What it is
ActionGate is a runtime security gateway that evaluates proposed tool calls with deterministic policy and Jev, then issues single-use permits bound to the exact action. It is for developers building AI agents that need to authorize tool calls.
How it uses Jev
ActionGate asks six narrow Jev questions in one request: alignment, target match, policy conflict, sensitive-data exposure, scope expansion, and missing intent. The model returns calibrated probabilities that supply evidence, but code owns authority; a model score never overrides RBAC, schema, limit, or duplicate failures. Only an enforced ALLOW produces a grant.
Primitives:choicescorenoul
Technique worth stealing
Asking six narrow typed questions instead of one vague safety question, and treating model scores as evidence while code retains authority.
Try it
git clone, pnpm install, pnpm dev; then pnpm refund:demo.
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-centricyes0.74
- Shows a System One patternno0.21
- Handles uncertaintyno0.06
- Measuredno0.04
- Runnableyes0.91
- Worth recommendingno0.38
- Model replicano0.04
- Problem scopescore on a 0–2 scale1.43
- About Jevyes0.98
Signals by Jev jev-1.13.0, card written by DeepSeek V4.1 Flash from the README on 20 Sept 2026.