How to use the LibreLeaf MCP server

Connect an MCP-capable client to LibreLeaf's read-only tools and inspect the source, rights, and ranking data it returns.

What the endpoint provides

LibreLeaf exposes a remote, read-only Model Context Protocol endpoint at https://libreleaf-books.netlify.app/mcp. MCP is an open protocol for giving compatible AI clients structured tools and data sources. The client, not the book site, decides how tool results are presented in a conversation.

The standard search and fetch tools return citation-ready work records. search_books supports focused catalogue queries, while resolve_access selects a canonical best match and returns every source-labelled offer with a ranking explanation. Results retain work IDs, source records, access type, and jurisdiction notes.

Connect and test

  1. In an MCP-capable client, add a remote Streamable HTTP server using the LibreLeaf /mcp URL. Client settings and workspace permissions vary, so use the current instructions for that client.
  2. List the available tools before calling one. This confirms that the endpoint and client agree on the live schemas.
  3. Try resolve_access with title Pride and Prejudice, author Jane Austen, and region GB. Review the selected match, why it ranked, and the applicability note on each offer.
  4. Use fetch with the returned stable LibreLeaf work ID when you need to refresh that canonical work rather than repeating an ambiguous text search.

Treat tool output as evidence, not clearance

MCP responses identify which upstream catalogue supplied a claim. They do not turn a US source assessment into a UK legal conclusion. Pass GB, US, or GLOBAL deliberately and keep the returned rights context with any recommended route.

Before production use, test the endpoint with MCP Inspector, handle partial source failures, and show source links to the user. Do not automatically download every offered file or describe a borrow or preview route as a permanent download. OpenAI's current guidance also recommends reviewing MCP server trust and tool behaviour before connecting it to a model workflow.

The endpoint does not require a LibreLeaf account. Availability of remote MCP connections in a particular client or managed workspace can depend on that product's current settings and administrator policy.