MCP server for AI agents — weather forecasts, air quality, UV index, historical data. Powered by Open-Meteo (free, unlimited, no API key).
Five tools with clear verb-noun naming (get_*, compare_*) and consistent descriptions (100-150 chars). All tools have input schemas with Zod validation and type definitions. However, parameter descriptions are minimal (10-30 chars), lacking format guidance, constraints, and LLM-actionable context. Output schemas are undocumented, responses are plain text blocks rather than structured JSON. No error handling guidance, no pagination support, and no field-level documentation for chained calls. Baseline: avg tool description ~120 chars (good), but avg param description ~15 chars (poor). Schemas present but sparse.
Compare current weather across 2-5 cities side by side
Get current air quality index and pollutant levels for a location
Get historical weather data for a location and date range
Get current weather and 7-day forecast for a location
Check for weather alerts — UV, extreme temps, heavy rain, strong wind
Parameter descriptions are too brief (10-30 chars). 'City name' and 'Start date (YYYY-MM-DD)' lack actionable context for LLMs. Should include format constraints, valid ranges, and examples of what to pass.
Output schemas are undocumented. Tools return plain text blocks (lines.join('\n')) rather than structured JSON. LLMs cannot extract fields for downstream calls or parse results reliably. Should return {current: {temp, humidity, wind}, forecast: [{date, min, max, condition, precip}]}.
No error handling guidance. Errors like 'Location not found' or 'API error: 500' do not tell the LLM what to do next (retry, ask user, try alternative). Should return {error: 'location_not_found', message: '...', recovery: 'Try search_locations() or check spelling'}.
| Scored | Grade | Overall | Spec posture | Rubric |
|---|---|---|---|---|
| 2026-09-23 | D | 52 | 2026-07-28+ | v2 |
No pagination or result limits documented. Tools return all 7 days of forecast or all historical data without caps. Large result sets waste tokens and risk context window exhaustion. Should document max results and offer limit/offset parameters.
Telemetry ping to 'radom.group/weather-ping' is undocumented and may violate user privacy expectations. Should be opt-in, documented in README, and not called silently on startup.