Luc's AI Agent - CrewAI Digital Twin

A comprehensive CrewAI-powered system that simulates Luc's learning process through collaborative AI agents. Features both a NANDA-based HTTP API server and a modular MCP (Model Context Protocol) package for flexible deployment.

🚀 Features

• Dual-Agent System: Research agent + Content mentor agent
• Multiple Deployment Options: NANDA HTTP API + MCP server package
• Topic-Driven Processing: Send a topic string; get back a clear summary
• Sequential Processing: Agents work together in a structured workflow
• Beginner-Friendly Output: Complex topics explained in simple terms
• Environment Configuration: Uses .env for keys and domain config
• Modular Architecture: Clean package structure with proper dependencies

📁 Project Structure

AI Agent/
├── Assignment_1/
│   ├── main.py                 # NANDA-based HTTP API server
│   ├── mentor_mcp/            # MCP server package
│   │   ├── __init__.py        # Package initialization
│   │   ├── main.py            # MCP server implementation
│   │   └── __main__.py        # Module entry point
│   ├── pyproject.toml         # Package configuration
│   ├── requirements.txt       # Python dependencies
│   ├── .gitignore            # Git ignore patterns
│   ├── simclr_explained.md   # Example output file
│   ├── ai_agent_explained.md # Example output file
│   ├── test_local.py         # Local testing script
│   ├── run_adapter.py        # NANDA adapter runner
│   └── README.md          

🛠️ Setup

Prerequisites

• Python 3.8 or higher
• Anthropic API key (used by the NANDA wrapper)

Installation

1. Navigate to the project directory:

   cd "/Users/lucchen/Desktop/AI Agent"
   

2. Create and activate a virtual environment:

   python3 -m venv .venv
   source .venv/bin/activate  # On Windows: .venv\Scripts\activate
   

3. Install dependencies and package:

   cd Assignment_1
   pip install -r requirements.txt
   pip install -e .  # Install mentor_mcp package in development mode
   

4. Set up environment variables:

   Create a .env file in the Assignment_1 directory with at least your Anthropic key. Optionally set a domain if you have one:

   cd Assignment_1
   echo "ANTHROPIC_API_KEY=your-anthropic-key-here" > .env
   echo "DOMAIN_NAME=your-domain.example" >> .env  # optional
   

   Or set them as environment variables in your shell:

   export ANTHROPIC_API_KEY="your-anthropic-key-here"
   export DOMAIN_NAME="your-domain.example"   # optional
   

🎯 Usage

Option 1: NANDA HTTP API Server

1. Start the NANDA server:

   python main.py
   

2. Send a message (topic) to the server:

   • Provide any technical topic string (e.g., "Machine Learning", "Neural Networks", "SimCLR").
   • If an empty message is provided, the default topic is "SimCLR".
   • The server returns the final explanation text generated by the crew.

3. Server details:

   • The NANDA adapter starts an HTTP API and handles routing
   • Check the terminal logs for the exposed URL and usage details

Option 2: MCP Server Package

1. Run as a module:

   python -m mentor_mcp
   

2. Run using the CLI command:

   mentor_mcp
   

3. Use in your code:

   from mentor_mcp import crew_agent_logic
   
   result = crew_agent_logic("Machine Learning")
   print(result)
   

Example Flow

# NANDA Server
$ python main.py
# NANDA starts and prints server information / public URL (if configured)
# Use that endpoint to POST a message containing your topic string

# MCP Server
$ python -m mentor_mcp
# MCP server starts and listens for stdio transport

Output

The server returns a text response with:

• Clear definitions
• Key concepts explained simply
• How things work (step-by-step)
• Real-world examples
• Why it's important

🔧 Configuration

Customizing Agents

You can modify the agents in either main.py (NANDA version) or mentor_mcp/main.py (MCP version):

• Research Agent: Change the role, goal, or backstory
• Mentor Agent: Adjust the simplification approach
• LLM Settings: Specify different models or temperature settings

Example customization:

def create_student_agent():
    return Agent(
        role="Harvard Data Science Student",
        goal="Summarize technical concepts clearly and accurately",
        backstory="""You are Luc, a Harvard Data Science student...""",
        verbose=True,
        allow_delegation=False,
        model="gpt-4o-mini",  # Specify model
        temperature=0.2       # Control randomness
    )

Package Development

To modify the MCP package:

1. Edit the package code in mentor_mcp/
2. Reinstall in development mode: pip install -e .
3. Test changes using any of the run methods above

🐛 Troubleshooting

Common Issues

1. "No module named 'crewai'"

   • Make sure you're in the virtual environment
   • Run pip install -r requirements.txt

2. API Key Error

   • Verify ANTHROPIC_API_KEY is set correctly in your .env or shell
   • Check that you have sufficient API credits
   • Ensure the .env file is in the Assignment_1 directory

3. Server Not Reachable

   • Confirm the server is running in the terminal
   • If using a custom DOMAIN_NAME, verify DNS/routing is correctly configured
   • Check local firewall or network restrictions

4. Import Errors

   • Make sure all dependencies are installed
   • Check that you're running from the correct directory

Performance Tips

• Use gpt-4o-mini for faster, cheaper runs
• Set temperature=0.2 for more consistent outputs
• Add specific output length requirements in task descriptions
• Use virtual environments to avoid dependency conflicts

📚 Dependencies

Core Dependencies

• crewai: Multi-agent orchestration framework
• python-dotenv: Environment variable management
• nanda-adapter: Wrapper that exposes Crew logic via an HTTP API
• requests: Utility HTTP client for integrations/tests

Package Dependencies (pyproject.toml)

• crewai>=0.30.0: Multi-agent orchestration framework
• mcp>=1.0.0: Model Context Protocol server framework
• requests: HTTP client library
• nanda-adapter: NANDA integration adapter

🎓 How It Works

NANDA HTTP API Flow

1. Research Agent: Acts as a Harvard Data Science student, researching and summarizing technical topics
2. Mentor Agent: Takes the research and rewrites it in beginner-friendly language
3. Sequential Process: Agents work one after another, with the mentor building on the research
4. HTTP Response: The final explanation is returned as text via the NANDA server API

MCP Server Flow

1. Same Agent Logic: Uses identical CrewAI agents and workflow
2. MCP Protocol: Exposes the functionality via Model Context Protocol
3. Tool Interface: Provides explain_topic tool for MCP clients
4. Stdio Transport: Communicates via standard input/output for MCP integration

🤝 Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

📄 License

This project is part of the MIT AI Studio - CrewAI Tech Track.

🆘 Support

If you encounter issues:

1. Check the troubleshooting section above
2. Verify your API keys and dependencies
3. Ensure you're running from the correct directory
4. Check the CrewAI documentation for advanced configuration

📝 Example Output

When you send a topic like "SimCLR" to either server, the response contains a clear explanation with:

• Definition: What the concept is
• Key Concepts: Core ideas explained simply
• How It Works: Step-by-step breakdown
• Example Applications: Real-world use cases
• Why It Matters: Importance and relevance

🧪 Testing

Local Testing

Use the included test_local.py script to test the CrewAI logic directly:

python test_local.py

NANDA Adapter Testing

Use run_adapter.py for testing the NANDA integration:

python run_adapter.py

🚀 Deployment Options

1. NANDA HTTP API: Perfect for web applications and external integrations
2. MCP Server: Ideal for AI tool integrations and MCP-compatible clients
3. Direct Import: Use crew_agent_logic function in your own Python applications

---

Happy Learning with Luc's AI Agents! 🚀✨