MCP Weather & Web Search Agent

A Model Context Protocol (MCP) server that provides weather information and web search capabilities, designed to work with Hugging Face's tiny-agents framework.

Features

• Weather Service: Get weather information for any location
• Web Search: Search the web for information (with special support for Hugging Face inference providers)
• AI Agent Integration: Works seamlessly with tiny-agents for conversational AI
• MCP Inspector Support: Debug and inspect server capabilities

Prerequisites

• Python 3.10 or higher
• uv package manager
• Node.js (for MCP inspector)
• Hugging Face account (for tiny-agents)

Installation

1. Clone or download this project

   git clone https://github.com/Deon62/mcp.git
   cd mcps
   

2. Install Python dependencies

   uv pip install mcp[cli] requests
   

3. Install tiny-agents (if not already installed)

   pip install tiny-agents
   

Quick Start

1. Run the MCP Server

Start the MCP server in one terminal:

uv run --with mcp mcp run server.py

The server will start and wait for connections.

2. Run the AI Agent

In another terminal, start the agent:

tiny-agents run agent.json

You should see:

Agent loaded with 3 tools:
 • get_weather
 • web_search
 • get_hf_inference_providers
»

3. Chat with the Agent

Once the agent is running, you can interact with it:

» Hello! Can you help me find information about Hugging Face inference providers?

Available Tools

1. Weather Service

» What's the weather like in New York?

2. Web Search

» Search for "Hugging Face inference providers"

3. HF Inference Providers

» Get me the list of Hugging Face inference providers

Configuration

Agent Configuration (agent.json)

{
    "model": "Qwen/Qwen2.5-72B-Instruct",
    "provider": "nebius",
    "servers": [
        {
            "type": "stdio",
            "command": "uv",
            "args": ["run", "--with", "mcp", "mcp", "run", "server.py"]
        }
    ]
}

Server Configuration (server.py)

The server provides three main tools:

• get_weather(location) - Returns weather information
• web_search(query) - Performs web searches
• get_hf_inference_providers() - Returns comprehensive list of HF inference providers

MCP Inspector Setup

The MCP Inspector allows you to debug and test your MCP server directly.

1. Install MCP Inspector

npm install -g @modelcontextprotocol/inspector

2. Run the Inspector

mcp-inspector

3. Connect to Your Server

In the inspector:

1. Click "Add Server"
2. Choose "stdio" transport
3. Set command: uv
4. Set args: ["run", "--with", "mcp", "mcp", "run", "server.py"]
5. Click "Connect"

4. Test Tools

Once connected, you can:

• View available tools in the sidebar
• Test each tool with different parameters
• See the JSON-RPC communication
• Debug any issues

Example Usage

Weather Queries

» What's the weather in Tokyo?
» Get weather for London
» How's the weather in San Francisco?

Web Search Queries

» Search for "latest AI developments"
» Find information about "MCP protocol"
» Look up "Hugging Face inference providers"

Specific HF Provider Queries

» Show me all Hugging Face inference providers
» What inference providers does HF support?
» List the available HF deployment options

Troubleshooting

Common Issues

1. "ModuleNotFoundError: No module named 'mcp'"

   uv pip install mcp[cli]
   

2. "KeyError: 'command'"

   • Check your agent.json configuration
   • Ensure the server configuration is correct

3. "Connection closed" errors

   • Make sure the MCP server is running
   • Check that all dependencies are installed

4. Agent shows "0 tools"

   • Verify the server is running
   • Check the agent.json configuration
   • Ensure the server command is correct

Debug Steps

1. Test the server directly:

   python server.py
   

2. Check MCP server with inspector:

   mcp-inspector
   

3. Verify dependencies:

   uv pip list | grep mcp
   

Project Structure

mcps/
├── server.py          # MCP server implementation
├── agent.json         # Agent configuration
├── requirements.txt   # Python dependencies
├── uv.lock           # Dependency lock file
└── README.md         # This file

Development

Adding New Tools

To add a new tool to the server:

@mcp.tool()
def your_new_tool(param: str) -> str:
    """Description of what this tool does"""
    return f"Result for {param}"

Modifying Agent Configuration

Edit agent.json to:

• Change the AI model
• Add more MCP servers
• Modify server configurations

Hugging Face Inference Providers

The server includes comprehensive information about HF inference providers:

1. Amazon SageMaker - Serverless inference with custom Inferentia2 chips
2. Novita AI - Integrated serverless inference directly on model pages
3. Together AI - Serverless inference with competitive pricing
4. Nscale - Official HF provider with high-performance GPU clusters
5. Inference Endpoints - Dedicated, fully managed infrastructure
6. Google Cloud - Vertex AI and other deployment options
7. Microsoft Azure - Azure Machine Learning services
8. Replicate - Easy-to-use model deployment platform
9. Banana - Serverless GPU inference platform
10. Modal - Serverless compute platform
11. RunPod - GPU cloud computing
12. Lambda Labs - GPU cloud infrastructure

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test with both the agent and inspector
5. Submit a pull request

License

[Add your license information here]

Support

For issues and questions:

1. Check the troubleshooting section
2. Use the MCP inspector to debug
3. Open an issue on GitHub
4. Check the MCP documentation

---

**Happy coding! **