Demo
The use-infrared skill
The public use-infrared skill and a recorded agent run show an agent running the engine end to end.
Let your agent run the physics
A language model can only guess the local climate. With infrared.city, your agent runs a simulation and reads the result.
Your agent runs a simulation in seconds and reads the result.
Ask a general model which street is hotter and you get a generic answer. It has no physics at street level.
Give your agent the public use-infrared skill. The agent calls the engine through the Python or TypeScript SDK and reads back felt heat, wind comfort, sun and shade, solar radiation and daylight. A run takes 1 to 4 s after the geometry is loaded.
You can run the engine and your agents on your own servers.
- NowPython and TypeScript SDK
- Nowuse-infrared agent skill
- SoonMCP server
- SoonAgent plugin
- 1–4 sRun time of one simulation on one 512 m tile, after the geometry is uploaded once.
- per simulation run by your agent
- 12Felt heat (UTCI) and its statistics, wind speed, pedestrian wind comfort, solar radiation, direct sun hours, daylight availability, daylight factor, sky view factor and energy balance.
- simulation types to call
- 0
- models to train or host yourself
Your agent answers with a simulation result, not a guess.
Funded and supported by
