What it is
usher-point is an on-demand CLI for developers that decides where a task should run — inline in Claude Code, via Codex CLI in a sandbox, or in an isolated Orca worktree — and optionally which skills and model to use, then dispatches it.
How it uses Jev
With --engine jev, usher-point optionally calls TypeSafe AI's Jev model via OpenRouter to return target + relevant skills + an optional model/effort override in one call. --engine auto tries Jev first and fails closed to the heuristic on any problem; --engine jev fails loudly.
Technique worth stealing
Fail-closed routing: try the model, fall back to inspectable heuristic rules on any problem.
Try it
npm install -g github:juanpiRiv/usher-point, then run usher or usher-point route "task".
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-centricno0.05
- Shows a System One patternno0.28
- Handles uncertaintyno0.05
- Measuredno0.04
- Runnableyes0.95
- Worth recommendingno0.16
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.16
- About Jevyes0.75
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.