What it is
A coding agent and personal assistant built on the Pi coding agent. It works with any LLM via OpenRouter or a direct provider, takes voice input, and includes macOS computer use, a browser agent, MCP connectors, scheduled agents and deploys.
How it uses Jev
Jev judges each action, turn, utterance and request: action gating on bash, edit, write, computer, browse and connector calls; progress and completion checks; model routing; skill and connector selection; voice intent; browser steps; credential requests; and agent workflow decide steps. Code owns thresholds, escalation and routing, and acts on the returned probabilities.
Primitives:choicescorenoul
Technique worth stealing
Code owns control flow: many narrow questions in one parallel request, mutually exclusive options with escapes, and confidence thresholds scaled to risk.
Try it
npm install && npm run build, npm link, then run reflex; onboarding asks for provider key, TypeSafe key and risk appetite.
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-centricunclear0.58
- Shows a System One patternyes0.93
- Handles uncertaintyno0.05
- Measuredno0.36
- Runnableyes0.86
- Worth recommendingno0.33
- Model replicano0.06
- Problem scopescore on a 0–2 scale1.48
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.