zotero-mcp-lite 0.1.3

Fast, Customizable & Light Zotero MCP server for AI research assistants

Zotero MCP Lite

A high-performance Model Context Protocol (MCP) server for Zotero with customizable research workflows.

• Full Local - No cloud, no API key; runs entirely via Zotero Desktop
• Atomic Tools - 9 composable tools; LLM orchestrates as needed
• MCP-Native - Works with Claude, Cursor, Gemini CLI, and any MCP client
• Extensible - User-editable prompts for customized workflows
• Easy Deploy - Single command install, auto-detects Zotero
• Parameterized - Customize analysis sections to match your research style

Architecture

flowchart LR
    subgraph MCP [Zotero MCP Lite]
        Read[Read Operations]
        Write[Write Operations]
    end
    
    subgraph Zotero [Zotero Desktop]
        LocalAPI["/api/ endpoint"]
        ConnectorAPI["/connector/ endpoint"]
        SQLite[(zotero.sqlite)]
    end
    
    Read -->|GET| LocalAPI
    Read -->|Direct SQL| SQLite
    Write -->|POST saveItems| ConnectorAPI
    LocalAPI --> SQLite
    ConnectorAPI --> SQLite

Operation	Endpoint	Method
Search, browse, read metadata	/api/users/0/items	GET
Read annotations	Direct SQLite query	SQL
Create notes/items	/connector/saveItems	POST

Quick Start

Prerequisites

1. Python 3.10+ with uv installed
2. Zotero 6 or 7 installed

Step 0: Enable Zotero Local API

The Local API allows third-party applications to communicate with Zotero. This is required.

Steps:

1. Open Zotero → Edit → Settings (or Preferences on macOS)
2. Go to Advanced tab
3. Under Miscellaneous, check "Allow other applications on this computer to communicate with Zotero"

┌─────────────────────────────────────────────────────────────┐
│ Advanced                                                    │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│ Miscellaneous                                               │
│                                                             │
│ [V] Automatically check for updated translators and styles  │
│                                                             │
│ [V] Report broken site translators                          │
│                                                             │
│ [V] Allow other applications on this computer to            │
│     communicate with Zotero                   <- Check this │
│                                                             │
│     Available at http://localhost:23119/api/                │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Once enabled, the API will be available at http://localhost:23119/api/

Note: This is different from the Zotero Connector (browser extension) which is enabled by default. The Local API requires explicit opt-in for security reasons.

Step 1: Install

uv tool install zotero-mcp-lite

Or with pip:

pip install zotero-mcp-lite

Alternative: Install from GitHub or source

# From GitHub (latest development version)
uv tool install "git+https://github.com/xmruuu/zotero-mcp-lite.git"

# From source (for development)
git clone https://github.com/xmruuu/zotero-mcp-lite.git
cd zotero-mcp-lite && uv sync
uv run zotero-mcp setup

# Direct run without install
uvx zotero-mcp-lite serve

Step 2: Setup

zotero-mcp setup

This detects your Zotero installation and configures MCP clients automatically.

Step 3: Connect to MCP Client

Client	Configuration
Claude Code	claude mcp add zotero -- zotero-mcp serve
Claude Desktop	Auto-configured by setup, or edit claude_desktop_config.json
Cursor	Settings → MCP Servers → add {"zotero": {"command": "zotero-mcp", "args": ["serve"]}}
Other	Use command zotero-mcp with args ["serve"] (stdio transport)

That's it! You're ready to use Zotero with AI assistants.

