Which integrations are available for this server?

Provides programmatic access to a [Zotero](/mcp/servers/integrations/zotero) library, allowing users to search papers, manage notes, and access repository content through the MCP server

Zotero MCP Server

A MCP (Model Context Protocol) server to let your MCP clients (e.g. Anthropic Claude App, Goose, possibly vscode Cline too) interact with your local Zotero repository. This server enables programmatic access to your Zotero library, allowing you to search papers, manage notes, and more.

Note: If you don't want to set up API keys, see the SQLite database server option below.

Setup

Install dependencies:

pip install -e .

Create a .env file in the root directory with your Zotero credentials:

ZOTERO_API_KEY=your_api_key_here ZOTERO_USER_ID=your_user_id_here

You can get your Zotero API key and user ID from Zotero's settings page.

Related MCP server: Zotero MCP Server

Integration with Anthropic Desktop App

To integrate with the Anthropic Desktop app, add the following configuration to ~/Library/Application Support/Claude/claude_desktop_config.json:

{ "mcpServers": { "zotero-mcp-server": { "command": "uv", "args": [ "--directory", "/Users/swairshah/work/research/zotero-mcp", "run", "python", "-m", "zotero_mcp.server" ] } } }

If this gives an error like

{"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"claude-ai","version":"0.1.0"}},"jsonrpc":"2.0","id":0} error: unexpected argument '--directory' found

Then use the following config, make sure to do uv venv; source .venv/bin/activate; uv pip install ".[dev]" to make sure the server can be run with all dependencies.

{ "mcpServers": { "zotero-mcp-server": { "command": "bash", "args": [ "-c", "cd /Users/shahswai/personal/zotero-mcp-server && source .venv/bin/activate && python -m zotero_mcp.server" ] } } }

Alternative: Direct SQLite Database Access

If you prefer to bypass the Zotero API entirely and work directly with the SQLite database, use zotero_mcp/db_server.py. This approach gives you full control over your Zotero data without API limitations. Note that you'll need to close Zotero completely before using this method since SQLite locks the database when Zotero is running.

Claude MCP config for the SQLite version:

{ "mcpServers": { "zotero-mcp-server": { "command": "uv", "args": [ "--directory", "/Users/swair/work/code/zotero-mcp-server", "run", "python", "-m", "zotero_mcp.db_server" ] } } }