Build a Weather Tool for AI Assistants
This playbook walks through creating a weather tool that AI assistants can discover and call via MCP.
For general setup, connecting clients, and authentication, see the MCP Server guide.
Create the Weather Tool
Open the Qaynaq UI, click Create New Flow, and configure each section:
Input - select MCP Tool
| Field | Value |
|---|---|
| Name | get_weather |
| Description | Get current weather for a city |
| Input Parameters | city (string, required) - "City name" |
Processor - select Mapping
For this example, we return a random temperature. In practice, you would call an external API or use other processors.
| Field | Value |
|---|---|
| Mapping | root.temperature = random_int(min:0, max:40).string() + "C" / root.city = this.city |
Output
The output is automatically set to Sync Response and locked when using MCP Tool input.
Click Save and then Start the flow.
Test the Tool
Once the flow is running, the get_weather tool appears automatically on the /mcp endpoint. You can verify it by asking your AI assistant something like "What's the weather in London?" and it will call the tool.
See MCP Server - Verifying for testing via the command line.