Zotero MCP Server

Overview of Zotero MCP

Zotero MCP bridges the gap between a local Zotero library and an AI assistant that communicates via the Model Context Protocol. By exposing Zotero’s rich bibliographic database as an MCP server, developers can give their assistants direct access to reference metadata and full‑text documents without relying on third‑party integrations such as the Claude app or Cursor. This eliminates the need for manual export/import workflows and allows AI agents to query, retrieve, and manipulate references in real time.

The server is lightweight yet feature‑rich. It authenticates with Zotero using an API key and library ID, then exposes two primary capabilities: resource retrieval and tool execution. The resource capability lets an AI client list collections, search for items by title or author, and fetch the full text of PDFs or other attached files. The tool capability provides actions such as adding a new reference, updating tags, or deleting an item. Because the server implements MCP’s standard and interfaces, any compliant AI client can discover and use these functions automatically.

Key features include:

• Seamless integration with any LLM that supports MCP, requiring only the model’s API base and key.
• Real‑time access to the latest library contents, ensuring that references are always up‑to‑date.
• Full‑text search across PDFs and other attachments, enabling AI assistants to pull out specific passages or citations.
• Modular design: the server can be extended with additional Zotero endpoints (e.g., collections, tags) without changing the client logic.

Typical use cases span research assistants, academic writing tools, and knowledge‑management bots. A researcher can ask an AI to “summarize the latest article on machine learning in their Zotero library,” and the assistant will retrieve, parse, and synthesize the content on demand. In a collaborative setting, an AI can automatically tag new imports or flag duplicates based on user-defined rules. For developers building custom workflows, the MCP interface allows embedding Zotero queries directly into conversational agents or automated scripts.

Compared to conventional approaches that rely on static exports or manual API calls, Zotero MCP offers a unified, protocol‑driven interface that reduces friction and enhances reliability. Its ability to expose both read and write operations within a single, well‑documented server makes it an attractive choice for any team that wants to harness the power of Zotero in AI‑driven applications.