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

spending-effort-with-jev

Claude Code mod where Jev judges each message's /effort level and offers a one-key switch for that turn.

PluginbetaMIT
View on GitHub ↗

Overview

spending-effort-with-jev is a Claude Code mod that sizes how much effort each message deserves. Every message you type is read by Jev, which judges the appropriate /effort level; before Claude starts you get a one-line verdict against the level Claude is about to use, and if it looks wrong one key switches it for that turn. With ask_first on, a dialog asks first; otherwise a band above the prompt offers the switch while Claude works, applied from the next step. The repo also includes an unrelated companion mod, a context-window progress bar.

Version 0.3 adds per-turn switches, separate effort for subagents (capped at high), mid-turn downgrade when the rest of a long turn looks mechanical, a short spec interview before long hands-off runs that leave requirements open, adaptation to how often you accept switches, and a ledger of turns per level and Jev's API cost. A list summary describes the hook as a Choice over low, medium, high, max or unclear plus a Noul on whether a hands-off request has a fuzzy spec, showing a tip only at 0.7 confidence or above, with 95% of tips pointing to the right level on a three-rater held-out set. Split votes produce no advice.

What the decision model decides

TypeQuestionAnswers
ChoiceWhat /effort level does this message need?low · medium · high · max · unclear
Yes / noDoes this hands-off request leave key requirements open (fuzzy spec)?—

How it works

  1. User submits a message
  2. Jev judges the effort level and spec clarity
  3. Show a verdict; offer a switch only when confident
  4. Apply the switch for that turn; log to the ledger

Strengths

  • Switches only with the user's OK
  • 95% of tips correct on a held-out set per the list summary
  • Tracks Jev's own API cost

Caveats

  • Needs Claude Code 2.1.287+ and a TypeSafe key
  • Only works on models with effort levels

Getting started

/plugin marketplace add Yaxin9Luo/spending-effort-with-jev then /plugin install spending-effort-with-jev@spending-effort-with-jev and paste a TypeSafe key.

At a glance

Type
Plugin
Status
beta
Language
Python
Runtime
Claude Code mod
Has code
yes
License
MIT

Use cases and integrations

How curated lists describe it

  • Coding agents: Claude Code plugin whose UserPromptSubmit hook asks Jev a `Choice` over `/effort` levels (low / medium / high / max / unclear) plus a `Noul` on whether a hands-off request has a fuzzy spec, showing a switch tip before Claude starts only at 0.7 confidence or above, with 95% of tips pointing to the right level on a three-rater held-out set.

    yibie/awesome-jev

  • Jev-powered /effort advisor for Claude Code: tells you when to switch effort, per prompt

    logicrw/awesome-jev-projects

  • Claude Code plugin whose UserPromptSubmit hook asks Jev a `Choice` over `/effort` levels and a `Noul` on whether a hands-off request has a fuzzy spec, while code shows a switch tip only at 0.7 confidence and a one-level gap and never changes effort itself; a three-rater eval in the repo puts tip precision at 95%.

    cobanov/awesome-jev

  • Jev-powered /effort advisor for Claude Code: tells you when to switch effort, per prompt.

    hellogumbo/awesome-jev

Frequently asked questions

What does Jev decide?

The effort level each message needs and whether a long hands-off request has an underspecified spec.

Does it change effort automatically?

No; it switches only when you confirm.

What happens when Jev is split?

It shows 'not sure' and advises keeping your level.

Related projects

  • jev-router★ 593

    Per-turn model routing for Claude Code and Codex: wraps the real CLI and lets Jev send simple turns to fast models and hard ones to strong.

    ProxybetaJavaScript
  • Claude Model Router

    Claude Code mod that sets model and effort per turn and subagent from a prompt classifier: Jev by default, or Clef, OpenAI or Ollama.

    Proxybeta
  • coding-router-jev

    TypeScript/Bun launcher that asks Jev which model and reasoning effort each Codex, Claude Code or Pi turn should use.

    ProjectbetaTypeScript
  • decision-router

    CLI, Claude Code plugin and Pi virtual model that pick the right model per prompt from one Jev call with a complexity floor.

    PluginbetaTypeScript
  • effortless

    Claude Code mod that picks reasoning effort per prompt with a Haiku judge, optionally Jev, and shows cache and context state.

    ProjectbetaTypeScript
  • frost

    Go CLI model router that uses one Jev analysis to recommend a model, harness and effort level, then selects deterministically.

    CLIbetaGo