Start¶
Four pages, in order, from an empty environment to an agent that asks Jev before it acts.
InstallUsing a coding agent? Copy this prompt into Claude Code, Codex, Kiro, or any coding assistant and it will walk you through these pages.
You leave with the package installed and your key in place, one decision made from Python with every answer field explained, an agent that routes a request through Jev, directly and then on the model's own initiative, and the line on the bill that all of it cost.
| page | you leave with |
|---|---|
| Install | strands-jev in a Python 3.10 or later environment, the key in TYPESAFE_API_KEY or ~/.typesafe, jev_models answering |
| First decision | one jev_ask call, three typed answers, what noul, choice, confidence, score and legend mean |
| First agent | Agent(tools=ALL_TOOLS), a direct route call, a chat turn where the model picks the tool |
| What it costs | the price line, the budget check that refuses an oversized request, the measured cost of the live suite |
Every python fence on these pages runs against this commit. The ones that talk to api.typesafe.ai spend money at $0.042 per million input tokens; the amounts are in the text next to them.