Live demo

See every call your agents make.

jev-gate judges each action — approve, hold, escalate — and logs all of it: confidence, violation, risk, transcript. No black box: audit everything below, tune thresholds against the muddy middle. Try a trace.

1

Agent trace

3 preset cases. Demo limit 30/hr. Powered by TypeSafe Jev. Preview only — get your own key for a persistent log.

2

Decision + why

Waiting for a trace…

3

Session preview

Preview only — rows stay in this browser tab. Your real, persistent log lives in your dashboard.

4

Confidence histogram

Uncertain calls (0.4–0.6) are the ones worth reviewing — that is your review queue.

POST /v1/gate → {action, confidence} · GET /v1/decisions → full log · GET /v1/confidence → histogram · GET /v1/review-queue → muddy middle first