Jev integration reference - Failproof AI

Jev has two uses in Failproof AI:

Use	When it runs	What it returns	Start here
Session evaluation	After a session finishes	A score for a fixed-answer question	Jev evaluations
Tool-call policy review	Before a gated tool call runs	A verdict alongside the installed policies	Jev policies

Reference pages

Topic	Details
Evaluation questions	Boolean and ordered-score criteria, results, limits, and backfill.
Provider comparison and own-key setup	TypeSafe, OpenRouter, Vercel, Cloudflare, and custom endpoints; URL inference, model IDs, `jev.json`, modes, and fallback codes.
FailproofAI Cloud route	Machine-key permissions, automatic observe setup, usage limits, connection state, and data handling.

The local CLI commands are listed in the Failproof AI CLI reference. The local dashboard reference describes its Jev settings and activity view.