Zotero MCP Guide

What is Zotero MCP?

Since BibGenie v0.4.0, we have introduced native Zotero MCP (Model Context Protocol) support. This feature aims to provide users with a more flexible way to integrate Zotero AI.

Simply put, the MCP Server acts like a bridge, allowing your favorite external AI tools (such as Claude, Cursor, VS Code, etc.) to directly "read" and "operate" your Zotero library.

Core Advantages

• Out-of-the-box: BibGenie's MCP Server is built-in. As long as you have the BibGenie plugin installed, the MCP service starts automatically, without needing extra Python or Node.js environment installation.
• Local & Secure: All data interactions happen directly on your local device, ensuring privacy and security.
• Efficient Streaming: Based on Streamable HTTP protocol for fast response speeds.

Core Features (Tools)

Currently, BibGenie MCP provides the following core tools, optimized for research scenarios:

1. Get Items in Collection
   • Retrieves the full list of items under a specific Zotero Collection.
   • Best for: Letting AI quickly understand all literature in a specific research direction.
2. Search Items
   • Quickly retrieves from the Zotero library via keywords.
   • Supports precise filtering via metadata like Tags, Date, etc.
   • Layered design: Search returns only brief information to maximize token saving and improve speed.
3. Get Item Details
   • Retrieves all metadata information for a specific Item (such as Notes, Tags, Abstract, citations, etc.).
   • Note: This tool does not return the full PDF text; instead, it guides the model via prompt injection to automatically call the Get PDF Text tool when needed to retrieve the full text.
4. Get PDF Text
   • Directly extracts and lets AI read the plain text content of PDF attachments in Zotero.
   • The model will automatically use this tool to retrieve full PDF text when needed.

How to Use

The MCP Server type provided by BibGenie is Streamable HTTP. The default service address is: http://127.0.0.1:23119/bibgenie/mcp

Preparation

Before connecting, please ensure that the BibGenie plugin is installed in Zotero and the version is v0.4.0 or above.

You can directly open http://127.0.0.1:23119/bibgenie/mcp in your browser to test if the MCP Server is running normally. If you see information similar to the following, it means zotero mcp is working properly:

{
  "name": "bibgenie-mcp-server",
  "version": "1.0.0",
  "protocolVersion": "2024-11-05",
  "capabilities": {
    "tools": {
    }
  },
  "status": "running"
}

Connect Cursor

Cursor is currently the best AI code editor experience. By connecting to MCP, it can directly reference your papers to answer questions.

{
  "mcpServers": {
    "ZoteroBibGenie": {
      "url": "http://127.0.0.1:23119/bibgenie/mcp"
    }
  }
}

Connect Claude Desktop

For Claude Desktop, the latest version supports managing MCP via a graphical interface. This makes integrating Zotero into your Claude Desktop easier than ever.

{
  "mcpServers":{
    "ZoteroBibGenie":{
      "command":"npx",
      "args":[
        "mcp-remote",
        "http://localhost:23119/bibgenie/mcp"
      ]
    }
  }
}

Connect Claude Code

If you use the command line tool claude-code, you can quickly add it using the following command:

claude mcp add ZoteroBibGenie --transport http http://127.0.0.1:23119/bibgenie/mcp

For more information, please refer to Claude Code MCP Docs.

Connect Cherry Studio

Cherry Studio is a powerful local AI client that supports multi-model management and also perfectly supports the MCP protocol.

Use Case Examples

Once connected, you can try the following Prompts to experience the power of Zotero Copilot:

Literature Retrieval

Deep Reading

Knowledge Base Management

FAQ