AI tools

These docs are readable by agents as well as people. Nothing here needs an account or a key.

MCP server

The endpoint at https://qrate.dvnl.work/mcp speaks Model Context Protocol over stateless HTTP. It answers with three tools: search_docs, get_doc, and list_docs.

Claude Code:

claude mcp add --transport http qrate-docs https://qrate.dvnl.work/mcp

Cursor, in .cursor/mcp.json:

{
  "mcpServers": {
    "qrate-docs": { "url": "https://qrate.dvnl.work/mcp" }
  }
}

A client that only speaks stdio can bridge with mcp-remote:

{
  "mcpServers": {
    "qrate-docs": {
      "command": "npx",
      "args": ["mcp-remote", "https://qrate.dvnl.work/mcp"]
    }
  }
}

The tool catalog is at /mcp-schema.json if you want to read what the server offers before connecting.

Plain markdown

Add .md to any docs URL for the source markdown, without navigation, styles, or scripts. For example /docs/columns.md.

llms.txt

• /llms.txt — the page index
• /llms-full.txt — every page in one file
• /llms-small.txt — the same, trimmed for smaller context windows

The Agent panel

Reading the docs is separate from reading a project. An agent running on your own machine can read the project open in qrate through the Agent panel, which is a different mechanism with its own permissions.