Model Context Protocol (MCP)

The Model Context Protocol (MCP) enables AI assistants to access WindBorne's real-time weather data through a secure, standardized interface.

Available Endpoints

All WindBorne API endpoints are accessible through MCP. With API keys, you get access to the full set:

• Observations: observations, soundings, super observations, constellation status, flying missions, flight paths, predicted paths, current location, launch sites
• Forecasts: point forecasts, interpolated point forecasts, gridded forecasts, station forecasts, forecasted soundings, initialization times, run information, variables, historical gridded and interpolated data
• Insights: population-weighted heating and cooling degree days (HDD/CDD), calculation times
• Severe Weather: tropical cyclone tracking and prediction

Endpoints marked [free] in the API docs (interpolated point forecast, initialization times, run information, variables, archived initialization times) are available without authentication alongside the documentation tools.

Two Modes of Operation

Getting Started

1. Install the required Python package

   Install Command

     pip install pydantic-ai>=0.3.2pip install pydantic-ai>=0.3.2

   Show moreShow less

2. Set API credentials as Environment Variables

   Shell Commands

     export WB_CLIENT_ID=your_client_idexport WB_CLIENT_ID=your_client_id

     export WB_API_KEY=your_wb_api_keyexport WB_API_KEY=your_wb_api_key

   Show moreShow less

   These commands are also available when the API key is created.

3. Connect to MCP Server using Python

   Python Connection

     from pydantic_ai.mcp import MCPServerStreamableHTTPfrom pydantic_ai.mcp import MCPServerStreamableHTTP

    

     windborne_server = MCPServerStreamableHTTP(windborne_server = MCPServerStreamableHTTP(

     url="https://mcp.windbornesystems.com/mcp", url="https://mcp.windbornesystems.com/mcp",

     headers={ headers={

     "wb-client-id": os.getenv("WB_CLIENT_ID"), "wb-client-id": os.getenv("WB_CLIENT_ID"),

     "wb-api-key": os.getenv("WB_API_KEY") "wb-api-key": os.getenv("WB_API_KEY")

     } }

     ))

   Show moreShow less

4. Complete Example

   Test Connection (Python)

     import asyncioimport asyncio

     import osimport os

     from pydantic_ai.mcp import MCPServerStreamableHTTPfrom pydantic_ai.mcp import MCPServerStreamableHTTP

    

     # Get credentials from environment variables# Get credentials from environment variables

     WB_CLIENT_ID = os.getenv("WB_CLIENT_ID")WB_CLIENT_ID = os.getenv("WB_CLIENT_ID")

     WB_API_KEY = os.getenv("WB_API_KEY")WB_API_KEY = os.getenv("WB_API_KEY")

    

     async def main():async def main():

     # Create MCP client # Create MCP client

     client = MCPServerStreamableHTTP( client = MCPServerStreamableHTTP(

     url="https://mcp.windbornesystems.com/mcp", url="https://mcp.windbornesystems.com/mcp",

     headers={ headers={

     "wb-client-id": WB_CLIENT_ID, "wb-client-id": WB_CLIENT_ID,

     "wb-api-key": WB_API_KEY "wb-api-key": WB_API_KEY

     } }

     ) )

    

     # Use the client as an async context manager # Use the client as an async context manager

     async with client: async with client:

     # Call a tool # Call a tool

     result = await client.call_tool( result = await client.call_tool(

     tool_name="get_interpolated_point_forecast", tool_name="get_interpolated_point_forecast",

     arguments={ arguments={

     "coordinates": "37.7749,-122.4194", "coordinates": "37.7749,-122.4194",

     "model": "wm-5c", "model": "wm-5c",

     "min_forecast_hour": "0", "min_forecast_hour": "0",

     "max_forecast_hour": "24" "max_forecast_hour": "24"

     } }

     ) )

    

     # Print result # Print result

     print("Forecast result:", result) print("Forecast result:", result)

    

     if __name__ == "__main__":if __name__ == "__main__":

     asyncio.run(main()) asyncio.run(main())

   Show moreShow less

MCP for Claude Code

Use the Claude Code CLI to connect with header-based auth (no credentials in the URL).

1. Set API credentials as environment variables

   Shell

     export WB_CLIENT_ID=your_client_idexport WB_CLIENT_ID=your_client_id

     export WB_API_KEY=your_wb_api_keyexport WB_API_KEY=your_wb_api_key

   Show moreShow less

2. Add the WindBorne MCP server

   Replace WB_CLIENT_ID and WB_API_KEY with your actual credentials, or use the env vars from the previous step.

   Shell

     claude mcp add --transport http -s user windborne-api https://mcp.windbornesystems.com/mcp \claude mcp add --transport http -s user windborne-api https://mcp.windbornesystems.com/mcp \

     --header "wb-client-id: $WB_CLIENT_ID" \ --header "wb-client-id: $WB_CLIENT_ID" \

     --header "wb-api-key: $WB_API_KEY" --header "wb-api-key: $WB_API_KEY"

   Show moreShow less

3. Test the connection

   Start Claude Code and ask something like "What's the weather in New York?" to verify connectivity.

MCP for Claude Paid Users

This requires the Claude desktop app. Complete these steps:

1. Access Connectors

   Under your account settings, open "Connectors" and press "+ Add custom connector". If you don't see this section, you are not on a paid plan. Use the free plan steps below.

   

2. Configure your Connector

   

   Give your connector a name and a user-specific URL. Replace WB_CLIENT_ID and WB_API_KEY with your actual credentials (available on the API Keys page).

     https://mcp.windbornesystems.com/mcp?client_id=WB_CLIENT_ID&api_key=WB_API_KEYhttps://mcp.windbornesystems.com/mcp?client_id=WB_CLIENT_ID&api_key=WB_API_KEY

   Show moreShow less

3. Test the Connection

   Activate the tool in prompt settings and ask for the weather to verify connectivity.

   

MCP for Claude Free Users

This requires the Claude desktop app. Complete these steps:

1. Create or Locate the Configuration File

   When you start the Claude desktop app it auto-detects MCP servers from claude_desktop_config.json . Locate it via the app settings or your file explorer, or create it if missing.
   
   For Windows: %APPDATA%/Claude/claude_desktop_config.json
   For macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

2. Add MCP Server Details

   Add this configuration to enable the WindBorne MCP server:

   claude_desktop_config.json

     {{

     "mcpServers": { "mcpServers": {

     "windborne": { "windborne": {

     "command": "npx", "command": "npx",

     "args": ["mcp-remote", "https://mcp.windbornesystems.com/mcp?client_id=WB_CLIENT_ID&api_key=WB_API_KEY"] "args": ["mcp-remote", "https://mcp.windbornesystems.com/mcp?client_id=WB_CLIENT_ID&api_key=WB_API_KEY"]

     } }

     } }

     }}

   Show moreShow less

3. Restart Claude & Test the Integration

   Restart the app and activate the WindBorne MCP in your prompt settings to test.

   

API Docs MCP

The MCP server also provides documentation tools, available both with and without API keys. Without authentication, these are the only tools exposed. With authentication, they are available alongside the live data tools.

• read_api_docs Search and read API documentation by topic
• list_api_endpoints See all available API endpoints

To connect without authentication (docs-only mode):