Watch Light Agent plan, run its tools, and answer in real time. Try the live agent below, or plug in your own model key to make it yours.
Every answer comes with the steps it took, so you can see it actually working.
Light Agent shows its working steps live: parsing your ask, running the right tool, formatting the result.
A built-in engine with its own personality handles math, time, conversions, and more, with zero setup.
Paste an OpenAI-compatible API key in Settings and Light Agent streams real model answers instead.
Responses type out word by word as they're produced, rendered with clean formatting as they land.
One HTML file, no backend, no build step. Drop it on any static host and it's live in minutes.
Your API key never leaves the browser except to call the model you chose. Nothing is collected.
The same loop a big agent runs, running right in your browser.
Light Agent reads what you want, not just what you typed.
It picks the right tool for the job, shows each step as it runs, and verifies the result.
Not just an answer: the steps it took and the result, formatted and streamed in.
Connect a model to make Light Agent fully yours. Leave it empty to use the built-in custom engine.
Stored only in this browser. Sent only to the endpoint above.