Zotero MCP - Model Context Protocol Integration for Zotero

Zotero MCP is an open-source project designed to seamlessly integrate powerful AI capabilities with the leading reference management tool, Zotero, through the Model Context Protocol (MCP). This project consists of two core components: a Zotero plugin and an MCP server, which work together to provide AI assistants (like Claude) with the ability to interact with your local Zotero library. This README is also available in: :cn: 简体中文 | :gb: English.

Fork us on Wechat

| MP | Forum | | :--- | :---: | | | |

📚 Project Overview

The Zotero MCP server is a tool server based on the Model Context Protocol that provides seamless integration with the Zotero reference management system for AI applications like Claude Desktop. Through this server, AI assistants can:

• 🔍 Smart Search: Multi-dimensional library search (title/creator/year/tags/fulltext/semantic) with boolean operators and relevance scoring
• 📖 Content Extraction: Extract PDF full-text, notes, abstracts, webpage snapshots with fine-grained mode control
• 📝 Annotation Analysis: Search and analyze PDF highlights and annotations by color, tags, and keywords
• 📂 Collection Browsing: Browse and search collection hierarchies, retrieve items within collections
• 🧠 Semantic Search: AI-powered concept matching via embedding vectors, discover related literature across languages
• ✏️ Write Operations: Create notes, manage tags, update metadata, create new items and attach PDFs
• 💾 Full-text Database: Access and search cached PDF full-text content

This enables AI assistants to help you with literature reviews, citation management, content analysis, annotation organization, knowledge base management, and more.

🚀 Project Structure

This project now features a unified architecture with an integrated MCP server:

• zotero-mcp-plugin/: A Zotero plugin with integrated MCP server that communicates directly with AI clients via Streamable HTTP protocol
• IMG/: Screenshots and documentation images
• README.md / README-zh.md: Documentation files

Unified Architecture:

AI Client ↔ Streamable HTTP ↔ Zotero Plugin (with integrated MCP server)

This eliminates the need for a separate MCP server process, providing a more streamlined and efficient integration.

🚀 Quick Start Guide

This guide is intended to help general users quickly configure and use Zotero MCP, enabling your AI assistant to work seamlessly with your Zotero library.

1. Installation (For General Users)

What is Zotero MCP?

Simply put, Zotero MCP is a bridge connecting your AI client (like Cherry Studio, Gemini CLI, Claude Desktop, etc.) and your local Zotero reference management software. It allows your AI assistant to directly search, query, and cite references from your Zotero library, greatly enhancing academic research and writing efficiency.

Two-Step Quick Start:

1. Install the Plugin:

   • Go to the project's Releases Page to download the latest zotero-mcp-plugin-x.x.x.xpi file.
   • In Zotero, install the .xpi file via Tools -> Add-ons.
   • Restart Zotero.

2. Configure the Plugin:

   • In Zotero's Preferences -> Zotero MCP Plugin tab, configure your connection settings:
     • Enable Server: Start the integrated MCP server
     • Port: Default is 23120 (you can change this if needed)
     • Generate Client Configuration: Click this button to get configuration for your AI client

2. Connect to AI Clients

Important: The Zotero plugin now includes an integrated MCP server that uses the Streamable HTTP protocol. No separate server installation is needed.

Streamable HTTP Connection

The plugin uses Streamable HTTP, which enables real-time bidirectional communication with AI clients:

1. Enable Server in the Zotero plugin preferences
2. Generate Client Configuration by clicking the button in plugin preferences
3. Copy the generated configuration to your AI client

Supported AI Clients

• Claude Desktop: Streamable HTTP MCP support
• Cherry Studio: Streamable HTTP support
• Cursor IDE: Streamable HTTP MCP support
• Custom implementations: Streamable HTTP protocol

For detailed client-specific configuration instructions, see the Chinese README.

👨‍💻 Developer Guide

Prerequisites

• Zotero 7.0 or higher
• Node.js 18.0 or higher
• npm or yarn
• Git

Step 1: Install and Configure the Zotero Plugin

1. Download the latest zotero-mcp-plugin.xpi from the Releases Page.
2. Install it in Zotero via Tools -> Add-ons.
3. Enable the server in Preferences -> Zotero MCP Plugin.

Step 2: Development Setup

1. Clone the repository:

   git clone https://github.com/cookjohn/zotero-mcp.git
   cd zotero-mcp
   

2. Set up the plugin development environment:

   cd zotero-mcp-plugin
   npm install
   npm run build
   

3. Load the plugin in Zotero:

   # For development with auto-reload
   npm run start
   
   # Or install the built .xpi file manually
   npm run build
   

Step 3: Connect AI Clients (Development)

The plugin includes an integrated MCP server that uses Streamable HTTP:

1. Enable the server in Zotero plugin preferences
2. Generate client configuration using the plugin's built-in generator
3. Configure your AI client with the generated Streamable HTTP configuration

Example configuration for Claude Desktop:

{
  "mcpServers": {
    "zotero": {
      "transport": "streamable_http",
      "url": "http://127.0.0.1:23120/mcp"
    }
  }
}

🧩 Features

zotero-mcp-plugin Features

• Integrated MCP Server: Built-in MCP server using Streamable HTTP protocol, no separate process needed
• Advanced Search Engine: Full-text search with boolean operators, relevance scoring, filtering by title, creator, year, tags, item type, and more
• Unified Content Extraction: Extract content from PDFs, attachments, notes, abstracts, webpage snapshots with four modes (minimal/preview/standard/complete)
• Smart Annotation System: Search and retrieve PDF highlights, annotations, and notes by color, tags, and keywords with intelligent ranking
• Collection Management: Browse, search collection hierarchies, get collection details, subcollections, and item lists
• Semantic Search: AI-powered semantic search using embedding vectors
  • Supports OpenAI and Ollama embedding APIs (auto-detection)
  • Vector indexing with SQLite-vec storage
  • Index status column in main library view
  • Collection/item context menu for index management
• Write Operations: Create/modify notes, manage tags, update metadata fields, create new items and reparent standalone PDFs
• Full-text Database: Cached PDF full-text database with list, search, get, and stats operations
• Standalone Attachment Management: Search and manage standalone PDF items without parent metadata
• Client Configuration Generator: Automatically generates configuration for various AI clients
• Security: Local-only operation ensuring complete data privacy
• User-Friendly: Easy configuration through Zotero preferences interface

📸 Screenshots

Here are some screenshots demonstrating the functionality of Zotero MCP:

| Feature | Screenshot | | :--- | :---: | | Feature Demonstration | | | Literature Search | | | Viewing Metadata | | | Full-text Reading 1 | | | Full-text Reading 2 | | | Searching Attachments (Gemini CLI) | | | Reading PDF (Gemini CLI) | |

🔧 API Reference (MCP Tools)

The integrated MCP server provides 20 tools in 5 categories:

1. Search & Query (7 tools)

search_library

Advanced library search with multi-dimensional filtering, boolean operators, relevance scoring, and intelligent mode control.

• q, title, titleOperator, yearRange, fulltext, fulltextMode, itemType, includeAttachments, mode (minimal/preview/standard/complete), relevanceScoring, sort, limit, offset

search_annotations

Search annotations by query, colors, or tags with intelligent ranking.

• q, itemKeys, types (note/highlight/annotation/ink/text/image), colors, tags, mode, limit, offset

search_fulltext

Full-text search across all document content with context snippets.

• q (required), itemKeys, mode, contextLength, caseSensitive

search_collections

Search collections by name. Params: q, limit.

get_item_details

Get complete metadata for a single item. Params: itemKey (required), mode.

get_item_abstract

Get item abstract/summary.