🚀 Prompt Optimizer MCP
A Model Context Protocol (MCP) server that provides intelligent tools for optimizing and scoring LLM prompts using deterministic heuristics.
🎯 Overview
The Prompt Optimizer MCP server offers two powerful tools:
optimize_prompt - Generate 3 optimized variants of a raw LLM prompt in different stylesscore_prompt - Evaluate the effectiveness of an improved prompt relative to the original
Perfect for developers, content creators, and AI practitioners who want to improve their prompt engineering workflow.
✨ Features
🎨 Prompt Optimization Styles
- Creative: Enhanced with descriptive adjectives and engaging language
- Precise: Concise and focused, removing redundant words
- Fast: Optimized for quick processing with shorter synonyms
📊 Intelligent Scoring Algorithm
The scoring system evaluates prompts based on:
- Length optimization (40%): Prefers shorter, more concise prompts
- Keyword preservation (30%): Maintains important terms from the original
- Clarity improvement (30%): Reduces redundancy and improves structure
🔧 Technical Features
- ✅ Stateless: No external dependencies or state management
- ✅ Deterministic: Same inputs always produce same outputs
- ✅ Error-free: Comprehensive input validation and error handling
- ✅ Fast: Simple heuristics for quick processing
- ✅ Extensible: Easy to add new styles and scoring metrics
- ✅ Dual Transport: Supports both STDIO (MCP) and HTTP (deployment)
📁 Project Structure
prompt-optimizer-mcp/
├── 📄 README.md # This file
├── 📄 server.py # Main MCP server (STDIO transport)
├── 📄 http_server.py # HTTP server for deployment
├── 📄 start.py # Startup script (auto-detects mode)
├── 📄 requirements.txt # Python dependencies
├── 📄 test_server.py # Test script
├── 📄 deploy.py # Deployment script
├── 📄 Dockerfile # Container configuration
├── 📄 .gitignore # Git ignore rules
├── 📁 tools/
│ ├── 📄 __init__.py # Package initialization
│ └── 📄 optimize.py # Core optimization logic
├── 📁 tests/
│ ├── 📄 __init__.py # Test package initialization
│ └── 📄 test_optimize.py # Unit tests
└── 📁 .github/
└── 📁 workflows/
└── 📄 ci.yml # CI/CD pipeline
🚀 Quick Start
1. Clone the Repository
git clone https://github.com/Mahad-007/Prompt-Optimizer-MCP-for-LLMs.git
cd Prompt-Optimizer-MCP-for-LLMs
2. Install Dependencies
pip install -r requirements.txt
3. Run Tests
4. Start the Server
# For local development (STDIO mode)
python server.py
# For deployment (HTTP mode)
python start.py
🛠️ Installation
Prerequisites
- Python 3.11 or higher
- pip package manager
Install Dependencies
# Install from requirements.txt
pip install -r requirements.txt
⚙️ Configuration
For Cursor IDE
Create .cursor/mcp.json:
{
"mcpServers": {
"prompt-optimizer": {
"command": "python",
"args": ["server.py"],
"env": {}
}
}
}
For Other MCP Clients
Configure your MCP client to use:
- Command:
python server.py - Transport: STDIO (default)
📖 Usage Examples
Using the MCP Server
Once configured, you can use the tools through any MCP client:
Optimize a Prompt
# Generate creative variants
variants = optimize_prompt(
raw_prompt="Write a story about a cat",
style="creative"
)
# Returns: [
# "Craft a compelling story about a cat",
# "Imagine you're an expert in this field. Write a story about a cat",
# "Write a story about a cat. in a way that captivates and inspires"
# ]
# Generate precise variants
variants = optimize_prompt(
raw_prompt="Please write a very detailed explanation about machine learning",
style="precise"
)
