Why not just use a normal linter?

Its rules need judgment rather than regex, such as whether a test can actually fail or whether client-sent user IDs are trusted; deterministic linters and human review still handle everything else.

Does it block the agent?

By default the check is advisory and a failed check lets the edit through unchanged; an optional before-the-write mode can hold an edit briefly, which the README says blocked 17 of 24 runs without deadlocks.

What happened when the hook ran on real agent tasks?

Violations per task fell from 2.38 to 1.21 across 96 Claude Code runs and from 1.67 to 0.92 across 48 Codex runs, both statistically significant. A second 48-run round with all 170 rules active showed the same direction, 2.71 to 1.54, from a higher starting point because more rules apply.

How accurate is the check on edits it was never tuned on?

The check itself scored 94-100% F1 on held-out edits in about 0.3 seconds, and across the packs held-out precision was 91-100% with 0-4% of clean edits getting any flag. The pattern gate is why that stays fast: only about a third of the rules are asked about on any given edit.

Should I add more rules to catch more mistakes?

Adding unrelated rules up to 100 per request left target-rule F1 at 97-99%, 99% with no padding, but median latency went from 319 ms to 458 ms, and splitting rules into parallel requests was slower and used 37% more tokens. The suggested 8-12 repo rules per language is about false alarms and the agent's attention, not Jev's capacity.

What does it cost to run, per edit and per task?

About $1.51 per 10,000 edits for the checks. Fixing a violation at the edit costs roughly another $0.04 and a few seconds per task, against about $0.18 and 70 seconds for a review, fix, re-review round, and review comments covered by the rules fell 41% with the hook on. It is MIT licensed.

What are my options besides a TypeSafe key, and how do I install it?

bun src/setup.ts offers TypeSafe Jev as the default, Cloudflare Clef and OpenAI Decisions with gpt-6-luna, checking existing credentials without printing them and asking before it applies. Options include --claude-only, --codex-only, --project, --pre, --async and --skills. A local fallback chain needs roughly 20 GB of memory, where a Kev-4B scored 95.2% F1 against Jev's 98.1% on 2,089 held-out edits.

CommunityJev Tools & CLIs

Jev Lint — ckorhonen

Fuzzy linter hook for Claude Code and Codex: after the agent edits a file, it sends the added code to Jev, which answers Noul checks against team rule packs in about 0.3 seconds, letting the agent fix violations before code review.

See highlights

Open github.com

Stars: 11
Language: HTML
License: MIT
Published by: Community
Last verified: Oct 2026

ckorhonen/jev-lint

User repository on GitHub

View on GitHub

A fuzzy linter for coding agents. It checks the code your agent writes against your team's best practices while the agent is still working, not at code review.

Forks: 1
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

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

About this project

A fuzzy linter for coding agents. Your agent writes a file; 0.3 seconds later it hears which of your team's rules it just broke, and fixes them before anyone reviews the code. ▶ Watch the 40-second demo (with sound; square cut). The editor scene is an illustrative session; the numbers are from the notebook. 🌐 Website: jevlint.dev · 📓 Results and method: experiment notebook (source in report/). Set it up with your agent. Options (environment variables on the hook command) For agents working in this repo.

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

Tagscommunity linter claude-code code-quality

Works forJev for Code Jev for Verification

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: CLI
Capabilities: Open sourceRuns locallyRequires API key

Highlights

Ships 177 rules across TypeScript/React, Swift, Kotlin, Rust, Python, Ruby and Bazel plus security, test-hygiene and performance packs.
In its runs, violations per task fell 2.38 to 1.21 across 96 Claude Code tasks and 1.67 to 0.92 across 48 Codex tasks.
Held-out checks scored 94-100% F1 in about 0.3 s; the pattern gate asks about only a third of the rules per edit.
Optional local fallback: a Kev-4B scored 95.2% F1 versus Jev's 98.1% on 2,089 held-out edits.

Watch out

MIT licensed. Default operation needs TypeSafe credentials (Cloudflare Clef and OpenAI Decisions are alternatives), and the local fallback chain needs roughly 20 GB of memory.

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.

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.

Reddit: A Claude Code mod replaces the auto-mode classifier with Jev, 2x faster.

More Jev resources like this

Sedum

Open-source, pre-alpha CLI that runs plain-English browser tests on Playwright, using Jev for bounded decisions: resolving each step, picking elements among offered candidates and judging verify claims with probabilities. Bring your own TypeSafe key; MIT licensed.

Jev Tools & CLIs#community#e2e-testing#playwright

Community

sedum-dev/sedumTypeScript

semcheck

Go linter whose rules are plain-English questions: AST matchers select nodes and Jev answers one Noul per fragment above the rule's threshold. It runs standalone, under go vet, and as a golangci-lint plugin, with cached answers so runs only ask about changed code.

Jev Tools & CLIs#community#linter#go

Community

arturobermejo/semcheckGo

JevCI

CLI quality gate that asks four typed questions about each change - three Score lenses for commit quality, doc alignment and breaking-API risk, plus a Noul secret audit - and fails a diff, commit message or doc set below the resulting 0-100 score. Runs from a terminal, a git pre-commit hook or a GitHub Action.

Jev Tools & CLIs#community#quality-gate#ci-cd

Community

sumant1122/jevciJavaScript

jevgrep (code search)

2.6k

A CLI that asks Jev which files and declarations answer a repository question, then prints verbatim excerpts for a coding agent to use.

Jev Tools & CLIs#community#typescript#cli

Community

dzhng/jevgrepTypeScript

surf

634

A CLI browser automation tool with optional Jev semantic mode: Surf observes, Jev selects the next action, Surf validates and acts, Jev checks the goal.

Jev Tools & CLIs#community#browser#cli

Community

nicobailon/surf-cliTypeScript

Jev-Code-Reviewer

140

A local Chrome extension plus CLI that classifies each agent PR hunk as P0, P1, or P2 with Jev, showing P0 by default and human-readable diff summaries.

Jev Tools & CLIs#community#code-review#typescript

Community

egma-ai/jev-code-reviewerJavaScript

From the Jev guides

Original write-ups on what this entry is for.

Jev on the Gateways: Vercel, Workers AI, and OpenRouter

Oct 12, 2026

The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.

Jev Tools & CLIs#gateway#vercel#cloudflare

Read article

Exporting Decision Models: GGUF, ONNX, and Device Runtimes

Oct 10, 2026

Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.

Jev Tools & CLIs#open-models#local#onnx

Read article

Records at Rest: Documents, Inboxes, and Risk Queues

Sep 28, 2026

Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.

Jev Tools & CLIs#classification#email#security

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai