Cursor

AI coding tools work best when given access to up-to-date documentation.

Lovely Docs places the documentation inside a folder ( lovely-docs/ by default) in your project. The documentation is available in both condensed and fulltext forms - noise is removed from pages, and content summaries are created for each directory.

Your coding agent should have an easy time searching and reading the documentation using its built-in tools, and you can also directly point it to any particular file using @lovely-docs/... or equivalent mechanism.

You can also use Lovely Docs MCP Server , but a local folder works better for AI coding agents.

Want me to add more libraries or have feedback? Reach out on Discord or .

Tokens ain't cheap, buy me a coffee .

Install directory summaries?

Example rules

You can add the following rule to your .cursorrules or similar configuration to help your AI agent use the documentation effectively:

We have dehydrated documentation in lovely-docs/. When you start a new session, check what's available. Refer to the docs first time you use a feature of a documented library.

Or a more agressive

We have curated, dehydrated documentation in lovely-docs/ WORKFLOW: 1. At the START of EVERY session, run: list_dir on lovely-docs/ 2. If documentation exists for a library you're about to use, read the relevant sections before using a feature for the first time.

{
  "mcpServers": {
    "lovely-docs": {
      "command": "npx",
      "args": [
        "-y",
        "lovely-docs@latest",
        "mcp"
      ]
    }
  }
}

File → Preferences → Cursor Settings → Features → MCP Servers → "Add New MCP Server" → Paste configuration