home / mcp / book fetch mcp server

Book Fetch MCP Server

Provides access to book content via MCP with local caching and paging for long texts.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "kinshukk-book-fetch-mcp": {
      "command": "uv",
      "args": [
        "--directory",
        "<PATH_TO_PARENT_DIR>/libgen-mcp",
        "run",
        "main.py"
      ]
    }
  }
}

Book Fetch MCP lets you talk to any published book inside your MCP client directly, with caching and paging for long texts. It provides a local MCP server you can run to fetch book content on demand, making it easy to ask questions and retrieve passages from a vast library without leaving your client environment.

How to use

You run the Book Fetch MCP locally and connect to it from your MCP client. The server exposes a toolchain that retrieves book content, caches long texts, and paginates results so you can read chapter-sized chunks without overwhelming the client. Use it to search, request specific pages or chapters, and prompt the client to summarize or extract quotes from a book. When you begin a long query, your MCP client can rely on the built-in cache to deliver results quickly and maintain context across pages.

How to install

Prerequisites you need before starting: you must have the uv runtime installed on your system.

Step 1: Verify uv is available on your path.

Step 2: Prepare the Book Fetch MCP directory and dependencies.

Step 3: Start the MCP server using the following runtime command specification shown in the configuration snippet.

{
  "mcpServers": {
    "book_fetcher": {
      "command": "uv",
      "args": [
        "--directory",
        "<PATH_TO_PARENT_DIR>/libgen-mcp",
        "run",
        "main.py"
      ]
    }
  }
}

Additional content

Configuration shows a local, stdio-based MCP server that you run with a runtime command. If you need to customize the path to the library or the script, replace the placeholder with your actual path. The server name is book_fetcher, and it runs via the uv runner with a directory pointing to the library content and a Python entry script main.py.

Available tools

book_fetch

Retrieves book content on demand from the library, caches it in MCP, and serves pages or chapters on request.

paginate_cache

Maintains a cache of long books and provides paginated responses to MCP clients to avoid exceeding context limits.

mini_rag_engine

Optionally spawns a lightweight retrieval-augmented generation style workflow to improve long-context answers by re-ranking chunks before returning results.

scihub_integration

Plan to integrate Sci-Hub sources in the future to broaden access to public-domain and accessible texts.