Greenlit Books

What Greenlit Books MCP server does

The Greenlit Books MCP server gives an AI client read-only access to the catalog and related material published on greenlitbooks.com. Its content covers practical books and reference pages about working with AI, including agent reliability, prompt injection, human involvement, durable execution, governance, and safety assessments.

The service can search the wider site for book pages, guides, field notes, and safety verdicts concerning AI agents, MCP servers, and coding tools. It can also search the catalog directly, identify books associated with a topic, return a series in reading order, and provide details for an individual book.

How it works

The Greenlit Books MCP server uses a stateless streamable HTTP connection at https://greenlitbooks.com/api/mcp. Responses are derived from the live catalog, and the repository does not run the server locally. No account, OAuth flow, API key, or rate-limit registration is required.

Most operations return source information alongside the requested content. Search results provide identifiers for use with fetch, which returns a page as Markdown with its canonical URL and dates. Book, concept, glossary, chapter, and claim results include links or citation details that can be used to inspect or reference the underlying material.

All tools are read-only. The plugin connection sends requests over HTTPS and does not read files from the local machine.

Setup and configuration

Add the endpoint as a remote MCP server in a compatible client. For Claude Code, the documented command is:

claude mcp add --transport http greenlit-books https://greenlitbooks.com/api/mcp

Claude Desktop or claude.ai can add it through Settings, Connectors, and Add custom connector. Clients using MCP JSON configuration can use:

{
  "mcpServers": {
    "greenlit-books": {
      "url": "https://greenlitbooks.com/api/mcp"
    }
  }
}

The repository also contains plugin manifests for Cursor, Grok Bot, and Grok Build. Those plugins package the remote connection and four skills; they do not install executable code, rules, commands, hooks, or scripts on the machine.

Tools and capabilities

The Greenlit Books MCP server exposes tools for several research workflows:

• Search the site or catalog, then fetch specific pages.
• Retrieve complete free first chapters, book metadata, concepts, reading paths, and topic matches.
• Define catalog terms, including related books, claims, and citation links.
• Check statements against the claim ledger, including the distinction between internal and external support.
• Find passages that answer a question in the free chapters, concept pages, glossary, guides, and claim ledger.
• Generate finished citations for books, free chapters, and coined terms.

Limitations and notes

The catalog intentionally does not provide publication dates, page counts, or ISBNs because the project says it has no verified source for them. define_term definitions are quoted and verified against the manuscript when the site is built. A search or claim check can return nothing when the catalog does not contain a relevant answer; that indicates the shelf has no matching material rather than a tool failure.

The Greenlit Books MCP server covers the catalog and site content, not general web search or write operations. Amazon links and book purchasing information may be returned by book-related tools, but the MCP tools themselves remain read-only.