LABSCO SUMMARY

Extracting your own annotations is the feature — the library metadata was never the hard part.

Titles and abstracts are searchable anywhere. What is trapped is the highlight you made on page fourteen three years ago, which is precisely the thing you are trying to find again. Pulling annotations out means a question can be answered from your own reading rather than from the paper's abstract. The local option matters for the same reason: the library is on your machine, and it does not need to leave it.

What it is

A connection to Zotero, the reference manager, reaching either the local application or the web API, and going into the documents rather than stopping at the metadata.

What you get

Access through the local Zotero or the web API
PDF annotation extraction — the highlights and notes you left in the papers
Advanced search across the library

Requirements

Python, installed from PyPI. MIT licensed.

Setup effort

One command — uv tool install zotero-mcp-server

See installation options

USE CASES

01

Finding what you already read

Search the library you built, not the whole literature.

zotero_search_itemszotero_advanced_searchzotero_semantic_searchzotero_search_by_tagzotero_get_recent

02

Recovering your own notes

Annotations lifted out of the PDFs.

zotero_get_noteszotero_get_annotationszotero_synthesize_annotations

03

Keeping it local

The local application rather than the web API.

41 tools

Listed in project docs. Your client may show a different list.

zotero_semantic_search

zotero_update_search_database

zotero_get_search_database_status

zotero_search_items

zotero_advanced_search

zotero_get_collections

zotero_get_collection_items

zotero_get_tags

zotero_get_recent

zotero_search_by_tag

zotero_get_item_metadata

zotero_get_item_fulltext

zotero_get_item_children

zotero_get_annotations

zotero_synthesize_annotations

BEFORE YOU INSTALL

What to know first

Tools you get

41 tools named in the project's documentation.

Requires

zotero-mcp on PATH
Python 3.10+

INSTALLATION

Connect your client

Server setup

TERMINAL

uv tool install zotero-mcp-server

MCP Clients

Claude Desktop, Claude Code

JSON

{
  "mcpServers": {
    "zotero": {
      "command": "zotero-mcp",
      "env": {
        "ZOTERO_LOCAL": "true",
        "ZOTERO_API_KEY": "<YOUR_ZOTERO_API_KEY>",
        "ZOTERO_LIBRARY_ID": "<YOUR_ZOTERO_LIBRARY_ID>"
      }
    }
  }
}

Autohand Code

AUTOHAND CODE · TERMINAL

autohand mcp add zotero env ZOTERO_LOCAL=true zotero-mcp

Cherry Studio

CHERRY STUDIO · JSON

{
  "mcpServers": {
    "zotero": {
      "name": "zotero",
      "type": "stdio",
      "isActive": true,
      "command": "zotero-mcp",
      "args": [],
      "env": {
        "ZOTERO_LOCAL": "true"
      }
    }
  }
}

Chorus

Find your installation path: run zotero-mcp setup-info to get the exact path and configuration details

Open Chorus.sh preferences and set Command to the full path to your zotero-mcp installation

Leave Arguments empty

Under Environment (JSON), paste your environment configuration as a single line: {"ZOTERO_LOCAL": "true"}

ChatGPT — Developer Mode

Install ngrok from https://ngrok.com/download

Start the server over streamable HTTP, with your Zotero variables set first (for example ZOTERO_LOCAL=true), and leave this terminal open

CHATGPT — DEVELOPER MODE · TERMINAL

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

In a second terminal, open the tunnel

CHATGPT — DEVELOPER MODE · TERMINAL

ngrok http 8000

Copy the HTTPS Forwarding URL ngrok prints

On chatgpt.com open Settings > Connectors, click Advanced and enable Developer Mode, then click Create

Set MCP Server URL to your ngrok URL followed by /mcp, set Authentication to No authentication, tick I trust this application and click Create

zotero-mcp serve has no authentication of its own: anyone with the tunnel URL can use every tool, so keep the URL private and stop the tunnel when you are not using it

Zotero Research Library