What is the Jev funnel?

A shape for bulk judging: code collects items such as reviews, logs, diffs, tickets or warnings, jev.py sends them all in bulk with typed yes/no, pick-one or rubric questions, confident answers are handled in code, and only the uncertain band goes to Claude or a human.

How does Jev fit into this?

Jev (System One) takes a state plus typed questions and returns calibrated probabilities — a noul yes/no, a choice from fixed options or a score on a rubric. The README calls it closer to a programmable if for judgments plain code cannot make.

What do I need to run it?

Claude Code to use the skill inside a session (not required for the CLI), Python 3.9+, and a key: an OpenRouter API key with prepaid credit as the default transport, or a TypeSafe key with --provider typesafe.

How fast and expensive is a batch?

In the repo's own live benchmark, batch judged 500 synthetic items with two questions each in 13 requests: a packed 40-item request took about 2.3 s at p50, the run took 3.92 s and cost $0.0061. List price is about $0.042 per million input tokens with output free.

Why batch instead of one call per question?

All questions in one request see the same state and run in parallel. The README cites TypeSafe's parallel-questions cookbook, where 13 questions in one request took 0.27 s and ran about 12x cheaper and 10x faster than one call per question with the same answers.

What is the cited A/B result?

On a bounded judging task, Jev's median answer took 378 ms against 3,554 ms for Claude Sonnet 5 at high reasoning, at 100% versus 99% accuracy over 306 decisions, per the Greenberg write-up the README cites.

How do I install and invoke it?

Add the marketplace and install the plugin with claude plugin marketplace add and claude plugin install jev-funnel@claude-jev-funnel, or the /plugin equivalents inside Claude Code. The skill is then invocable as /jev-funnel:jev.

Jev directory

CommunityJev Agents & MCP

Claude Jev Funnel

A Claude Code plugin and zero-dependency Python CLI that judges items in bulk with calibrated Jev answers, resolves the confident ends in code, and sends only the uncertain band to Claude or a human.

See highlights Open github.com

Stars: 2
Language: Python
License: APACHE-2.0
Reading time: 3 min
Published by: Community
Author: shakibuzzaman3104
Last verified: Oct 2026

shakibuzzaman3104/claude-jev-funnel

User repository on GitHub

View on GitHub

Claude Code plugin + zero-dependency CLI for TypeSafe's Jev: judge items in bulk with calibrated yes/no, pick-one and rubric answers; handle the confident ends in code, review only the uncertain band.

Forks: 0
Last pushed: Updated Sep 2026

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

Tagscommunity catalog-import logicrw claude-code batch triage

Works forJev for Classification Jev for Decision gating

Used inJev for Software & DevTools

Languages: Python
Platforms: CLI OpenRouter
Capabilities: Open sourceRequires API key

Highlights

One request judges many items: all questions in a batch see the same state and run in parallel, and usage.cost on each response reports the real spend.
Repo benchmark: batch judged 500 synthetic items with two questions each in 13 requests, 3.92 s wall time and $0.006094 total.
Cited A/B: Jev's median answer 378 ms versus 3,554 ms for Claude Sonnet 5 at high reasoning, 100% vs 99% accuracy over 306 decisions.
Confident answers — YES/NO/ACT/CONFIDENT — are handled in code with no further model call; UNCERTAIN/REVIEW/ABSTAIN/UNSURE go to Claude or a human.
Python 3.9+ CLI is standard library only and needs an OpenRouter key with prepaid credit by default, or a TypeSafe key with --provider typesafe.

Quickstart

bash

claude plugin marketplace add Shakibuzzaman3104/claude-jev-funnel
claude plugin install jev-funnel@claude-jev-funnel

Watch out

Copy reviewed against the truncated repo README on 2026-10-15; the excerpt cuts off inside the install section. The latency, accuracy, batching and cost figures are the README's own benchmark and a cited external A/B, not an independent evaluation.

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.

Instagram: A claim triage tool that routes suspicious claims and never denies one.

Instagram

X: Intent-based search in Gmail

X: 400 companies matched to one candidate

X: Watching 25 customer groups on WhatsApp

From the Jev guides

Original write-ups on what this entry is for.

MCP tool routing: Jev in front of the tool call

Oct 16, 2026

Tool menus can outgrow the context that reads them. The catalog's Jev-based routers and gates narrow the list: benchmarked tool search, per-tool scoring, fail-closed approvals, and WebMCP in the browser.

Jev Agents & MCP#mcp#tool-routing#agents

Read article

Approval policies for coding agents: hooks, guards, and gateways