# Returns: [
# "Write a detailed explanation about machine learning",
# "• Write a detailed explanation about machine learning",
# "Write a detailed explanation about machine learning Be specific and concise."
# ]
Score a Prompt
score = score_prompt(
raw_prompt="Please write a very detailed explanation about machine learning",
improved_prompt="Write an explanation about machine learning"
)
# Returns: 0.85 (high score due to length reduction and clarity improvement)
HTTP API Usage
When deployed, the server also provides HTTP endpoints:
# Health check
curl http://localhost:8000/health
# Optimize prompt
curl -X POST http://localhost:8000/optimize \
-H "Content-Type: application/json" \
-d '{"raw_prompt": "Write about AI", "style": "creative"}'
# Score prompt
curl -X POST http://localhost:8000/score \
-H "Content-Type: application/json" \
-d '{"raw_prompt": "Write about AI", "improved_prompt": "Write about artificial intelligence"}'
Direct Python Usage
from tools.optimize import optimize_prompt, score_prompt
# Optimize a prompt
variants = optimize_prompt("Write about AI", "creative")
print(f"Optimized variants: {variants}")
# Score a prompt
score = score_prompt("Write about AI", "Write about artificial intelligence")
print(f"Score: {score}")
🧪 Testing
Run the comprehensive test suite:
# Run all tests
python test_server.py
# Run unit tests
python -m unittest tests.test_optimize -v
# Run specific test classes
python -m unittest tests.test_optimize.TestOptimizePrompt
python -m unittest tests.test_optimize.TestScorePrompt
python -m unittest tests.test_optimize.TestIntegration
🚀 Deployment
Automated Deployment
Use the deployment script:
This will:
- Run all tests
- Install dependencies
- Run linting checks
- Build Docker image (if available)
- Create deployment package
Manual Deployment
Deploy to Smithery
- Install Smithery CLI:
npm install -g @smithery/cli
- Authenticate:
- Deploy:
# Windows
.\deploy.bat
# Linux/macOS
chmod +x deploy.sh
./deploy.sh
Deploy with Docker
# Build the image
docker build -t prompt-optimizer-mcp:latest .
# Run the container
docker run -p 8000:8000 prompt-optimizer-mcp:latest
The server supports both STDIO (for MCP clients) and HTTP (for web deployment) transports:
- STDIO Mode:
python server.py (for MCP clients) - HTTP Mode:
python start.py (for web deployment)
Your MCP server will be available at: https://prompt-optimizer-mcp.smithery.ai
For detailed deployment instructions, see DEPLOYMENT.md.
🔧 Development
Adding New Optimization Styles
- Add the new style to the
Literal type in server.py - Implement the style function in
tools/optimize.py - Add corresponding tests in
tests/test_optimize.py
Extending the Scoring Algorithm
Modify the score_prompt function in tools/optimize.py to include additional metrics or adjust weights.
Running Locally
# Start the MCP server (STDIO mode)
python server.py
# Start the HTTP server (deployment mode)
python http_server.py
# Auto-detect mode based on environment
python start.py
- Response Time: < 100ms for most operations
- Memory Usage: ~50MB typical
- CPU Usage: Minimal (stateless operations)
- Scalability: Auto-scales from 1-5 replicas on Smithery
🤝 Contributing
We welcome contributions! Please follow these steps:
- Fork the repository
- Create a feature branch (
git checkout -b feature/amazing-feature) - Commit your changes (
git commit -m 'Add amazing feature') - Push to the branch (
git push origin feature/amazing-feature) - Open a Pull Request
Development Setup
# Clone your fork
git clone https://github.com/yourusername/Prompt-Optimizer-MCP-for-LLMs.git
cd Prompt-Optimizer-MCP-for-LLMs
# Install dependencies
pip install -r requirements.txt
# Run tests
python test_server.py
# Make your changes and test
python demo.py
📝 License
This project is licensed under the MIT License - see the LICENSE file for details.
🙏 Acknowledgments
📞 Support
⭐ Star History
Made with ❤️ for the AI community