calibre-mcp

A Model Context Protocol (MCP) server for searching and reading books from your Calibre ebook library.

Features

• Full-text search with automatic phrase matching and fuzzy fallback
• Search by author or title with partial matching
• Extract text excerpts from books with keyword highlighting and pagination
• Pure bash implementation using `calibredb` CLI
• Returns Calibre deep links (`calibre://`) and file URLs (`file://`)
• Robust timeout handling for macOS compatibility

Prerequisites

• Calibre installed at `/Applications/calibre.app`
• A Calibre library at `~/Calibre Library`
• Text (.txt) exports of your books for excerpt functionality
• Bash 4.0 or later

Installation

1. Clone this repository:

git clone https://github.com/trieloff/calibre-mcp.git
cd calibre-mcp

2. Make the script executable:

chmod +x calibre-mcp.sh

Usage

With MCP Inspector

Test the server using the MCP Inspector:

npx @modelcontextprotocol/inspector /path/to/calibre-mcp.sh

With Claude Desktop

Add to your Claude Desktop configuration:

{
  "mcpServers": {
    "calibre": {
      "command": "/path/to/calibre-mcp.sh"
    }
  }
}

Available Tools

search-fulltext

Search books using Calibre's full-text search engine.

Parameters:

• `query` (required): Search query. Multi-word queries are automatically phrase-searched
• `limit` (optional): Maximum results (default: 50)
• `fuzzy_fallback` (optional): Fallback query with related terms if exact search fails

Example:

{
  "query": "machine learning",
  "fuzzy_fallback": "AI artificial intelligence ML neural networks deep learning"
}

search-author

Search books by author name with partial matching.

Parameters:

• `author` (required): Author name (partial match supported)
• `limit` (optional): Maximum results (default: 50)

search-title

Search books by title with partial matching.

Parameters:

• `title` (required): Book title (partial match supported)
• `limit` (optional): Maximum results (default: 50)

get-excerpt

Extract text excerpts from books with keyword context.

Parameters:

• `book_id` (required): Calibre book ID
• `keyword` (optional): Search term to find in the book
• `context_lines` (optional): Lines of context around matches (default: 10)
• `max_results` (optional): Results per page (default: 10)
• `page` (optional): Page number for pagination (default: 1)

Response Format

Search results include:

• Book metadata (title, authors, series, tags, publisher)
• Calibre deep link for opening in Calibre
• File URL for direct EPUB access
• Available formats
• Whether text format is available for excerpts

Logging

The server logs all requests and responses to `requests.log` for debugging.

Limitations

• Requires Calibre to be closed or may experience timeouts
• Full-text search requires indexed books
• Excerpt extraction only works with .txt exports

License

Apache License 2.0 - See LICENSE file for details.