jevbooks

← All projects

jev-axi

shiftynick/jev-axi · Homepage

Agent-ergonomic CLI for TypeSafe's Jev calibrated judgments.

What it is

jev-axi is a CLI that gives coding agents a second opinion using TypeSafe's Jev, a model that answers typed questions with calibrated probabilities instead of text. It provides commands like pick, rate, check, rank, triage, and guard, plus a safety hook for Claude Code and Codex.

How it uses Jev

Jev scores shell commands for destructiveness, exfiltration, remote code execution, and security weakening. Routine commands are decided locally; others are sent to Jev with credentials redacted. The resulting probabilities drive allow, ask, or block decisions in the PreToolUse hook.

Primitives:choicescorenoul

Technique worth stealing

Local pre-filtering of routine commands so only risky ones reach Jev, keeping cost and latency near zero.

Try it

echo '{"tool_name":"Bash","tool_input":{"command":"pnpm test"}}' | jev-axi hook pre-tool-use --explain

View on GitHub

judged by Jevjev-1.13.0

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-centricyes0.94
  • Shows a System One patternyes0.89
  • Handles uncertaintyno0.08
  • Measuredyes0.96
  • Runnableyes0.98
  • Worth recommendingunclear0.59
  • Model replicano0.03
  • Problem scopescore on a 0–2 scale1.58
  • About Jevyes0.98

Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.