How the LLM Uses These Tools
Example 1: Weather Alerts
You: "Are there any weather alerts in California?"
Claude's process:
Recognizes it needs weather alert info
Calls
get_alerts(state="CA")Gets the formatted response
Presents it to you in natural language
Example 2: Weather Forecast
You: "What's the weather forecast for San Francisco?"
Claude's process:
Knows SF coordinates (or looks them up)
Calls
get_forecast(latitude=37.7749, longitude=-122.4194)Gets 5-period forecast
Summarizes it for you
Update the claude config file claude_desktop_config.json to below content
{ "mcpServers": { "weather": { "command": "/Users/santhosh.sharma/.local/bin/uv", "args": [ "--directory", "/Users/santhosh.sharma/Repositories/mcp-weather", "run", "weather.py" ] } } }
Reference : https://modelcontextprotocol.io/docs/develop/build-server#python
Analyze logs in ~/Library/Logs/Claude/mcp.log
When you ask Claude (with this MCP server connected):
local-only server
The server can only run on the client's local machine because it depends on local resources.
Provides weather forecast and alert information from the National Weather Service API, allowing users to query weather forecasts by coordinates and check weather alerts by state.