Build Your First Agentic AI App with MCP

Maven Lightning Lesson

A Python project demonstrating the use of OpenAI GPT models and MCP (Model Context Protocol) agents to fetch and process real-world data.

Overview

This project provides an example of using a custom agent to retrieve weather forecasts for Hintertux, Austria, by fetching data from the meteoblue.com website. It leverages the openai-agents library and its capabilities to run and expose MCP Servers.

Features

• Custom agent implementation using OpenAI GPT models
• Integration with MCP server for data fetching
• Example function tool (get_time) for current time retrieval
• Returns structured JSON output

Requirements

• Python 3.13+
• OpenAI API key

Installation

1. Clone the repository:

   git clone ...
   cd mcp-lightning-lesson

2. Create a venv using uv:

   uv venv

3. Copy the example environment file and add your OpenAI API key:

   cp .env.example .env
   # Edit .env and set your OPENAI_API_KEY

Usage

Run the main script:

uv run python main.py

The agent will:

• Fetch the weather forecast for Hintertux, Austria
• Get the current time
• Return the results as a JSON payload

Project Structure

• main.py – Main application logic
• pyproject.toml – Project dependencies and metadata
• .env.example – Example environment variable file

Dependencies

• openai-agents >= 0.0.15
• python-dotenv >= 1.1.0

Frontend

Coming soon!