| name | weather-signal |
| description | Open-Meteo weather forecasts, geocoding, saved locations, demand signal features, and threshold filtering via MCP + CLI. Use when a task needs current/daily/hourly weather, weather-derived demand features for forecasting, city geocoding, or repeatable business-location aliases. |
Weather Signal
Weather and demand-signal features via Open-Meteo. Prefer MCP tools when available; fall back to the weather-signal CLI.
Tool Map
All MCP tools are prefixed weather-signal_:
| Need | MCP Tool |
|---|
| Current conditions | current_weather |
| Daily forecast (1-16d) | daily_forecast |
| Hourly forecast (1-384h) | hourly_forecast |
| Compact summary + risk/warm/wet flags | weather_summary |
| Demand forecasting features | demand_signal |
| Filter days by thresholds | threshold_days |
| Historical/archive data | historical_weather |
| Resolve place names | geocode |
| Saved location aliases | list_places |
| Cache inspection | cache_status |
Escalation Pattern
weather_summary — compact overview (risk days, warm/hot/wet/windy counts)
demand_signal — feature enrichment for forecasting
daily_forecast / hourly_forecast — raw weather variables
threshold_days — filter to days matching decision rules
current_weather — now/current context
historical_weather — backtesting and feature engineering
Defaults
- Output: JSON
- Forecast horizon: 7 days
- Hourly horizon: 48 hours
- Country hint for London:
--country GB
- Cache: prefer cached for repeatable analysis; use
--refresh for live decisions
Guardrails
- Weather signals are features for forecasting, not standalone demand forecasts.
- Ambiguous place names: use
geocode with --country, never guess London/Paris/Springfield.
- Saved locations: MCP can list but cannot add or remove — use CLI for mutations.
- For batch output, treat item-level
error values as partial failures.
- MCP HTTP server loopback-only unless an authenticating proxy controls access.
Sub-topics