jev-pilot

Claude Code plugin that asks Jev before each turn for reasoning effort, subagent model, strategy and skill, with a decision log.

Agent skillbetaMIT

View on GitHub ↗

Overview

jev-pilot is a Claude Code plugin that asks TypeSafe's Jev a few typed questions about each prompt before the turn and sets the turn up from the answers, while you keep Opus for the conversation. Decisions include reasoning effort from low to xhigh at turn start, raising effort up to max once mid-turn when tool calls keep failing, the model and effort for subagents (Haiku, Sonnet or Opus), an advisory strategy (do it directly, delegate, parallelise or plan a graph), and the one skill the prompt needs, if any.

Every decision is recorded with tuning suggestions available through a report command. Jev never writes in the conversation; its decisions are shown by an animated pet above the prompt. It works through OpenRouter or TypeSafe, and the README covers custom models and other agents. Tests run in CI.

What the decision model decides

Type	Question	Answers
Choice	What reasoning effort should this turn use?	low · medium · high · xhigh
Choice	Which model and effort should this subagent use?	Haiku · Sonnet · Opus
Choice	What strategy fits: direct, delegate, parallel or plan a graph?	direct · delegate · parallel · graph
Choice	Which skill, if any, does this prompt need?	—

How it works

Before each turn, send the prompt to Jev with typed questions
Set effort, suggest strategy and pick a skill
On subagent start, pick its model and effort
Raise effort once mid-turn if tool calls keep failing
Log decisions and offer tuning suggestions

Strengths

Easy work runs cheaper; hard work gets more thinking
Decision log with tuning report
Works with OpenRouter or TypeSafe

Caveats

Requires Claude Code 2.1.278 or newer
Strategy is advisory only
Licence listed differently across sources (badge shows MIT)

At a glance

Type: Agent skill
Status: beta
Language: TypeScript
Runtime: Claude Code plugin
Has code: yes
Has tests: yes
License: MIT

Use cases and integrations

Classification & Routing306

Claude Code373

How curated lists describe it

Let Jev steer Claude Code: the right reasoning effort, subagent model and skill for every prompt. A Claude Code plugin powered by TypeSafe's Jev (OpenRouter / TypeSafe).
logicrw/awesome-jev-projects
Claude Code plugin that asks Jev before each turn for the reasoning effort, subagent model and skill to use, with a decision log, tuning suggestions and an animated pet that says what Jev decided.
MrJev/awesome-jev

Frequently asked questions

What does Jev decide?

Reasoning effort, subagent model, an advisory strategy and which skill to use for each prompt.

Does Jev write in the conversation?

No. Its decisions are shown by an animated pet above the prompt.

Can I check if it's paying off?

Yes. /jev-pilot:report shows every decision with tuning suggestions.

Related projects

Helm
Orchestrator that probes which coding agents are installed and logged in, then has Jev pick the best one for a task.
Agent skillbetaPython
jev-agent-hooks
Hooks for Claude Code, Codex and pi that use Jev to suggest a skill each turn, plus a rule gate for subagent model routing.
HookbetaJavaScript
jev-for-all
Experimental plugin wiring Jev into coding-agent harnesses to pick skills, tools, models and browser moves, starting with OpenCode.
Agent skillprototypeTypeScript
jev-harness-router
Per-turn router: one ~350 ms Jev call picks model tier, effort, tools and skill, with a hard deadline and regex fallback.
Agent skillbetaTypeScript
loadout
Chooses the model tier, skills and MCP servers for each coding-agent session with two Jev requests, then launches Claude Code or Codex.
MCP serverbetaPython
JevRouter
Router that puts models, subagents, skills, MCP tools and CLIs in one candidate set and asks Jev one Choice about which handles the step.
MCP serverbetaTypeScript