Skip to content

Zotero MCP

by Steven YuConnectorResearch something

Lets your AI assistant search, read, summarise and annotate papers in your Zotero library

View on GitHub

About Zotero MCP

Connects your Zotero reference library to AI assistants such as Claude, ChatGPT and Codex: search papers by title, tag or meaning, read them, add highlights and notes, and add papers by DOI or link.

For Graduate studentsResearchersResearch labs

Made by Steven Yu

An open-source project by developer Steven Yu, published on GitHub.

Stars
★ 5.7k
License
MIT
Cost
Free
Level
Intermediate
Runs
Local

Screenshots

From the project's own repository · 2 images, swipe or use the arrows.

Zotero with a paper open and the Zotero Agent chat panel explaining one passage, with page citations that highlight the quoted sentence
Zotero’s library view with the Zotero Agent chat panel listing which papers in a collection trace one research idea
Image: Steven Yu (MIT). Original file

Overview

Zotero MCP connects your Zotero reference library to AI assistants such as Claude, ChatGPT, Codex and Cursor. Your assistant can search by title, tag, full text or meaning, read papers and their BibTeX, add highlights and notes, and add new papers by DOI or URL. The same package also offers a command-line route for agents and a chat panel inside Zotero. It suits students, researchers and labs that already keep their papers in Zotero.

It works on the library you already have, and in local mode reads come straight from Zotero on your own computer.

Try these first

Start with the first one: it is a small task you can check easily.

  1. List five papers in my Zotero library about [topic], with authors and year. Don't change anything.First try · read-only

  2. Summarise the key findings of [paper title] and give me its BibTeX.

  3. Highlight the main claims in the PDF of [paper title] and add a short note.writes to Zotero

  4. Add this paper to my [collection] collection: [DOI]writes to Zotero

Example prompts written by Agent Flows Hub.

Capabilities

From the creator's documentation.

ToolsetWhat your agent can doAccess
SearchLook up papers by title, author, tag, collection, words in the text, or meaningRead
ReadItem details, BibTeX citations, full text and chosen pages of a PDFRead
AnnotateHighlights, boxes and notes on specific words, figures or tables in a PDFRead + write
Add and organiseNew papers from a DOI, URL, ISBN, BibTeX or file; collections, tags and duplicate mergingRead + write
Semantic searchOptional search by meaning with local, OpenAI, Gemini or Ollama embeddingsRead

Good to know

Ask questions about the papers you already keep in Zotero and get answers that point back to items in your library. The connector can search, read full text and BibTeX, and, once writes are enabled, add highlights, notes and new items.

Limits

  • Zotero must be installed with its local API enabled. Changes to the library need Zotero 10 with a one-time authorization, or a Zotero web API key on older versions.
  • Optional semantic search can use OpenAI or Gemini embeddings, which sends text to those services; local embeddings are also offered.

Notes by (original project name: zotero-mcp), written from the creator's documentation, not from running it ourselves. Zotero MCP README · Zotero MCP: getting started

Stuck setting it up?

Tell us what happened. We take a look first; if it needs paid work, you get an estimate before anything starts.

Get help

Other connectors

  • Airtable MCP Lets your AI agent find, add and update records in your Airtable bases
  • AntV Chart MCP Lets your AI agent turn your numbers into charts, mind maps and flow diagrams
  • Apify MCP Server Lets your AI agent collect data from maps, online shops, search results and social sites
  • Asana MCP Lets your AI assistant look up, create and update tasks and projects in Asana
  • Atlassian MCP Server Lets your AI agent find, summarize and update Jira work items and Confluence pages
  • Attio MCP Lets your AI assistant look up and update contacts, deals and notes in your Attio CRM

See all 52 connectors · An MCP server is missing or disconnected

GitHub