Weather Prediction MCP Server & Databricks Agent

An MCP server deployed on Databricks Apps backed by the Open-Meteo REST API, integrated with Databricks Agent Bricks to answer natural language weather queries.

File Structure

• weather_adapter.py: HTTP request and geocoding adapter module.

• weather_mcp_server.py: FastMCP tool definitions (get_current_weather, get_forecast, predict_umbrella_needed).

• app.yaml: Databricks Apps deployment runtime config.

• requirements.txt: Python package requirements.

• .gitignore: Git exclusion file.

Related MCP server: mcp-meteo

Setup & Deployment Instructions

1. Git Folder Setup:

   • Push this directory to your GitHub repository.

   • In Databricks, navigate to Workspace > Users > <your_user>, click Create > Git Folder, and clone your repository URL.

2. Deploy Databricks App:

   • In Databricks sidebar, go to Compute > Apps > Create App.

   • Name the app mcp-weather-server (must start with mcp-).

   • Set the source code path to your cloned Git folder location.

   • Deploy the app and copy its runtime URL.

3. Register MCP Server & Connect Agent Bricks:

   • In Workspace Settings / External Connections, register the app SSE URL (https://<app-url>/sse).

   • In Agent Bricks, create a new agent, attach WeatherPredictionServer as a tool, and paste the system prompt below.

Agent System Prompt

You are WeatherBot, an expert weather advisor assistant deployed on Databricks.

GUIDELINES & GUARDRAILS:
1. Rely EXCLUSIVELY on information returned by your weather tools. Never invent weather metrics.
2. For real-time queries, call `get_current_weather`.
3. For multi-day queries, call `get_forecast`.
4. For recommendation queries, call `predict_umbrella_needed`.
5. If a tool returns an error status, notify the user clearly and request location clarification.

Proof of Functionality (Sample Agent Execution Traces)

Query 1: Current Weather

• Prompt: "What is the current weather in Chicago right now?"

• Tool Called: get_current_weather(location="Chicago")

• Response: "Currently in Chicago, United States, it is 22.5°C (feels like 22.1°C) with 58% humidity and no precipitation."

Query 2: Multi-day Forecast

• Prompt: "Give me a 3-day forecast for Austin, TX."

• Tool Called: get_forecast(location="Austin, TX", days=3)

• Response: "3-Day Forecast for Austin, TX:\n- Aug 8: High 36.1°C, Low 24.2°C (10% rain chance)\n- Aug 9: High 37.0°C, Low 25.0°C (5% rain chance)\n- Aug 10: High 35.8°C, Low 24.5°C (45% rain chance)"

Query 3: Recommendation Logic

• Prompt: "Should I bring an umbrella to Austin tomorrow?"

• Tool Called: predict_umbrella_needed(location="Austin, TX", target_date_offset=1)

• Response: "You do not need an umbrella in Austin tomorrow. The rain probability is 5% with expected precipitation of 0.0 mm."