mcp-tutorial

# MCP Tutorial: Weather Server Example A practical demonstration of the Model Context Protocol (MCP) using a simple weather server. ## Overview This tutorial demonstrates the core concepts of MCP: - **Tools**: Model-controlled functions (get_weather) - **Resources**: Application-controlled data (all_cities) - **Prompts**: User-controlled templates (weather_report, packing_list) ## Files - `weather_server.py`: A simple MCP server that provides weather data - `mcp_utils.py`: Shared utilities, including the MCPClient class - `mcp_client.py`: A client that connects to the server and demonstrates basic interactions - `weather_agent.py`: An agent that uses the weather server to suggest activities - `weather_prompt.py`: A demo focusing on user-controlled prompts ## Getting Started 1. Clone this repository 2. Create a virtual environment: `python -m venv venv` 3. Activate the environment: - Windows: `venv\Scripts\activate` - Mac/Linux: `source venv/bin/activate` 4. Install dependencies: `pip install anthropic mcp requests` ## Running the Examples ### Basic MCP Client (`mcp_client.py`) This example demonstrates all three core MCP components: tools, resources, and prompts. ```bash python mcp_client.py ``` **What you'll see:** - List of available tools, resources, and prompts from the server - Retrieving all cities (resource) - Getting weather data for different cities (tool) - Generating a weather report and packing list (prompts) ### Weather Activity Agent (`weather_agent.py`) This example shows how an agent can use MCP tools to provide intelligent recommendations. ```bash python weather_agent.py ``` **What you'll see:** - Interactive prompt asking for a city name - Agent retrieving weather data using the MCP tool - Agent suggesting activities based on weather conditions - Type 'quit' to exit ### User-Controlled Prompts Demo (`weather_prompt.py`) This example focuses on the "prompts" aspect of MCP, showing how users can directly invoke predefined prompt templates. ```bash python weather_prompt.py ``` **What you'll see:** - List of available prompt commands (like slash commands) - Interactive interface where you can type commands like: - `/weather_report tokyo` - Generate a weather report for Tokyo - `/packing_list london 3` - Generate a packing list for 3 days in London - Type 'quit' to exit ## What You'll Learn ### From `mcp_client.py`: - Basic MCP protocol interactions - How to discover and call tools - How to retrieve resources - How to list and invoke prompts ### From `weather_agent.py`: - How to build an agent that uses MCP tools - How tools can provide data for intelligent processing - How to create interactive agents with MCP ### From `weather_prompt.py`: - How prompts work as user-controlled templates - How to implement slash commands with MCP prompts - How prompts differ from tools (user vs. model control) ## MCP Protocol Details Each file demonstrates a different aspect of the MCP protocol: 1. **Tools** - Model-controlled functions that the AI can invoke when needed - Example: `get_weather` tool that retrieves weather data for a city 2. **Resources** - Application-controlled data exposed to the AI - Example: `all_cities` resource that provides a list of available cities 3. **Prompts** - User-controlled templates for common interactions - Examples: `weather_report` and `packing_list` prompts that generate formatted output ## Learn More For more information about MCP, visit [Anthropic's MCP documentation](https://modelcontextprotocol.io/introduction).