Yaxin9Luo/spending-effort-with-jev
User repository on GitHub
Jev-powered /effort advisor for Claude Code: tells you when to switch effort, per prompt
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Tools & CLIs
Spending Effort With Jev
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, then offers a one-key switch before Claude starts when the tip clears its confidence threshold.
- Stars
- 7
- Language
- Python
- License
- MIT
- Published by
- Community
- Last verified
- Oct 2026
About this project
Two Claude Code mods that live above your prompt. Install either or both; they share the band and stack. Both need Claude Code 2.1.287 or later, in the terminal or the desktop app. Know which /effort level each message needs, and switch with one key. Every message you type is read by Jev, which judges how much effort the task deserves. Before Claude starts, you get a one-line verdict against the level Claude is about to run on, and when that level looks wrong, one key switches it.
Quoted from the project's README. Read the full README.
Highlights
- Judges every prompt's /effort level with Jev and offers a one-key switch; a switch lasts one turn
- Subagents are sized separately (never above high), and a mid-turn downgrade can offer low for the rest
- A fuzzy-spec Noul triggers a two-or-three-question interview before a long hands-off run
- On the held-out half, 90% of switch tips named the right level and 98% pointed the right way
- A ledger pane (/effort-ledger) tracks turns per level, how often you followed Jev, and Jev's API cost
Watch out
MIT license. Needs Claude Code 2.1.287+ and a TypeSafe API key, and the README notes the published evaluation ran on the v0.2 Python hook over a 160-message set.
Questions about this entry
Short answers drawn from the linked project or page.
No. Claude Code has no plugin API to change the setting, so the mod rewrites the level on the requests of that turn only; the control under the input box keeps showing your setting.
On a held-out half of 160 labeled messages, 90% of switch tips named the right level and 98% pointed the right way. When Jev's answer is split, the status line says so instead of advising.
On Claude Code 2.1.287 or later, in the terminal or the desktop app, run /plugin marketplace add Yaxin9Luo/spending-effort-with-jev then /plugin install spending-effort-with-jev@spending-effort-with-jev. Paste your TypeSafe API key when Claude Code asks; it is kept in secure storage and the plugin reads it from nowhere else. Finish with /reload-plugins. Jev is a hosted API, so nothing else has to be installed.
It decides who waits. With ask_first on, a dialog asks before Claude starts whenever the level looks wrong in either direction, and the whole turn then runs on the level you pick. With it off, nothing is held up: a band above the prompt offers Switch, Keep or Close while Claude works, and the switch applies from Claude's next request. Set it with --config ask_first=true at install time or later under /plugin.
The band shows Jev's API cost for the day, and the ledger at /effort-ledger lists calls and cost today and over the last seven days, turns per level, and how often you ran the level Jev named. The figure is priced at TypeSafe's list rate of $0.042 per million input tokens with output free, and Claude's own token spending is deliberately not counted.
Each subagent's task is sized separately by Jev and runs on that level, capped at high so max stays yours to choose, and subagents started together are announced in a single toast. Mid-turn, Jev is asked whether the rest of a long high or max turn looks mechanical, such as applying a decided change or writing the commit, and the band can offer low for the remainder of that turn only.
A Noul on whether the spec is fuzzy fires first, and the plugin runs a two or three question interview drafted by the small model from your message. Your answers are passed to Claude alongside the run, then max is offered for it as any switch is; Up to Claude skips a question and Start now the rest. With the interview off you get a warning line and a toast instead, on the reasoning that more effort will not fix a wrong reading of the task.
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
21GitHub stars
MIT-licensed Python tool router for coding agents that keeps obvious calls on a fast path, asks Jev (default typesafe/jev-1.13 via OpenRouter) for genuinely ambiguous choices, and can run bounded MCTS when multi-step consequences matter. Ships a CLI, MCP server and hook integrations.
Jev Tools & CLIs#community#routing#mcp
0GitHub stars
Claude Code PreToolUse hook that asks Jev one Noul on whether a Bash command is strictly read-only and auto-approves at p of 0.95 or higher. It can only allow or stay silent, never deny, so unknown or risky commands fall back to Claude Code's normal permission prompt.
Jev Tools & CLIs#community#claude-code#hooks
39.8kGitHub stars
A teams-first multi-agent orchestration plugin for Claude Code with opt-in Jev hooks for sub-millisecond judgment points, decision caching and per-point egress controls. TypeScript, MIT, with a staged team pipeline and tmux CLI workers.
Jev Tools & CLIs#community#claude-code#orchestration
13GitHub stars
Claude Code plugin and CLI that runs Claude Code on Opus 5.5 and has Jev re-decide the reasoning effort when a prompt arrives and after every tool batch, sending each change as a per-message effort statement inside one unbroken prompt cache.
Jev Tools & CLIs#community#claude-code#routing
Community
WXK-AI/jev-opusTypeScript
WXK-AI/jev-opusTypeScript8GitHub stars
Supervisor for unattended coding agents: it tracks cross-turn state and asks one batched Jev Choice whether a trajectory should continue, pause or escalate, while deterministic code owns dangerous-command blocking, budget caps and Ed25519-signed receipts.
Jev Tools & CLIs#community#guardrails#safety
Community
VeridicalTech/EdwardPython
VeridicalTech/EdwardPython4GitHub stars
Python harness for GUI grounding: it divides a screenshot into grid tiles, has a vision model describe each tile neutrally, then asks typesafe-ai/jev to choose among the descriptions. Local confidence and margin gates decide whether to descend or refuse, and it emits only a point and bounding box.
Jev Tools & CLIs#community#gui-grounding#computer-use
Community
ZihuaEvan/GUI_JEVPython
ZihuaEvan/GUI_JEVPythonFrom the Jev guides
Original write-ups on what this entry is for.
The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.
Jev Tools & CLIs#gateway#vercel#cloudflare
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.
Jev Tools & CLIs#classification#email#security
Read article
From the Jev community
Posts from builders shipping with Jev right now.