A professional-grade MCP (Model Context Protocol) server that provides cutting-edge prompt optimization tools with research-backed strategies delivering 15-74% performance improvements.
- Clarity: Simplifies prompts for directness and precision
- Specificity: Adds detailed constraints and requirements
- Chain of Thought: Incorporates step-by-step reasoning
- Few-Shot: Includes example formats for guidance
- Structured Output: Defines clear output organization
- Role-Based: Adds expert role context
- Tree of Thoughts (ToT): Multi-path reasoning with 74% success rate on complex tasks
- Constitutional AI: Self-critique and alignment with safety principles
- Automatic Prompt Engineer (APE): AI-discovered optimal instruction patterns
- Meta-Prompting: AI generates its own optimized prompts
- Self-Refine: Iterative improvement with 20% performance gains
- TEXTGRAD: Natural language feedback as optimization gradients
- Medprompt: Multi-technique ensemble achieving 90%+ accuracy
- PromptWizard: Feedback-driven self-evolving prompts
Production-ready templates across 11 domains:
- Business Analysis: Competitive analysis frameworks
- Product Management: User research synthesis
- Content Creation: Technical blog posts with SEO optimization
- Development: Comprehensive code review checklists
- Communication: Stakeholder updates and project reports
- Strategy: OKR planning frameworks
- Operations: Standard Operating Procedures (SOPs)
- Legal: Contract termination and compliance
- Customer Experience: Feedback surveys and insights
- Data Analysis: Data insights and reporting
- Meeting Management: Effective meeting agendas
# Clone the repository
git clone <repository-url>
cd mcp-prompt-optimizer
# Create virtual environment (recommended)
python3 -m venv venv
source venv/bin/activate # On Windows: venv\Scripts\activate
# Install dependencies
./install.sh
# Or install manually
pip install -r requirements.txt
# Configure Claude Desktop
python3 setup_interactive.pyAdd to your Claude Desktop configuration file:
macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json
Linux: ~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"prompt-optimizer": {
"command": "python3",
"args": ["/path/to/mcp-prompt-optimizer/prompt_optimizer.py"],
"env": {}
}
}
}# Analyze prompt quality
"Analyze this prompt: write a blog post about AI"
# Apply specific optimization
"Optimize this prompt using chain_of_thought: explain machine learning"
# Auto-select best strategy
"Auto-optimize: help me debug this code"
# Get domain template
"Get domain template for code_review_checklist"
# Use Tree of Thoughts for complex problems
"Apply advanced optimization with tree_of_thoughts: design a microservices architecture"
# Use Constitutional AI for safety-critical tasks
"Apply advanced optimization with constitutional_ai: create content moderation guidelines"
# Use Medprompt for high-accuracy classification
"Apply advanced optimization with medprompt: categorize customer support tickets"
# List available templates
"List all domain templates"
mcp-prompt-optimizer/
├── prompt_optimizer.py # Main MCP server
├── advanced_strategies.py # Research-backed optimization strategies
├── domain_templates.py # Professional domain templates
├── examples.py # Usage examples and demonstrations
├── setup_interactive.py # Automated setup script
└── README.md # This file
# Run basic tests
./test.sh
# Run usage examples
python3 examples.py| Strategy | Use Case | Performance Improvement |
|---|---|---|
| Tree of Thoughts | Complex reasoning | 70-74% success rate |
| Medprompt | Classification tasks | 90%+ accuracy |
| Self-Refine | Iterative improvement | 20% per iteration |
| Constitutional AI | Safety alignment | High compliance |
| Chain of Thought | Step-by-step tasks | 15-25% improvement |
- analyze_prompt: Analyzes prompt quality and identifies issues
- optimize_prompt: Applies specific optimization strategies
- auto_optimize: Automatically selects optimal strategy
- get_prompt_template: Returns basic templates
- advanced_optimize: Applies research-backed strategies
- get_domain_template: Returns professional domain templates
- list_domain_templates: Lists available templates by domain
| Prompt Type | Recommended Strategy |
|---|---|
| Complex problems | tree_of_thoughts |
| Classification tasks | medprompt |
| Safety-critical | constitutional_ai |
| Vague requirements | meta_prompting |
| Needs refinement | self_refine |
| General optimization | auto |
We welcome contributions! Please:
- Fork the repository
- Create a feature branch
- Add tests for new functionality
- Update documentation
- Submit a pull request
- New Strategy: Add to
advanced_strategies.py - New Template: Add to
domain_templates.py - Examples: Add to
examples.py
MCP not working?
- Check Python version:
python3 --version(requires 3.8+) - Install dependencies: Run
./install.shorpip install -r requirements.txt - Verify MCP installation:
pip show mcp - Check Claude Desktop logs
- Restart Claude Desktop
Commands not recognized?
- Verify configuration file location
- Check file paths in configuration
- Run setup script again
# Test server directly
python3 prompt_optimizer.py
# Verbose logging
export MCP_LOG_LEVEL=debug
python3 prompt_optimizer.pyThis project is licensed under the MIT License - see the LICENSE file for details.
- Research from Princeton, Google DeepMind, Microsoft Research
- Anthropic's Constitutional AI framework
- Stanford's DSPy framework
- OpenAI's prompt engineering guidelines
If you use this tool in your research or projects, please cite:
@software{mcp_prompt_optimizer,
title={MCP Prompt Optimizer: Research-Backed Prompt Optimization for AI Systems},
author={Bubobot},
year={2024},
url={https://github.com/Bubobot-Team/mcp-prompt-optimizer}
}Built with ❤️ for the AI community
For questions, issues, or contributions, please visit our GitHub repository.