Does the guard ever block my agent when the scorer is down?

No. Every failure path passes the call through: no key, endpoint down, or malformed response means the tool call proceeds, so the guard degrades to no enforcement rather than blocking work.

How do I stop it from nagging about deploys I already approved?

The deploy gate is deliberately literal: a general go-ahead does not satisfy the authorization question, so you confirm in the command's own words; tune denyThreshold around the 0.7 default or narrow the rule set if the friction is wrong.

What does the answer check actually verify?

It asks how likely a claim is unsupported by evidence in the same turn, using the answer, recent tool results, and the prompt; it catches finished-sounding answers over turns that never ran the check, but it does not verify truth.

Will my code leave my machine?

Yes, when Jev rules run: each guarded tool call sends the file, command, prompt, or answer plus tool results to the configured Jev endpoint, which defaults to a third-party URL; point the base URL somewhere you control or leave JEV_RULES empty for the regex tier alone.

What does the deploy gate check before denying a command?

Two noul questions on every Bash command: whether it deploys, publishes, merges, or sends a credential, and whether the user's last prompt asked for exactly that. It denies when the first is at or above denyThreshold and the second is below 0.5.

Can the rule checks run without sending anything to Jev?

Yes. REGEX_RULES decide locally with no network and no cost, and leaving JEV_RULES empty runs the regex tier alone. JEV_RULES are the ones sent to Jev, batched into one call per tool call.

How does the CI twin differ from the editor plugin?

scripts/jev-lint.ts runs the same JEV_RULES over the added lines of a diff, one request per file. It is advisory by default and exits 1 only with --strict, and it skips silently when JEV_API_KEY is unset.

CommunityJev Agents & MCP

Jev Guard — muratcakmak

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.

See highlights

Open github.com

Stars: 7
Language: TypeScript
License: MIT
Published by: Community
Author: muratcakmak
Last verified: Oct 2026

muratcakmak/jev-guard

User repository on GitHub

View on GitHub

Probability-scored guardrails for Claude Code: deny rule-breaking edits and unasked-for deploys, route your docs into each prompt, and check the final answer against the turn's own evidence.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

This is one of several Jev listings for the same project. The canonical entry is Jev Guard — leepokai.

About this project

A Claude Code plugin that scores what the model is about to do — and what it just claimed — with a probability model, and denies the call when the score is high enough. Conventions written in CLAUDE.md are advice. The model follows them until the context gets long, then quietly stops. jev-guard moves the ones that matter out of the prompt and into hooks, where a broken rule is a denied tool call with the correction attached, not a review comment three days later.

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

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Routing Jev for Ranking Jev for Moderation

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

Rule guard checks Edit, Write, and Bash calls with local REGEX_RULES plus batched Jev JEV_RULES.
Deploy gate denies publish, merge, deploy, or credential-send commands unless the prompt asked for exactly that.
Answer check scores finished answers for unbacked claims, merged-claim errors, and hidden failures, then nudges once.
Every failure path passes the call through, so a missing key or scorer outage never blocks the agent.

Watch out

Ships example React rules meant to be replaced, sends edited files, commands, prompts, and answers to the Jev endpoint, and the doc router and subagent routing stay off by default. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-test-filter

Diff-aware test selector that sends the git diff to Jev with one question per test, then prints the native filter flags and hands them to vitest, jest, node, bun, Playwright, cargo, or go test.

Jev Agents & MCP#community#jevbest#catalog-import

Community

mizchi/jev-test-filterTypeScript

pijev

Terminal coding agent on Pi where Jev shortlists skills, ranks retrieved excerpts, and triages tool failures while the coding model handles reasoning and edits.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

tonyzdev/pijevTypeScript

typesafe-mod

A Claude Code function hook that sends AskUserQuestion dialogs to Jev and draws a probability per option above the dialog, so you still choose while seeing how confident the model is.

Jev Agents & MCP#community#jevbest#catalog-import

Community

BeLazy167/typesafe-modTypeScript

actiongate-jev

A runtime authorization gateway for agent tool calls: deterministic policy plus six narrow Jev questions produce evidence, then a single-use grant bound to the exact approved action is consumed at execution.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

omkarghugarkar007/actiongate-jevTypeScript

jev-code

TypeScript SDK and CLI coding agent where Jev only picks options while your code runs effects, building programs as syntax trees.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

rhighs/jev-codeTypeScript

pi-jev-router

Pi extension that classifies each task with Jev over OpenRouter and routes it to a Pareto-optimal model balancing measured quality, cost, and latency under a role policy.

Jev Agents & MCP#community#jevbest#catalog-import

Community

philippdubach/pi-jev-routerTypeScript

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai