ci.dev/docs/llms.txt" \
    --transport sse \
    --port 8082 \
    --host localhost

Then launch the MCP Inspector:

npx @modelcontextprotocol/inspector

In the Inspector UI, set the transport to SSE, enter http://localhost:8082/sse, and click Connect. Use the Tools tab to call list_doc_sources and fetch_docs.

How llms.txt is generated

The file is produced by hack/generate-llms-txt.sh, which scans the Antora content tree and extracts each page’s title and introductory paragraph from the AsciiDoc source files. It uses the nav.adoc files to determine page ordering and section structure.

The script runs automatically as part of the CI build pipeline, so any change to the documentation — new pages, updated titles, restructured navigation — is reflected in llms.txt on the next deploy.