Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

better-call-jev

Claude Code plugin that sends judgment calls to Jev, acts at 0.8+ for yes or 0.2- for no, and gathers more evidence in between.

Agent skillbeta
View on GitHub ↗Live demo ↗

Overview

better-call-jev is a Claude Code plugin that lets Claude decide judgment calls by a number instead of a hunch. Claude asks Jev a yes/no, pick-one or rate-it question about the evidence and acts on the calibrated probability.

Every boolean answer gets a verdict from two cut-offs: at or above 0.8 is yes, at or below 0.2 is no, and anything between is undecided, in which case Claude splits the question or adds evidence. The README's example shows a refund question at 0.31 leading Claude to ask the customer what they want instead of acting. Thresholds can be changed through environment variables. It is one file with zero dependencies on Node 18+, invoked with /jev or by asking Claude to use Jev, and can be made the default through CLAUDE.md. It reads JEV_API_KEY and supports Vercel AI Gateway.

What the decision model decides

TypeQuestionAnswers
Yes / noIs the customer asking for a refund?yes · no · undecided
ChoicePick-one judgment over evidence—
ScoreRate-it judgment over evidence—

How it works

  1. Claude hits a judgment call or the user types /jev
  2. Skill sends evidence and question to Jev
  3. Verdict: yes at 0.8+, no at 0.2-, else undecided
  4. Claude acts, or splits the question and gathers evidence

Strengths

  • Explicit thresholds with an undecided band
  • Zero dependencies
  • Configurable risk tolerance

Caveats

  • Claude Code only
  • Needs a Jev key or AI Gateway

Getting started

/plugin marketplace add jukkatupamaki/better-call-jev then /plugin install better-call-jev@better-call-jev; set JEV_API_KEY.

At a glance

Type
Agent skill
Status
beta
Language
JavaScript
Runtime
Claude Code plugin (Node 18+)
Has code
yes

Use cases and integrations

How curated lists describe it

  • A skill for integrating Jev model calls to any Claude Code session. Supports Vercel AI Gateway.

    hellogumbo/awesome-jev

Frequently asked questions

What happens between 0.2 and 0.8?

The verdict is undecided and Claude splits the question or adds evidence.

Can I change the thresholds?

Yes, via environment variables.

How do I make it the default?

Say so in your CLAUDE.md.

Related projects

  • SkillRanker★ 134

    Rust CLI that ranks which agent skills fit the next step from live session context with Jev, can abstain, and hooks into Claude Code.

    CLIbetaRust
  • augustus

    Agent skills for placing, building, evaluating and improving decision-model systems, with Jev as the default hosted example.

    Agent skillbeta
  • jev-agent-skill

    Portuguese-language agent skill for millisecond Jev decisions via OpenRouter, with validation, an MCP mode and calibration evals.

    Agent skillbetaJavaScript
  • jev-mcp

    Eval-first MCP server for Jev: prototype questions, map them over many items, and measure accuracy and thresholds on labelled data.

    MCP serverbetaTypeScript
  • tenbin

    MCP server and agent skill for designing Jev judgments: question linting, batch evaluation, threshold calibration and reranking.

    MCP serverbetaTypeScript
  • typesafe-mcp (MarkChu-git)

    Bun-based MCP server wrapping Jev with compact answers and a code-computed act, review or abstain verdict.

    MCP serverbetaTypeScript