Cookbook
Four recipes that put Zen, Enso and Kai in one loop: route a request, triage with typed answers, gate an agent's command, and drive an agent from programs instead of prompts.
Each recipe is one job, end to end, in requests you can paste. Together they show the pattern the platform is built around: Kai answers the bounded questions, Zen writes where writing is the work, Enso decides which is which, and your policy keeps the last word.
| Recipe | What it shows | Runs on |
|---|---|---|
| Route with Enso | one request for every model, a price cap per call, the model that answered, feedback that trains the router | api.hanzo.ai |
| Triage with Kai, reply with Zen | three decisions with no generated tokens, then one generation | Kai runtime + api.hanzo.ai |
| Gate an agent's command | a Decision Program's thresholds and verdict rule, joined with your policy by mode | Kai runtime |
| Drive an agent loop | the loop's control flow as programs — preflight, generate, check done — not as prompts | Kai runtime + api.hanzo.ai |
Kai runtime: POST /v1/decisions is served by the decision runtime,
decision serve, which listens on 127.0.0.1:8080 by default. It is not
public yet and api.hanzo.ai does not serve the route yet — see
status. The Kai responses quoted in these recipes
are real, from that runtime serving the kai-1 baseline on CPU, not from a
trained Kai checkpoint.
export HANZO_API_KEY=... # console.hanzo.ai
export DECISIONS=http://127.0.0.1:8080 # decision serveHow is this guide?
Last updated on