🤖 Promptbook MCP

Your personal cookbook for AI prompts with RAG-powered semantic search

✨ What is this?

Promptbook MCP is a plug-and-play server that helps developers who use AI coding assistants (like GitHub Copilot, Claude, etc.) to:

• 📚 Store prompts from your AI sessions automatically
• 🔍 Search prompts by meaning, not just keywords (RAG-powered)
• 🤖 Access your prompt library from any MCP-compatible tool
• 📊 Organize prompts by category (refactoring, testing, debugging, etc.)

Perfect for: Developers who reuse AI prompts and want a searchable knowledge base.

---

🚀 Quick Start

Get running in 30 seconds:

Option 1: Automated Setup (Recommended)

git clone https://github.com/isaacpalomero/promptbook-mcp.git
cd promptbook-mcp
./setup.sh

That's it! 🎉

Option 2: Docker

git clone https://github.com/isaacpalomero/promptbook-mcp.git
cd promptbook-mcp
docker-compose up -d

Done! Your server is running.

---

💡 Use Cases

Problem: You asked ChatGPT/Copilot the perfect prompt for refactoring last week. Now you can't find it.

Solution: Promptbook MCP auto-saves and indexes all your prompts.

# Later, search by meaning
search_prompts("refactor typescript to use dependency injection")
→ Returns your exact prompt from last week

Real Examples

1. Refactoring patterns - Store your best "clean code" prompts
2. Testing strategies - Find that perfect test structure prompt
3. Debugging workflows - Access proven debugging prompts
4. Code review - Reuse comprehensive review prompts

---

📦 Installation

Prerequisites

• Python 3.9+ OR Docker
• 2GB RAM minimum
• macOS, Linux, or Windows

Detailed Setup

Automated Setup (Recommended)

# Clone repository
git clone https://github.com/isaacpalomero/promptbook-mcp.git
cd promptbook-mcp

# Run setup script
chmod +x setup.sh
./setup.sh

# Activate virtual environment
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Start server
python mcp_server.py

Docker Method

# Clone repository
git clone https://github.com/isaacpalomero/promptbook-mcp.git
cd promptbook-mcp

# Copy environment file
cp .env.example .env

# Start services
docker-compose up -d

# Verify
docker-compose logs

Manual Setup

# Create virtual environment
python3 -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Create directories
mkdir -p prompts sessions

# Index existing prompts (if any)
python prompt_rag.py --index

# Start server
python mcp_server.py

---

🎯 Features

1. Semantic Search (RAG)

Find prompts by meaning, not exact words:

search_prompts("how to add unit tests")
→ Finds prompts about "testing", "jest", "pytest", etc.

2. Auto-Organization

Drop AI session files → Auto-categorized and indexed:

sessions/
└── copilot-session-abc123.md  → Auto-processed into:
    ├── prompts/refactoring/prompt1.md
    ├── prompts/testing/prompt2.md
    └── Updated RAG index

3. Multi-Provider Embeddings

Choose your embedding backend:

• Sentence-Transformers (default, local, CPU)
• LMStudio (GPU-accelerated, better quality)

# Use local embeddings (default)
EMBEDDING_PROVIDER=sentence-transformer

# Or use LMStudio
EMBEDDING_PROVIDER=lmstudio
LMSTUDIO_URL=http://localhost:1234

4. MCP Tools (13 Available)

Access via any MCP client:

Tool	Description
search_prompts	Semantic search by meaning
create_prompt	Add new prompt directly
update_prompt	Modify existing prompt
delete_prompt	Remove prompt safely
get_prompt_by_file	Get full content
list_prompts_by_category	Browse by category
find_similar_prompts	Find related prompts
get_library_stats	View statistics
index_prompts	Rebuild search index
organize_session	Process AI session file
get_prompt_index	View full metadata index

Available categories:

• refactoring
• testing
• debugging
• implementation
• documentation
• code-review
• general

---

🔌 MCP Client Setup

Claude Desktop

1. Open Claude config file:

   # macOS
   ~/Library/Application Support/Claude/claude_desktop_config.json
   
   # Windows
   %APPDATA%\Claude\claude_desktop_config.json
   

2. Add Promptbook MCP server:

   {
     "mcpServers": {
       "promptbook": {
         "command": "python",
         "args": ["/path/to/promptbook-mcp/mcp_server.py"]
       }
     }
   }
   

3. Restart Claude Desktop

Other MCP Clients

Any MCP-compatible client can connect using the same pattern. See MCP Protocol docs for details.

---

📖 Documentation

• Setup Guide - Detailed installation steps
• Deployment Options - Docker, local, and production setups
• Embeddings Guide - Configure RAG providers
• Contributing - How to contribute
• Changelog - Version history

---

👨‍💻 For Developers

⚙️ Configuration

All runtime settings are centralized in config.py and exposed through an immutable Config dataclass. The server loads environment variables once at startup.

