Paper Zotero

Paper Zotero

Overview

Imports selected papers into Zotero via mcp-zotero. Supports DOI-based import (preferred — auto-fetches full metadata + OA PDF via Unpaywall) and PDF import with fulltext indexing. Auto-creates a named collection for the batch.

Required MCP

mcp-zotero (Xevos117) must be installed and configured with ZOTERO_API_KEY and ZOTERO_USER_ID.

Actual MCP Tool Names

Use these exact tool names when calling mcp-zotero:

| Action | Tool | |--------|------| | List collections | get_collections | | Create collection | create_collection | | Import by DOI | add_items_by_doi | | Import PDF from path/URL | import_pdf_to_zotero | | Search library | search_library | | Get item details | get_items_details |

Input

Accepts paper-fetch output JSON, or ask the user for:

• PDF file paths and/or DOI list
• Target Zotero collection name (default: derive from search topic, e.g. RAG for Code Generation)

Steps

1. Resolve collection

• Call get_collections to check if a collection with the target name already exists
• If not, call create_collection with the topic name → save returned collection_key

1b. Deduplication pre-check (run before any import)

• For each paper with a DOI or arXiv ID, call search_library with the arXiv ID or DOI as the query
• If a result is returned: skip add_items_by_doi for that paper, use the existing zotero_key instead, log as "already in library"
• Only call add_items_by_doi for papers not already in the library

1. Import each paper

• If DOI available → call add_items_by_doi with dois array and collection_key
• This auto-attaches OA PDF via Unpaywall when a direct PDF link is found
• If only local PDF path available → call import_pdf_to_zotero with the file path

1. PDF fallback (run after step 2)

• For each paper where pdf_attached: false in the add_items_by_doi response:
• If paper-fetch downloaded a local PDF for this paper, call import_pdf_to_zotero with that path
• Otherwise log it as "PDF unavailable" — do not block the workflow

1. Report summary

• "Imported N papers to Zotero collection '{{topic}}'"
• List papers with PDF status: ✅ PDF attached / ⚠️ metadata only

Output

{
  "collection_key": "NUZMGKE7",
  "collection_name": "RAG for Code Generation",
  "imported": [
    {
      "title": "Attention Is All You Need",
      "doi": "10.48550/arXiv.1706.03762",
      "zotero_key": "ABC123DE",
      "pdf_attached": true
    }
  ],
  "failed": []
}

Pass zotero_key values to downstream Obsidian sub-skills so notes can link back to Zotero.

Common Issues

• Duplicate detection: Always run the deduplication pre-check in step 1b before importing. If add_items_by_doi still returns a duplicate warning despite the pre-check, skip and log the existing key.
• Missing metadata: If DOI import returns incomplete metadata, inform the user which fields are missing rather than silently leaving them blank.
• PDF quota exceeded: Zotero free storage is 300 MB. If upload fails with a quota error, log the paper title and continue — do not abort the whole batch.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Geek96
• Source: Geek96/paper-research-skill
• License: MIT

Install and usage instructions live in the source repository linked above.