Weather Prediction MCP Server + Agent Bricks

A Model Context Protocol (MCP) server that exposes weather forecast tools, integrated with Databricks Agent Bricks for natural-language weather queries.

Architecture

┌─────────────────────────────────────┐
│  User Question                      │
│  "Will it rain in Chicago tomorrow?"│
└───────────────┬─────────────────────┘
                │
                v
┌─────────────────────────────────────┐
│  Agent Bricks Agent                 │
│  (Databricks AI)                    │
│  - Interprets question              │
│  - Selects appropriate tools        │
│  - Formats natural-language answer  │
└───────────────┬─────────────────────┘
                │
                v
┌─────────────────────────────────────┐
│  Weather MCP Server                 │
│  (FastMCP, Databricks App)          │
│  Tools:                             │
│  - get_current_weather()            │
│  - get_forecast()                   │
│  - predict_umbrella_needed()        │
└───────────────┬─────────────────────┘
                │
                v
┌─────────────────────────────────────┐
│  Open-Meteo API                     │
│  (Free weather data, no auth)       │
│  - Geocoding                        │
│  - Current conditions               │
│  - Forecast data                    │
└─────────────────────────────────────┘

Related MCP server: Weather Prediction MCP Server

Components

1. Weather MCP Server (weather_mcp_server/)

A FastMCP server deployed as a Databricks App that exposes three weather tools:

Tools:

• get_current_weather(location) - Real-time weather conditions

  • Returns: temperature, humidity, wind speed, conditions

  • Example: get_current_weather("Chicago")

• get_forecast(location, days=7) - Multi-day forecast (1-16 days)

  • Returns: daily high/low temps, precipitation chance, conditions

  • Example: get_forecast("Austin, TX", days=5)

• predict_umbrella_needed(location, target_date=None) - Smart recommendation

  • Decision logic: Umbrella needed if precipitation > 40% OR rain-related conditions

  • Returns: boolean recommendation + reasoning

  • Example: predict_umbrella_needed("Seattle", "2026-08-10")

2. Weather Broker (weather_broker.py)

Adapter module for Open-Meteo API:

• get_current_conditions() - Fetches current weather

• get_multi_day_forecast() - Fetches forecast data

• _geocode_location() - Converts city names to lat/lon

• _map_weather_code() - Translates WMO codes to human-readable conditions

API Choice: Open-Meteo

• Free and open source

• No API key required

• ~10,000 calls/day

• High-quality data from official weather services

3. Agent Bricks Agent

A Databricks AI agent configured with:

• System prompt explaining weather-answering capabilities

• External MCP tool registration pointing to the deployed MCP server

• Natural-language interface for weather questions

Setup Instructions

Step 1: Deploy MCP Server

1. Create Databricks secret (if not exists):

   databricks secrets create-scope mcp_server
   databricks secrets put-secret mcp_server databricks_token

   (Paste your Databricks token when prompted)

2. Deploy the app from GitHub:

   databricks apps create weather-mcp-server \
     --source-code-url https://github.com/amalphonse/weather-mcp-server.git

3. Start the app:

   databricks apps start weather-mcp-server

4. Get the app URL:

   databricks apps get weather-mcp-server

   Note the url field - you'll need this for Agent Bricks.

Step 2: Register MCP Server with Agent Bricks

1. Navigate to the Agent Bricks UI in Databricks

2. Create a new agent or edit an existing one

3. Under "External Tools", add a new MCP server:

   • Name: Weather Forecast Server

   • URL: <your-app-url> (from Step 1)

   • Description: Provides weather forecasts and recommendations

Step 3: Configure Agent System Prompt

Add this system prompt to your agent:

You are a helpful weather assistant with access to real-time weather data.

Capabilities:
- Answer questions about current weather conditions
- Provide multi-day forecasts (up to 16 days)
- Make practical recommendations (e.g., umbrella needed, jacket needed)

Guidelines:
- Always use the available weather tools to get real data
- Never guess or make up weather information
- If a location cannot be found, ask the user to clarify
- If the API is unavailable, inform the user clearly
- Explain your reasoning when making recommendations

When a user asks about weather:
1. Use get_current_weather for "now" / "current" / "today" questions
2. Use get_forecast for future dates or multi-day questions
3. Use predict_umbrella_needed when asked about rain gear or travel prep

Step 4: Test the Agent

Try these example queries:

1. Current conditions: "What's the weather like in Chicago right now?"

2. Forecast: "Will it be hot in Austin this weekend?"

3. Recommendation: "Should I bring an umbrella to Seattle tomorrow?"

Example Agent Conversations

(To be filled in after testing)

Example 1: Current Weather

User: "What's the weather like in Chicago right now?"

Agent: [Calls get_current_weather("Chicago")]

Response: "In Chicago, it's currently 68°F (20°C) with partly cloudy skies. The humidity is 55% and winds are calm at 7 mph. Last updated at 2:30 PM local time."

Example 2: Forecast

User: "Will it rain in Austin tomorrow?"

Agent: [Calls get_forecast("Austin, TX", days=2)]

Response: "Tomorrow in Austin, the forecast shows a high of 92°F and a low of 74°F with mainly clear skies. The precipitation chance is only 10%, so rain is unlikely. It should be a hot, sunny day."

Example 3: Recommendation

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

Agent: [Calls get_forecast("Seattle", days=5) and predict_umbrella_needed("Seattle", target_date)]

Response: "This weekend in Seattle, temperatures will range from 58-65°F. You should definitely bring a light jacket for the cooler mornings and evenings. Also, I recommend bringing an umbrella - Saturday shows a 65% chance of rain with moderate showers expected."

Files

• app.yaml - Databricks App configuration

• requirements.txt - Python dependencies

• weather_mcp_server.py - FastMCP server with tool definitions

• weather_broker.py - Open-Meteo API adapter

• README.md - This file

• .gitignore - Git ignore rules

Dependencies

• fastmcp>=0.4.0 - MCP server framework

• fastapi>=0.115.0 - Web framework

• uvicorn>=0.32.0 - ASGI server

• requests>=2.32.0 - HTTP client

• databricks-sdk>=0.35.0 - Secrets management

Error Handling

• Location not found: Returns clear error asking user to clarify

• API outage: Returns error message explaining the service is unavailable

• Invalid date range: Returns error with valid date range

• Network timeout: 10-second timeout with error message

Notes

• Open-Meteo is free for non-commercial use (~10k calls/day)

• No API key required - zero credentials to manage

• Data sourced from official weather services (NOAA, DWD, etc.)

• Forecast accuracy: Best within 7 days, still useful up to 16 days