Environment Variables

Variable	Description	Default
PROMPTS_DIR	Root folder for categorized prompts	./prompts
SESSIONS_DIR	Directory watched for exported sessions	./sessions
VECTOR_DB_DIR	Persistent ChromaDB path	<PROMPTS_DIR>/.vectordb
EMBEDDING_PROVIDER	sentence-transformer or lmstudio	sentence-transformer
EMBEDDING_MODEL	Sentence Transformers model name	all-MiniLM-L6-v2
LMSTUDIO_URL / LMSTUDIO_MODEL	LMStudio endpoint + model	http://localhost:1234 / nomic-embed-text
LMSTUDIO_DIMENSION	Expected LMStudio embedding size	768
CHUNK_SIZE / CHUNK_OVERLAP	Prompt chunking parameters	500 / 100
ENABLE_RAG	Toggle RAG initialization	true
AUTO_REINDEX_INTERVAL	Seconds between auto-index checks	30
LOG_LEVEL	Python logging level	INFO

Configuration file:

# Copy example
cp .env.example .env

# Edit settings
vim .env

Access config anywhere in code:

from config import CONFIG

print(CONFIG.prompts_dir)  # Validated Path object
print(CONFIG.embedding_provider)  # Type-safe enum

---

🧪 Testing & Quality

We enforce strict quality gates:

# Run all tests
pytest

# Run with coverage
pytest --cov=. --cov-report=html

# Run only unit tests
pytest tests/unit/

# Run only integration tests
pytest tests/integration/

# Style check
flake8 --max-line-length=100

# Type check
mypy --strict mcp_server.py prompt_rag.py prompt_organizer.py

Quality Standards

• Test Coverage: Minimum 80%
• Type Safety: mypy --strict must pass
• Code Style: Flake8 compliant
• CI Pipeline: All checks run on Python 3.9-3.12

A GitHub Actions workflow (.github/workflows/ci.yml) runs these checks automatically.

---

🐳 Docker Advanced

Multi-Stage Build

The Dockerfile uses a multi-stage build for optimized image size:

# Stage 1: Builder (installs dependencies)
FROM python:3.11-slim as builder

# Stage 2: Runtime (slim final image)
FROM python:3.11-slim
COPY --from=builder /app/.venv /app/.venv

Result: Final image < 800 MB

Health Checks

Docker includes automatic health monitoring:

HEALTHCHECK --interval=30s --timeout=10s --start-period=5s --retries=3 \
  CMD python -m prompt_rag --health || exit 1

Volume Mounts

Persist data outside containers:

volumes:
  - ./prompts:/app/prompts        # Prompt storage
  - ./sessions:/app/sessions      # Session import
  - ./prompts/.vectordb:/app/prompts/.vectordb  # RAG database

Building & Running

# Build image
docker build -t promptbook-mcp:latest .

# Run container
docker run --rm -i \
  -v "$(pwd)/prompts:/app/prompts" \
  -v "$(pwd)/sessions:/app/sessions" \
  -e EMBEDDING_PROVIDER=sentence-transformer \
  promptbook-mcp:latest

---

🏗️ Architecture

Components

┌─────────────────────────────────────┐
│         MCP Client (Claude)         │
└──────────────┬──────────────────────┘
               │ MCP Protocol
┌──────────────▼──────────────────────┐
│         mcp_server.py               │
│  - 13 MCP tools                     │
│  - Request routing                  │
│  - Error handling                   │
└──────────────┬──────────────────────┘
               │
       ┌───────┴────────┐
       │                │
┌──────▼──────┐  ┌─────▼────────┐
│prompt_rag.py│  │prompt_org.py │
│- RAG search │  │- Session     │
│- Embeddings │  │  parsing     │
│- ChromaDB   │  │- Auto-org    │
└─────────────┘  └──────────────┘
       │                │
       └───────┬────────┘
               │
┌──────────────▼──────────────────────┐
│       prompts/                      │
│       ├── refactoring/              │
│       ├── testing/                  │
│       ├── debugging/                │
│       └── .vectordb/                │
└─────────────────────────────────────┘

Data Flow

1. User asks Claude to search prompts
2. Claude sends MCP request to mcp_server.py
3. Server calls prompt_rag.py for semantic search
4. RAG queries ChromaDB vector database
5. Results returned to Claude with metadata
6. User sees relevant prompts instantly

---

🤝 Contributing

We love contributions! 🎉

Quick Start

1. Fork the repository
2. Create feature branch: git checkout -b feature/amazing-feature
3. Make changes and add tests
4. Run tests: pytest
5. Ensure quality: flake8 && mypy --strict
6. Commit: git commit -m 'feat: add amazing feature'
7. Push: git push origin feature/amazing-feature
8. Open Pull Request

See CONTRIBUTING.md for detailed guidelines.

Commit Convention

We follow Conventional Commits:

• feat: New feature
• fix: Bug fix
• docs: Documentation only
• style: Code style changes
• refactor: Code refactoring
• test: Test changes
• chore: Build/tooling changes

---

📄 License

This project is licensed under the MIT License - see LICENSE file for details.

---

🙏 Acknowledgments

• Built with MCP Protocol
• Powered by ChromaDB and Sentence-Transformers
• Inspired by the need for better prompt management in AI-assisted development

---

📞 Support

• 🐛 Report bugs
• 💡 Request features
• 💬 Discussions
• 📖 Documentation

---

Made with ❤️ for the AI development community