Weather MCP Server & Databricks Agent

A Model Context Protocol (MCP) server built with FastMCP exposing weather intelligence tools backed by the Open-Meteo API, integrated with Databricks Agent Bricks.

📁 Repository Structure

weather-mcp/
├── src/
│   └── weather_mcp/
│       ├── __init__.py          # Package exports
│       ├── weather_broker.py    # Adapter module & WeatherClient / WeatherRuleEngine
│       └── mcp_server.py        # FastMCP server & tool definitions
├── tests/
│   └── test_weather.py          # Unit & integration test suite
├── AGENT_PROMPT.md              # Databricks Agent system prompt & registration guide
├── CONTEXT.md                   # Domain model glossary
├── README.md                    # Project documentation
├── app.yaml                     # Databricks Apps manifest
├── pyproject.toml               # Python package configuration
└── requirements.txt             # Production dependencies

Related MCP server: MCP Weather Server

🏛️ Architecture Overview

 ┌────────────────────────────────────────────────────────┐
 │            Databricks Agent Bricks Agent               │
 └──────────────────────────┬─────────────────────────────┘
                            │ Model Context Protocol (MCP / HTTP)
                            ▼
 ┌────────────────────────────────────────────────────────┐
 │      FastMCP Server (src/weather_mcp/mcp_server.py)    │
 └──────────────────────────┬─────────────────────────────┘
                            │ Thin wrapper calls
                            ▼
 ┌────────────────────────────────────────────────────────┐
 │     Broker Adapter (src/weather_mcp/weather_broker.py) │
 └──────────┬─────────────────────────────────┬───────────┘
            │ Geocoding API                   │ Forecast API
            ▼                                 ▼
 ┌──────────────────────┐         ┌───────────────────────┐
 │ Open-Meteo Geocoding │         │  Open-Meteo Forecast  │
 └──────────────────────┘         └───────────────────────┘

🛠️ MCP Tools Reference

🚀 Setup & Local Execution

1. Install Dependencies

uv sync

2. Run Test Suite

PYTHONPATH=src uv run python tests/test_weather.py

3. Start FastMCP Server Locally

uv run fastmcp run src/weather_mcp/mcp_server.py --transport http --port 8000

The server will listen at http://127.0.0.1:8000/mcp.

☁️ Databricks Apps Deployment & Agent Bricks Registration

1. Deploy App: Deploy this repository to Databricks Apps using app.yaml:

   command:
     - "fastmcp"
     - "run"
     - "src/weather_mcp/mcp_server.py"
     - "--transport"
     - "http"
     - "--host"
     - "0.0.0.0"
     - "--port"
     - "8000"

2. Register External MCP Tool:

   • In Databricks Workspace -> AI Playground / Agent Bricks.

   • Click Add Tool -> External MCP Server.

   • Enter App URL: https://<databricks-app-url>/mcp.

3. Configure System Prompt:

   • Copy system prompt from AGENT_PROMPT.md.

🧪 Demonstrated Agent Queries & Responses

Query 1: Rain & Umbrella Prediction

User: "Will it rain in Chicago tomorrow?"

• Tool Call: predict_umbrella_needed(location="Chicago", date="tomorrow")

• Agent Response:

  "No, it is very unlikely to rain in Chicago tomorrow (2026-08-09). The precipitation probability is only 2% with clear skies. High: 79.0°F (26.1°C), Low: 64.5°F (18.1°C). You won't need an umbrella!"

Query 2: Travel Packing Advice

User: "Should I bring a jacket to Austin this weekend?"

• Tool Call: get_travel_recommendation(location="Austin, TX", days=3)

• Agent Response:

  "You will not need a heavy jacket in Austin this weekend! Temperatures will be very warm with highs reaching 99.6°F (37.5°C) and lows around 76.1°F (24.5°C). I recommend packing light summer clothing and staying hydrated."

Query 3: Multi-City Weather Comparison

User: "Is it warmer right now in Chicago, Austin, or Tokyo?"

• Tool Call: compare_city_weather(cities="Chicago, Austin, Tokyo")

• Agent Response:

  "Currently, Austin, Texas is the warmest at 96.6°F (35.9°C). Comparison:

  1. Austin, TX: 96.6°F (35.9°C)

  2. Chicago, IL: 79.0°F (26.1°C)

  3. Tokyo, Japan: 73.9°F (23.3°C)"

Query 4: Edge Case Error Handling

User: "What's the weather in NonExistentCityXYZ?"

• Tool Call: get_current_weather(location="NonExistentCityXYZ")

• Agent Response:

  "I couldn't find a location matching 'NonExistentCityXYZ'. Could you please check the spelling or provide the city name alongside its state or country?"