Skip to content
JevDirectory.orgTraffic count unavailable. View site traffic

Jev directory

CommunityJev Tools & CLIs

Jev Guard — leepokai

A security hook for coding agents that asks Jev three typed questions before every tool call and scans tool results for prompt injection, with adapters for Claude Code, Codex, Copilot CLI, Gemini CLI, Cursor, pi, OpenCode and ACP.
See highlightsOpen github.com
Stars
94
Language
JavaScript
License
MIT
Reading time
3 min
Published by
Community
Author
leepokai
Last verified
Sep 2026

About this project

jev-guard A security hook for coding agents, powered by Jev .

Quoted from the project's README. Read the full README.

Tagscommunityjavascriptagentscoding-agentsecurityguardrailsclaude-codecodex
Works forJev for ModerationJev for CodeJev for Tool callingJev for Decision gating
Used inJev for Security & Privacy

Highlights

  • Before a tool runs, Jev scores the harm, whether the user requested it, and whether it came from untrusted content; destructive calls are denied.
  • After a tool returns, results are scanned for prompt injection and canaries, and hits are remembered for the rest of the session.
  • Instruction files such as skills and CLAUDE.md are checked for exfiltration, covert execution, instruction override, and canaries.
  • A typical call is about 1k tokens, roughly $0.00004 per tool call, with measured p50 around 580 ms through the Vercel AI Gateway.
  • On a machine with 662 installed skills, none crossed the flag line, while planted exfiltration samples scored 0.99.

Quickstart

bash
npm i -g jev-guard
jev-guard key YOUR_API_KEY
jev-guard install claude

Watch out

MIT-licensed. Needs Node.js 20.3+ and a TypeSafe or Vercel AI Gateway key; it is a guardrail, not a sandbox, and tool results can contain anything the agent read.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: Vercel's fx safety reviewer, measured at 18× faster

Reddit: Early experiments wiring Jev into Pi as a safety layer and model router.

YouTube: Ray Amjad puts Jev inside an agentic coding loop and looks at what it costs to run.

YouTube: RoboNuggets puts Jev to work inside Claude Code.

Other listings of this project

Forks and copies of this project in the directory. This entry is the canonical one.

jev-guard
7GitHub stars
Claude Code plugin that scores pending tool calls and finished answers with Jev noul probabilities, denying the call with the correction attached when a rule threshold is met.
Jev Agents & MCP#community#awesomejev#catalog-import
Communitymuratcakmak/jev-guardTypeScript
jev-guard
3GitHub stars
A safety gate for AI coding agents that checks every shell command, file write, and file read before it runs and returns ALLOW, ASK, or DENY: most calls are decided locally in under a millisecond, while ambiguous or destructive ones go to TypeSafe's System One model.
Jev Tools & CLIs#community#catalog-import#logicrw
Communityclemensschartmueller/jev-guardGo
jev-guard
1GitHub stars
PI agent extension pairing in-flight terminal-output pruning with a surgical diff and secret guard, both driven by typed Jev questions plus a deterministic secret-literal layer.
Jev Tools & CLIs#community#github
Communityyelkhanyergali-sys/jev-guardJavaScript
JEV Guard
0GitHub stars
Open, configurable guard for company AI agents built on Jev. Agents call it before sending email, launching campaigns, or editing a CRM, and receive allow, block, or hold. Companies are described in a YAML profile; includes a demo UI, human review inbox, and eval scripts.
Jev Agents & MCP#guardrails#agents#python
CommunityAlexBzal/jev-guard

From the Jev guides

Original write-ups on what this entry is for.

From the Jev AI model community

Posts from builders shipping with Jev right now.

Follow @typesafeai