User repository on GitHub
프롬프트를 jev로 판단해 알맞은 모델을 고르는 Claude Code / Codex 스킬. 의존성 0개, 능력치는 공개 벤치마크 출처 명시.
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Agents & MCP
Jev Router — heyman333
A Claude Code and Codex skill with a CLI that answers narrow agent judgments - is this file relevant, which team owns this request, how severe is this - with Jev's calibrated probabilities instead of another LLM call. Zero dependencies, MIT.
- Stars
- 1
- Language
- JavaScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- heyman333
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Router — gargpratyush.
Used inJev for Software & DevTools
- Languages
- JavaScript
- Platforms
- CLI
- Capabilities
- Open sourceRuns locallyRequires API key
Highlights
- One jev call can carry several questions: --bool, --pick and --score are judged together, billed once at $0.042 per 1M input tokens with output free.
- Batch mode reads stdin line by line, keeps input order and emits JSONL, so an agent can narrow 184 files down to the 12 that matter.
- Confidence under 60% prints a warning and the skill stops to show candidates or read the item itself instead of silently proceeding.
- The README documents TYPESAFE_API_KEY from console.typesafe.ai/keys; a Vercel AI Gateway vck_ key does not work and returns a 401.
- The made-with-Jev listing describes it as picking the model for a prompt and routing there; heyman333 also built jev-model-classifier.
Quickstart
bash
jev "결제가 3일째 실패하고 있어요. 급합니다." \
--bool urgent "시간에 쫓기는 요청인가?" \
--pick team "어느 팀이 맡나?" billing=결제·환불 technical=버그·장애 sales=가격 \
--score severity "얼마나 심각한가?" 사소함 불편함 업무정지Watch out
Grounded on the truncated README, the made-with-Jev listing blurb and GitHub metadata reviewed 2026-10-15; the README is written in Korean and the listing text is cut off mid-sentence. Pricing and latency figures (about $0.00004 per case, 300-500 ms) are the project's own.
Questions about this entry
Short answers drawn from the linked project or page.
A Claude Code and Codex skill with a zero-dependency CLI that answers narrow agent judgments - is this file relevant, which team owns this request, how severe is it - with Jev's typed values and calibrated probabilities instead of LLM calls.
The CLI sends one Jev request per judgment set with --bool, --pick and --score questions. The README notes several questions in one call cost one request, roughly $0.00004 per case at $0.042 per 1M input tokens with free output.
Clone the repository and run npm test to check it with no dependencies, then npm link for the jev command; if you skip the link you can run node bin/jev.mjs instead, and the skill copies into ~/.claude/skills/jev.
A TypeSafe API key from console.typesafe.ai/keys exported as TYPESAFE_API_KEY in ~/.zshrc or ~/.bashrc. The README warns that a Vercel AI Gateway vck_ key does not work because it is issued elsewhere and returns a 401.
It reads one input per line from stdin, asks the same question for each, keeps input order and emits JSONL. The README example filters git ls-files output, and a failing line only becomes an error object, leaving the rest alive.
For --pick and --score it reports confidence and warns below 60 percent. The skill then shows the user the candidates or reads that item itself instead of quietly taking the top choice; the README contrasts this with LLMs asserting even when unsure.
Per the README: Jev cannot write or fix code, cannot explain why it judged something, sees only text you pass it, and is not worth the round trip for one-off judgments - the benefit grows with repetition.
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
Prompt reviewer for Claude Code and Codex CLI: Jev first decides whether a submitted prompt needs a review, and when it does, a lightweight model flags overlooked context, unstated assumptions, or the real problem in a local web UI.
Jev Agents & MCP#community#catalog-import#logicrw
Community
jsk4581/jev-blindspotTypeScript
jsk4581/jev-blindspotTypeScriptA Codex plugin that prepares traceable context handoffs: Jev answers one narrow classification question per old, verified read-only tool pair, and code validates the answers, preserves protected content and builds a JSON or Markdown packet.
Jev Agents & MCP#community#catalog-import#logicrw
An agent skill for Claude Code, Codex and other coding agents, with a standard-library Python CLI, 5 tested question sets and findings from 19,367 Jev calls and 90.4M metered input tokens over 4 days.
Jev Agents & MCP#community#catalog-import#logicrw
Community
marcodicesare-dev/jev-skillPython
marcodicesare-dev/jev-skillPythonA local MCP gateway that connects your MCP servers once and exposes just two tools to agents: find_tool discovers a capability through Jev Choice questions and execute_tool runs it with arguments validated against the original schema, managed from a CLI and a local dashboard.
Jev Agents & MCP#community#catalog-import#madewithjev
Community
prasanth263/mazaJavaScript
prasanth263/mazaJavaScriptA free Claude Code skill that puts Jev in front of the expensive model: thirteen commands sort, check, score, gate and verify with Jev, so Claude reads only the unsure slice and writes the output. Needs python3 and an OpenRouter key.
Jev Agents & MCP#community#python#claude-code
Community
charlesdove977/claude-x-jevPython
charlesdove977/claude-x-jevPythonOpen-source Codex plugin that consults TypeSafe Jev about tools, models, tasks, skills, context, and strategies, diagnoses failed commands, and checks completion claims against evidence. The current package is version 0.4.0.
Jev Agents & MCP#community#catalog-import#logicrw
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.