What it is
A Claude Code, Codex, and pi plugin plus library that routes agent steps needing no text output to Jev, TypeSafe's judgment model, so LLMs and Jev collaborate: Jev decides fast, the LLM writes.
How it uses Jev
Jev judges typed questions (check/pick/rate) over a supplied state; answers carry confidence and an escalate flag. Untypeable or generation-needing questions are rejected before the call, and low-confidence answers return as flagged priors.
Primitives:choicescorenoul
Technique worth stealing
Pre-call routing rejects untypeable questions and returns low-confidence answers as flagged priors with typed escalation.
Try it
npx -y jev-use install, set a provider key, then npx -y jev-use doctor.
Evidence
Each line is one question put to Jev about the README. ≥ 0.60 reads as yes, ≤ 0.40 as no; in between Jev is not making a call.
- Jev-centricyes0.87
- Shows a System One patternyes0.95
- Handles uncertaintyyes0.92
- Measuredyes0.94
- Runnableyes0.94
- Worth recommendingyes0.66
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.10
- About Jevyes0.99
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.