Zotero MCP Server for Poke

An MCP server that gives Poke access to your Zotero research library. Built with FastMCP following the InteractionCo template.

Tools

Tool	Description
search_items	Search the library by keyword
get_recent_items	Get recently added items
search_by_tag	Filter items by tag
get_tags	List all tags in the library
get_item_metadata	Get full metadata for an item by key
get_item_fulltext	Extract full text from a paper (local mode)
get_item_children	List attachments and notes for an item
get_item_notes	Get note content for an item
get_item_annotations	Get PDF highlights and comments (local mode)
get_collections	List all collections
get_collection_items	Get items in a specific collection
create_note	Create a new note, optionally attached to an item
add_item_by_doi	Add a paper to Zotero by DOI
get_library_info	Show connection status and library details

Setup

1. Copy the example env file and fill it in:

cp .env.example .env

You need:

• Library ID — the number in your Zotero profile URL, or at zotero.org/settings/keys
• API key — create one at zotero.org/settings/keys/new with read (and optionally write) access

2. Install dependencies:

/opt/homebrew/bin/python3.11 -m venv .venv
.venv/bin/pip install -r requirements.txt

3. Run locally:

.venv/bin/python src/server.py

The server starts at http://localhost:8000/mcp.

Local Mode (full-text + annotations)

Set ZOTERO_LOCAL=true to connect directly to the Zotero app running on your machine. This enables:

• get_item_fulltext — extract text from PDFs
• get_item_annotations — read PDF highlights and comments

In Zotero: Edit → Preferences → Advanced → Allow other applications to access Zotero

Connecting Poke

Add this to your Poke MCP configuration:

{
  "mcpServers": {
    "zotero": {
      "url": "http://localhost:8000/mcp"
    }
  }
}

Or if deployed to Render, replace localhost:8000 with your Render URL.

Deploy to Render

Push this repo to GitHub, connect it to Render, and set the environment variables (ZOTERO_LIBRARY_ID, ZOTERO_API_KEY) in the Render dashboard. The render.yaml handles everything else.