Watch an agent plan dog walks

A small language model with three tools: an address lookup, a route optimizer, and a weather service that returns deterministic safety verdicts. Below, you can watch it work — every tool call, every validation bounce, every audit veto, streamed live. The interesting part isn't that it works; it's the machinery that keeps an 8-billion-parameter model honest.

Run a demo

loading presets…

…or build a custom roster

Up to 6 dogs. Walks are 20, 30, or 60 minutes — the three products a dog walker actually sells. Custom runs are rate-limited; one plan runs at a time.

What you just watched

The agent is a hand-rolled loop — no framework — around a small model (qwen3-8b) doing native tool calling. Three mechanisms keep it honest: a referee that validates every tool call against its schema before anything executes (violations bounce back as errors the model reads and corrects); an auditor — plain code, no AI — that refuses to accept a finished plan until every dog's actual walk interval has a weather check at that dog's location; and a structured finish line: the model ends by calling a submit_plan tool whose validated arguments are the answer you see above. Weather verdicts come from threshold tables in code, not from the model's judgment — the model relays them and plans around them.

The service runs on a retired laptop in a closet (the inference is rented per-token); the whole stack costs about a dollar a month. Source, tests, and the full build history: the repo.