jev-zig-cli
Auditable Zig terminal agent pairing Codex for coding with Jev typed preflight and postflight checks under deterministic policy gates.
Overview
jevx is an unofficial terminal agent written in Zig that uses the authenticated OpenAI Codex CLI for generative coding work and restricts Jev, accessed through OpenRouter, to bounded Choice, Noul and Score decisions for routing, hazard detection, impact scoring and verification. Ordinary Zig code, not Jev, owns permissions and final control flow.
Each prompt is redacted, then a typed preflight batches action routing, five hazard checks and impact scoring into one validated Jev request; a deterministic policy gate with tested thresholds can deny or ask for confirmation; a durable pre-action audit record is written; Codex executes; and a postflight verification step allows at most one repair turn. Journal records are redacted, encrypted, authenticated, sequenced and hash-chained, credentials are stripped from output, recovery preserves history without unsafe replays, and output is versioned JSONL with stable exit classes. The README explicitly claims no better intelligence, only auditability and deterministic policy. v1.0.0 ships verified release builds for macOS and Linux with CI and security workflows.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which action route does this prompt require? | — |
| Yes / no | Does this request carry hazard X? (five hazard checks) | — |
| Score | What is the impact of this action? | — |
| Yes / no | Did the Codex result satisfy the request (postflight verification)?(inferred) | — |
How it works
- Redact the prompt
- Typed Jev preflight: routing, five hazards, impact in one request
- Deterministic policy gate denies, confirms or allows; audit record written
- Codex executes; Jev postflight verification with at most one repair turn
Strengths
- Permissions never delegated to a model
- Encrypted, hash-chained audit journal
- Verified release builds and CI
- Scriptable JSONL output
Caveats
- Unofficial; depends on Codex CLI and OpenRouter
- No claim of better coding quality
Getting started
curl -fsSL https://raw.githubusercontent.com/SupratimSircar05/jev-zig-cli/v1.0.0/install.sh | sh, then `jevx doctor`.At a glance
- Type
- CLI
- Status
- production
- Language
- Zig
- Runtime
- CLI
- Has code
- yes
- Has tests
- yes
Use cases and integrations
How curated lists describe it
Unofficial Jev-powered Zig terminal agent with deterministic policy gates and encrypted audit trails.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev do in jevx?
Routing, hazard checks, impact scoring and postflight verification; it never controls permissions.
Who writes the code?
The OpenAI Codex CLI.
How is the audit trail protected?
Records are redacted, encrypted, authenticated, sequenced and hash-chained.
Related projects
Risk gate for AI coding agents: Jev decides at five layers and Claude Code hooks enforce deny or ask deterministically.
ProjectbetaJavaScriptLocal guard layer for coding agents that blocks dangerous tool calls by deterministic rules, with optional Jev judgments.
CLIbetaTypeScriptAgent-ergonomic Jev CLI and PreToolUse hook that scores shell commands for hazards and screens fetched text for injection.
PluginbetaTypeScriptClaude Code plugin adding typed Jev decisions to hooks: task kind, command risk, failure class, completion; never grants permissions.
PluginbetaJavaScriptPreToolUse firewall for Claude Code and Codex: shell-aware rules first, then Jev returns allow, ask or block, failing closed.
HookbetaTypeScriptSafety gate plugin for Claude Code, Codex CLI and Antigravity: local sensitive-file checks, then Jev judges blast radius.
PluginbetaTypeScript