54yyyu/zotero-mcp

What it solves

Zotero MCP allows AI assistants (like Claude, ChatGPT, and Cursor) to interact directly with a user's Zotero research library. It eliminates the need to manually export and upload papers or notes, enabling AI to search, summarize, and analyze a personal research collection in real-time.

How it works

The project implements the Model Context Protocol (MCP), acting as a bridge between Zotero and AI clients. It can connect via a local API for fast read-only access, a web API for cloud library management, or a hybrid mode that combines both.

For advanced discovery, it includes a semantic search system that uses embedding models (local via Ollama/sentence-transformers or cloud via OpenAI/Gemini) to index the library in a vector database (ChromaDB), allowing the AI to find papers based on conceptual meaning rather than just keywords.

Who it’s for

Researchers, academics, and students who use Zotero to manage their citations and PDFs and want to leverage LLMs to synthesize information across their entire library.

Highlights

• Semantic Search: Vector-based similarity search to find conceptually related research.
• Library Management: Ability to add papers via DOI or URL, manage collections, and merge duplicates.
• Annotation Extraction: Retrieves PDF annotations and outlines for AI analysis.
• Citation Intelligence: Optional integration with Scite to track citation tallies and retraction alerts.
• Flexible Access: Supports local, web, and hybrid connection modes.
• Agent Skill: A lightweight CLI installation option that reduces token cost for agents with shell access.