User repository on GitHub
Claude Code plugin that picks the right model and effort for each task, using TypeSafe Jev judgments.
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Sommelier
A Claude Code plugin that routes each request to the right model and effort — Haiku for quick lookups, Sonnet for well-specified changes, Opus for investigation or design — by asking TypeSafe Jev three questions about the request: reasoning depth, whether it starts a new task, and whether it is high-risk.
- Stars
- 7
- Language
- TypeScript
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- wide3751
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- TypeScript
- Platforms
- CLI
- Capabilities
- Open source
Highlights
- Jev is asked three questions per new request: how much reasoning it needs, whether it starts a new task, and whether it is high-risk.
- Answers map onto a configurable model/effort ladder; Jev returns probabilities for the classification.
- Default mapping: quick lookups to Claude Haiku, well-specified changes to Sonnet, investigation or design to Opus.
- Built on Claude Code mods, using the mods API's turn.step.
- Stated goal: save Claude Code usage without giving up quality; READMEs ship in five languages.
Watch out
README reviewed 2026-10-17; the digest cuts off inside the reasoning-depth ladder table, so the higher rungs, thresholds and configuration format are not captured, and the README reports no measured savings. The Haiku/Sonnet/Opus choices are the plugin's own defaults.
Questions about this entry
Short answers drawn from the linked project or page.
A Claude Code plugin — built on Claude Code mods via the mods API's turn.step — that picks the model and reasoning effort for each task automatically, the way a sommelier picks a wine for each dish.
By default: quick lookups go to Claude Haiku, well-specified changes to Claude Sonnet, and investigation or design work to Claude Opus. The mapping is described as a configurable ladder, so the rungs can be retuned.
For each new request it asks TypeSafe Jev three questions — how much reasoning it needs, whether it starts a new task, and whether it is a high-risk operation — and maps the returned probabilities onto the ladder.
The README's goal is saving Claude Code usage without giving up quality: instead of running everything on the strongest model, routine work lands on cheaper rungs automatically. No measured savings numbers appear in the captured text, so the claim is unquantified here.
A Claude Code environment that supports mods, and Jev access for the three classification questions. The digest does not show the install steps or how the TypeSafe key is configured; the README has translations in Traditional and Simplified Chinese, Japanese and Korean.
The visible table row defines L0 as a lookup or mechanical edit mapped to haiku at medium effort, with checking the version in package.json as the example. The digest truncates the table before the higher rungs.
Wide3751 built it. Metadata: TypeScript, MIT, 6 stars, 0 forks, created 2026-10-13 and pushed 2026-10-16. The README links a Ko-fi page for support, which is separate from any licensing.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
More Jev resources like this
A Claude Code mod that scores each prompt with TypeSafe Jev and sets Claude's effort level for that turn, mapping four narrow questions, reasoning depth, scope, stakes and ambiguity, onto an effort level without ever touching the model.
Jev Tools & CLIs#community#github#claude-code
A Claude Code mod that routes each request to haiku, sonnet or opus with a reasoning effort and confidence score using local heuristics, and can optionally let Jev (TypeSafe System One) pick the tier instead. Adds a live status band above the prompt.
Jev Tools & CLIs#community#github#claude-code
Community
visualife/ClaudeRouterTypeScript
visualife/ClaudeRouterTypeScriptA Claude Code mod that asks Jev (TypeSafe) how much effort each prompt needs and sends every request of that turn at that effort, never changing the main loop's model. Off by default; a fork of jjjjjjjjjjjjjjjjacob/jev-router.
Jev Tools & CLIs#community#github#claude-code
Community
vntrungld/jeffortTypeScript
vntrungld/jeffortTypeScriptA Claude Code plugin whose /fc command shrinks old tool output in about a second: it keeps the start and end, saves every cut to a file Claude can read back, and an optional Jev pass picks which outputs to keep whole. No API key needed; /compact stays untouched.
Jev Tools & CLIs#community#github#typescript
Community
nodardavituri/fast-compactTypeScript
nodardavituri/fast-compactTypeScriptA TypeScript/Bun launcher that routes coding-agent user turns through JEV, the TypeSafe classifier, with model and reasoning-effort selection: codex-jev wraps the Codex CLI, claude-jev wraps Claude Code, and a Pi extension adds a jev/auto virtual model.
Jev Tools & CLIs#community#github#cli
A Claude Code plugin that asks Jev, TypeSafe's System One model, which model tier each prompt needs — haiku, sonnet or opus — and switches automatically, and routes classification, scoring and yes/no judgments to Jev via scripts/jev.py.
Jev Integrations & Adapters#community#github#claude-code
Community
ricky94511/jevTypeScript
ricky94511/jevTypeScriptFrom 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.