home / mcp / zotero mcp server

Zotero MCP Server

Provides local AI access to search, retrieve, and cite items from your Zotero library via Streamable HTTP.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "cookjohn-zotero-mcp": {
      "url": "http://127.0.0.1:23120/mcp"
    }
  }
}

Zotero MCP connects your AI client directly to your local Zotero library through the Model Context Protocol, enabling fast searches, detailed reference data, and precise citations while keeping data local for privacy.

How to use

You will use an MCP client to talk to the integrated Zotero MCP server inside the Zotero plugin. Enable the internal server, generate a client configuration, and supply that configuration to your AI client. With Streamable HTTP you’ll get real-time bidirectional communication, so your AI assistant can search your library, fetch item details, locate items by identifiers like DOI or ISBN, and filter results by tags, year, and more.

Typical usage patterns include: searching your Zotero library for relevant papers, opening item metadata, retrieving PDF highlights and notes, and matching items by identifiers to ensure precise citations. Your AI client uses the generated configuration to connect to the local MCP endpoint at http://127.0.0.1:23120/mcp and exchanges requests with the Zotero plugin.

How to install

Prerequisites and setup steps ensure you have the right environment to run the integrated MCP server inside the Zotero plugin.

1) Install prerequisites:

- Node.js 18.0 or higher

- npm or yarn

- Git

2) Install and configure the Zotero plugin

- Download the latest Zotero MCP plugin package and install it through Zotero’s Add-ons manager.

- In Zotero, open Preferences, find the Zotero MCP Plugin section, and enable the integrated server. The default port is 23120. Ensure the server is started so the AI client can connect.

3) Generate client configuration

- In the Zotero MCP Plugin preferences, click Generate Client Configuration. This creates the configuration you will load into your AI client to connect via Streamable HTTP.

4) Optional development workflow (for developers)

- Clone the project repository to your local machine.

- Install plugin development dependencies and build the project.

- Start the development server and use auto-reload during work.

Additional configuration and example

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

Notes on security and usage

This integration operates locally, ensuring complete data privacy. The MCP server runs on your machine and communicates with AI clients over the Streamable HTTP protocol. Treat the local endpoint as a trusted interface and keep your environment secure.

Available tools

search_library

Searches the Zotero library with filters like q, title, creator, year, tag, itemType, and supports limiting and sorting.

get_item_details

Retrieves full information for a single item by its itemKey.

find_item_by_identifier

Finds an item by DOI or ISBN, requiring at least one identifier.