Cool Route Planner is a heat-aware routing and scheduling agent for people who work, ride or walk outdoors. Ask it for the coolest path between two points or the coolest hour for an outdoor task. It answers from hyperlocal 2m temperature, not a city-wide forecast.
"42 degrees in Delhi today" cannot tell a delivery rider which of three roads bakes hardest at 2pm. It cannot tell them the 7am window feels far cooler than the afternoon. Heat is hyperlocal. Shade, surface and humidity move the felt temperature more than the map does. That gap is a health risk. It is exactly what this agent reads.
Two planners running against the deployed backend right now. Each number is the felt temperature returned by the data source, placed on the NWS heat-risk scale. The coolest option is marked.
The agent is a reasoning model that calls the FortyGuard Temperature API as its tools. It never invents a temperature. It ranks options on the returned field, then leads with one recommendation and a safety note.
A route between two points or a time window for an outdoor task, in plain language.
Hyperlocal 2m temperature, felt-heat index, wet-bulb and street-level shade at each sampled point.
Deterministic planners rank routes or hours by felt temperature. The numbers come from the API, not the model.
One recommendation, the felt temperature and an NWS heat-risk band with a safety note.
Felt temperature uses the NWS heat-index regression. Wet-bulb uses Stull (2011). Risk bands follow the NWS heat-index categories.
Every recommendation lands on the NWS heat-index bands, read in felt temperature. This is the scale behind each dot and pill above.
Reading the backend from the live health check.