MCP Gradio Multi-Tool Assistant

A Gradio web interface for interacting with a Model Control Protocol (MCP) server. This application uses Claude 3.5 Sonnet to process user queries and invoke specialized tools through the MCP.

Features

• Image Generation: Generate images from text descriptions using multiple fallback options:

  • Hugging Face Inference API
  • Replicate API
  • SanaSprint Gradio model

• Weather Information: Get current weather data for any city using OpenWeatherMap API

• Text Translation: Translate text between languages (currently using mock implementation)

Setup

1. Clone this repository

   git clone <repository-url>
   cd mcp-gradio-poc
   

   Copy

2. Install dependencies

   pip install -r requirements.txt
   

   Copy

3. Create a .env file in the project root with the following variables:

   # Required
   ANTHROPIC_API_KEY=your_anthropic_api_key
   
   # Required for weather tool
   OPENWEATHER_API_KEY=your_openweather_api_key
   
   # Optional for image generation (need at least one)
   HF_API_TOKEN=your_huggingface_token
   REPLICATE_API_TOKEN=your_replicate_token
   

   Copy

4. Run the application

   python app.py
   

   Copy

Usage

1. Launch the application with python app.py
2. Connect to the MCP server by clicking "Connect" (using the default path "gradio_mcp_server.py")
3. Start chatting with the assistant to use the available tools:
   • For images: "Generate an image of mountains at sunset"
   • For weather: "What's the weather in Tokyo?"
   • For translation: "Translate 'hello' to French"

Project Structure

• app.py: Main Gradio interface application
• gradio_mcp_server.py: MCP server implementation with tool definitions
• requirements.txt: Dependencies
• .env: Environment variables and API keys

Adding New Tools

To add new tools to the MCP server, add a new function with the @mcp.tool() decorator in gradio_mcp_server.py. The function should:

1. Use async/await syntax
2. Have typed parameters with docstrings
3. Return a JSON string with properly formatted results

Example:

@mcp.tool()
async def my_new_tool(param1: str, param2: int = 42) -> str:
    """Tool description here.
    
    Args:
        param1: Description of parameter 1
        param2: Description of parameter 2 (default: 42)
    """
    result = {}
    # Tool logic here
    return json.dumps(result)

Troubleshooting

• If image generation fails, make sure you have at least one of the API keys set up
• For weather data, an OpenWeatherMap API key is required
• For any API errors, check the console output for detailed error messages