pyzotero

Trigger phrases

Phrases that activate this skill when typed to Claude Code:

• access my Zotero library
• export citations from Zotero
• search my Zotero collection
• add items to Zotero
• upload PDF to Zotero

What it does

pyzotero is a Claude Code skill from K-Dense AI’s scientific-agent-skills repo. It turns Claude into a Zotero API client that can read and write your Zotero reference library programmatically — retrieving items, searching by tag or collection, exporting formatted citations, uploading PDF attachments, and building automation workflows that integrate with Zotero’s cloud sync.

A session produces working Python code that talks to the Zotero Web API v3 using the pyzotero client, along with the actual data retrieved or modified in your library.

When to use it

Reach for it when:

• You want to export a filtered subset of your Zotero library (e.g., all items tagged “to-read” or in a specific collection) as a BibTeX file
• You’re building a research automation workflow that needs to write new items or attachments back to Zotero
• You need to batch-update tags, collections, or metadata across many library items

When not to reach for it:

• Simple citation formatting for a manuscript — use citation-management
• You don’t have a Zotero library or API key configured

Install

Copy the SKILL.md from K-Dense AI’s pyzotero folder into .claude/skills/pyzotero/ in your project. You’ll need a Zotero API key from zotero.org/settings/keys and your library ID.

Trigger phrases: “access my Zotero library”, “export citations from Zotero”, “search my Zotero collection”.

What a session looks like

A typical session has three phases:

1. Authentication setup. Claude checks for your Zotero API key and library ID, either from environment variables or prompts you to provide them. It confirms API access with a test call.
2. Operation scoping. You describe what you want — search, export, update, or upload. Claude maps the request to the appropriate pyzotero methods and pagination strategy for large libraries.
3. Code execution and output. Claude runs the pyzotero calls, handles rate limits, and returns the results: exported BibTeX, a list of matched items, or confirmation of writes.

Receipts

Where it works well:

• Exporting a collection to BibTeX for a LaTeX project — pyzotero handles pagination cleanly for large libraries
• Batch tagging items based on metadata criteria (year range, author, journal)

Where it backfires:

• PDF full-text extraction is not part of pyzotero — it only manages metadata and attachment links, not file content
• Rate limits on the Zotero API can slow batch operations on libraries with thousands of items

Pattern that works: always check collection IDs and item keys before running write operations; read-then-write with a dry-run confirmation step prevents unintended library modifications.

Source and attribution

Originally authored by K-Dense Inc.. The canonical SKILL.md lives in the pyzotero folder of their public scientific-agent-skills repository.

License: MIT. Install, adapt, and redistribute with attribution preserved.

This page documents the skill from a practitioner’s perspective. For the formal spec and any updates, defer to the source repo.