Reproducible bibliographic workflow with AI-queryable knowledge repository

Transform Zotero into a meta-knowledge hub that AI agents can understand and reference.

Problem: AI agents can answer anything, but that doesn't build your knowledge.

You read a paper. An AI summarizes it. Great! But where does that leave you? Disconnected facts. No personal archive. No conversation context.

Solution: Zotero as your queryable knowledge repository

• 📚 Your archive: Not just papers—videos, podcasts, cafes, life moments
• 🤖 AI agents query your library during conversations (via MCP)
• 🔗 Bibliographic notes become shared context between you and agents
• 📖 Reading workflow: capture → query → connect → emerge

Philosophy: Life is a book. Everyone is an author. Build your anthology.

For Humans:

• ✅ Reproducible Zotero setup (NixOS-ready, syncs via Git)
• ✅ Korean Dewey Decimal Classification citation keys (book-pkm2024)
• ✅ Auto-export BibTeX for Org-mode/LaTeX
• ✅ Self-contained workspace (no scattered configs)

For AI Agents:

• ✅ MCP integration: Semantic search over your library
• ✅ PDF annotation queries: "What did I highlight in this paper?"
• ✅ Context-aware conversations: "Find related papers I've read"
• ✅ Knowledge archaeology: "What was I thinking 2 years ago?"

Simple flow:

You → Zotero → MCP Server → AI Agent

1. You capture papers/videos/notes in Zotero
2. Better BibTeX generates citation keys (book-pkm2024)
3. MCP Server indexes your library (semantic search)
4. AI agents query during conversations

Example conversation:

You: "I'm researching knowledge graphs" Agent (queries your Zotero via MCP): "You have 3 papers on this topic. One from 2020 highlights the connection to PKM systems. Want me to reference your annotations?"

• ✅ Korean Dewey Decimal Classification (KDC)

  • Citation key format: book-title2024, blog-title2024, wiki-title, ...
  • Type-based prefixes: book, blog, wiki, film, doc, web, news, person, dict, art
  • Reproducible keys for LaTeX/Org-mode integration

• ✅ Better BibTeX Integration

  • Auto-export to .bib files (tracked in Git)
  • Org-mode Quick Copy: Ctrl+Shift+C → citation key
  • Custom postscript: callnumber, datemodified, dateadded fields

• ✅ MCP (Model Context Protocol) Support

  • Semantic search: concept-based paper discovery
  • PDF annotation extraction and search
  • Full-text indexing for deep queries
  • Integration with Claude Desktop, Cline, etc.

• ✅ Attanger Plugin

  • Auto-organize attachments by item type
  • Source directory monitoring
  • Configurable file type filters

• ✅ Reproducible Setup

  • NixOS/Home-Manager compatible
  • Self-contained workspace structure
  • Plugin XPI files versioned in Git
  • Configuration templates with path variables

zotero-config/
├── plugins/               # Zotero plugins (XPI files)
│   ├── [email protected]
│   └── [email protected]
├── config/                # Better BibTeX configuration
│   └── betterbibtex-preferences-nixos.json
├── docs/                  # Documentation
│   ├── MCP-INTEGRATION.md
│   ├── SETUP-GUIDE.md
│   └── DEWEY-CLASSIFICATION.md
├── scripts/               # Automation scripts
│   ├── setup.sh
│   └── mcp-setup.sh
└── workspace/             # Local workspace (not tracked)
    ├── data/              # Zotero database (.gitignore)
    ├── files/             # Attachments (.gitignore)
    ├── exports/           # Auto-exported BibTeX
    └── incoming/          # Attanger source (.gitignore)

git clone https://github.com/junghan0611/zotero-config.git ~/zotero
cd ~/zotero

NixOS/Home-Manager:

home.packages = with pkgs; [ zotero ];

Ubuntu/Debian:

sudo apt install zotero

# Install 54yyyu/zotero-mcp for AI agent integration
uv tool install "git+https://github.com/54yyyu/zotero-mcp.git"

# Setup with local Zotero database
zotero-mcp setup --local

# Or use Zotero Web API (for cloud sync)
zotero-mcp setup --no-local \
  --api-key YOUR_API_KEY \
  --library-id YOUR_LIBRARY_ID

# Build semantic search database
zotero-mcp update-db --fulltext

# Run setup script (coming soon)
./scripts/setup.sh

# Or manually:
# 1. Install plugins from plugins/
# 2. Import Better BibTeX config from config/
# 3. Set data directory to workspace/data/

Model Context Protocol (MCP) enables AI agents to access external data sources. With Zotero MCP:

• 🔍 Semantic Search: Ask agents to find papers by concept, not just keywords
• 📝 Annotation Access: Query your PDF highlights and notes
• 🔗 Context Awareness: Agents reference your reading history during conversations
• 📊 Meta-Analysis: Generate insights from your entire library

# AI agent queries your Zotero library
"Find papers about knowledge graphs published after 2020"
"What are my annotations in the paper about PKM systems?"
"Connect this concept to books in my Collection C3"

See docs/MCP-INTEGRATION.md for detailed setup.

Before (Traditional):

1. Read paper → manual notes → disconnected knowledge

After (Agent-Integrated):

1. Capture: Save to Zotero with citation key (book-pkm2024)
2. Query: Ask AI agent to find related papers
3. Connect: Agent references your annotations and library
4. Emerge: New insights from conversation context

• Org-mode Integration: Citation keys directly in your notes
• Denote Linking: [[book-title2024]] links to bibliographic entries
• Emacs Workflow: Quick Copy → insert citation → auto-update .bib

AI agents help you rediscover forgotten connections:

• "What did I read about this topic 2 years ago?"
• "Find patterns in my reading history"
• "Which books influenced my thinking on X?"

• Digital Garden: notes.junghanacs.com
• Zotero Meta Note: meta/20240320t110018
• Bib Folder: notes.junghanacs.com/bib/
• Zotero Group Library: @junghanacs

🚧 Work in Progress - Building in public!

Completed:

• Repository structure
• Plugin backup (Better BibTeX v7.0.50, Attanger v1.3.8)
• Configuration templates
• MCP architecture design

In Progress:

• MCP setup scripts
• Documentation (MCP integration, setup guide)
• Korean Dewey Classification guide
• NixOS Home-Manager module

This is a personal configuration, but feel free to:

• Open issues for questions about the workflow
• Fork and adapt for your own bibliographic system
• Share your MCP integration experiences

MIT License - Feel free to use and adapt!

Author: @junghanacs Created: 2025-10-11 Philosophy: 인생은 한 권의 책 (Life is a book)