What it is
jevwire wires TypeSafe AI's Jev classifier into agent harnesses. It offers 6 MCP tools, an embeddable JevDecisionModel library, and a Claude Code plugin whose hooks judge tool calls, fetched results, and turn endings.
How it uses Jev
Jev answers typed questions over a state: tool-call descriptions, user prompts, fetched results, and final messages. The returned probabilities drive notes, tripwires, stop blocks, injection flags, and contradiction notes addressed to Claude, not the user.
Primitives:choicescore
Technique worth stealing
Put mandatory checks in the harness via pure run* functions, not in a tool an agent may decline to call.
Try it
npm i jevwire, or /plugin marketplace add Brainwires/jevwire then /plugin install jev@brainwires-jevwire.
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-centricyes0.92
- Shows a System One patternyes0.91
- Handles uncertaintyno0.18
- Measuredno0.06
- Runnableyes0.97
- Worth recommendingunclear0.46
- Model replicaunclear0.42
- Problem scopescore on a 0–2 scale1.47
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.