Create the "get_weather" tool

Click "Create Tool" on the MCP Server in RapidMCP and add the following details:

Tool Name: get_weather

Tool Description: Get the weather for a given location

Tip: description should be description so the model has context what this tool call does

Tool Arguments:

{
  "latitude": 52.52,
  "longitude": 13.419998,
}

Tip: Tool Arguments is a json markup that has test values to specify the expected arguments from the model. This markup will be replaced with real values when the model calls the tool. RapidMCP automatically transforms this json markup into a ZodObject for you!

Define the tool trigger

This REST API is taken from OpenMetro for illustrative purposes. You should put your own API details here.

Tool Method: GET

Tool Endpoint: https://api.open-meteo.com/v1/forecast?latitude={{properties.latitude}}&longitude={{properties.longitude}}&current=temperature_2m,wind_speed_10m

Tip: notice the use of {{ and }} to template the tool arguments from the model into query args.

Request Body:

{
  "lat": "{{properties.latitude}}",
  "long": "{{properties.longitude}}",
}

Tip: notice the use of {{ and }} to template the tool arguments from the model into request body.

Test the tool call by clicking the play button.

Click the play button to test the tool trigger. If the templating looks good, you're ready to add the server url to the AI Agent.

Since this is just a test, we want the AI Agent to now call this tool. To do so, add the server url into AI Agent MCP configuration. See next steps.

Add the MCP Json spec to an AI Agent

Find the server URL from the "Setup" tab. Add this spec to your AI agent. Note: the agent must support custom MCP servers (not all models support this yet).

{
  "mcpServers": {
    "RapidMCP": {
      "url": "https://rapid-mcp.com/mcp/<serverId>/stream",
    }
  }
}