Features

9 Atomic MCP Tools

Search and Navigation

• zotero_search_items - Keyword search with tag filtering
• zotero_get_recent - Recently modified/added items (excludes notes by default)
• zotero_get_collections - List all collections
• zotero_get_collection_items - Items in a collection (excludes notes by default)
• zotero_search_annotations - Search all PDF highlights across library

Content Reading

• zotero_get_item_metadata - Metadata, authors, abstract, tags
• zotero_get_item_children - Attachments, notes, and PDF annotations
• zotero_get_item_fulltext - Full text extraction

Writing (via local Connector API)

• zotero_create_note - Create note with full formatting support (tables, lists, line breaks)

4 Research Skills (MCP Prompts)

Pre-defined workflows that guide AI through common academic tasks:

flowchart LR
    A[knowledge_discovery] --> B[literature_review]
    B --> C[comparative_review]
    C --> D[bibliography_export]
    
    A -.- A1["Find relevant papers"]
    B -.- B1["Analyze single paper"]
    C -.- C1["Compare multiple papers"]
    D -.- D1["Export citations"]

Skill	Use Case	What It Does
knowledge_discovery(query)	Explore a topic	Searches titles AND your annotations
literature_review(item_key)	Deep-dive one paper	Structured analysis from annotations or full text
comparative_review(item_keys)	Compare papers	Table-rich synthesis with insights
bibliography_export(item_keys)	Prepare citations	APA, IEEE, and BibTeX formats

Key Features:

• Dual-mode analysis: Works with or without user annotations
• Full formatting preserved: Tables, line breaks, and lists write correctly to Zotero
• Fully customizable: Edit prompt files to match your workflow
• See Customizing Skills for details

For Claude Desktop Users: These 4 MCP Prompts automatically become powerful Skills in Claude. Simply ask Claude to help with literature review, paper comparison, or bibliography export - it will invoke the appropriate skill with your Zotero library.

Advanced

Testing & Debugging

npx @modelcontextprotocol/inspector zotero-mcp serve

Opens a web UI to test tools interactively.

Alternative Transports

zotero-mcp serve --transport streamable-http --port 8000
zotero-mcp serve --transport sse --port 8000

Environment Variables

# Direct path to Zotero database (most specific)
export ZOTERO_DATABASE_PATH=/path/to/zotero.sqlite

# Or specify Zotero data directory (will find database automatically)
export ZOTERO_DATA_DIR=/path/to/Zotero

# Group libraries
export ZOTERO_LIBRARY_ID=12345
export ZOTERO_LIBRARY_TYPE=group

Technical Notes

Annotation Retrieval

Annotations are fetched directly from Zotero's SQLite database using a three-level join:

Item -> PDF Attachment -> Annotation

This approach is significantly faster than the Web API and works offline.

Cross-Platform Support

Automatically detects Zotero data directory on Windows, macOS, and Linux.

Troubleshooting

Problem	Solution
"Local API is not enabled"	Enable in Zotero → Settings → Advanced (see Step 0)
"Database is locked"	Normal when Zotero is running; API tools still work
General diagnostics	Run zotero-mcp setup

Customizing Skills

Prompts are fully customizable. Copy from the package defaults and edit:

~/.zotero-mcp/prompts/
├── literature_review.md      # Single paper analysis skill
├── comparative_review.md     # Multi-paper comparison skill
├── knowledge_discovery.md    # Topic exploration skill
└── bibliography_export.md    # Citation export skill

Loading order: User files (~/.zotero-mcp/prompts/) take priority over package defaults.

How It Works

flowchart LR
    Prompt[prompt.md] --> LLM[LLM Analysis]
    LLM --> Review[Formatted Review]
    Review --> Note["zotero_create_note()"]
    Note --> Zotero[Zotero Note]

The LLM generates the complete review text with your desired formatting (tables, lists, etc.), then saves it directly via zotero_create_note. All formatting is preserved.

Prompt Files (.md)

Define how the AI should analyze papers:

## Analysis Instructions

Analyze the paper covering:
- Research Objective
- Methods
- Contribution
- Research Gaps

Use tables and bullet points as appropriate.

## Save to Zotero

Use `zotero_create_note(content=<your review>, parent_key=<item_key>)`

You can customize the analysis sections, add new ones, or change the output format entirely

Credits

Thanks to @54yyyu for the original zotero-mcp project.

License

MIT License - See LICENSE file.