This server exposes two simple weather tools with basic descriptions and schemas. Both tools are read-only and directly callable via the fastmcp framework. The tool definitions are visible in weather.py and properly registered with @mcp.tool() decorators. However, descriptions lack depth and actionability, parameter descriptions are minimal, output schemas are undocumented, and error handling is perfunctory. The tools follow a verb_noun naming pattern (get_alerts, get_forecast), which is good, but descriptions don't explain WHEN to use each tool, what data structures are returned, or how to handle failures. Parameter validation is weak, latitude/longitude accept any float, state accepts any string without enum validation. Output is free-form text rather than structured JSON. Error messages are generic ('Unable to fetch alerts or no alerts found') and don't guide recovery.

• Add explicit parameter validation: get_alerts should validate state is exactly 2 uppercase letters (enum or regex), get_forecast should validate latitude in [-90, 90] and longitude in [-180, 180]. Return clear validation errors: 'Invalid state code: got "CA1". Must be 2-letter US state code (e.g. CA, NY, TX).'
• Expand tool descriptions to 100-200 chars and include WHEN to use each. Example: 'Get active weather alerts (severe storms, flood warnings, etc.) for a US state. Use this when the user asks about dangerous weather conditions. Returns alerts with severity, affected area, and instructions.'
• Expand parameter descriptions with constraints and format. Example for state: 'Two-letter US state code (e.g. CA, NY, TX). Must be uppercase.' For latitude/longitude: 'Geographic coordinate (latitude -90 to 90, longitude -180 to 180 in decimal degrees)'.
• Return structured JSON instead of formatted text. Define a schema for alert objects: {event, area, severity, description, instructions}. Define forecast schema: {periods: [{name, temperature, temperatureUnit, windSpeed, windDirection, detailedForecast}]}. Document the output schema in the tool description or via a MCP response type annotation.
• Implement error classification and recovery guidance. Catch network timeouts and return '{"error": "Service timeout. The weather API took too long to respond. This is typically retryable. Wait a few seconds and try again." "retryable": true}'. For invalid coordinates, return '{"error": "Unable to fetch forecast for location ({latitude}, {longitude}). Check coordinates are valid (lat -90..90, lon -180..180). Did you mean a nearby city?"}' .