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

cursor-jev

Cursor plugin whose hooks ask Jev whether subagent spawns, file reads, writes and shell commands are worth taking.

Pluginbeta
View on GitHub ↗

Overview

cursor-jev routes Cursor subagents with Jev and lets any agent ask typed Choice, Score and Noul questions through MCP tools (jev_judge, jev_ask, jev_route, jev_auth). Jev never writes code; it decides whether an expensive hop is worth taking, answering in about 100 ms according to the README.

Hooks intercept Task, Read, Write and shell calls and return allow, deny, ask or rewrite. The README describes three savings: fewer extra subagents (local edits stay on the current agent, a second Task in one turn is blocked, nested subagents are denied), fewer extra file reads (after Grep or Glob, Jev keeps about three relevant files), and less scope creep (writes the user did not ask for are denied, and destructive shell commands ask the user). It is honest that it is almost useless on one-file tweaks and that hooks fail open without a key.

What the decision model decides

TypeQuestionAnswers
Yes / noIs spawning this subagent worth it?(inferred)yes · no
ChoiceWhich of these files are relevant to read?(inferred)—
Yes / noIs this write within the scope the user asked for?(inferred)yes · no

How it works

  1. Cursor agent proposes Task, Read, Write or shell
  2. Hook sends the context to Jev
  3. Jev's answer becomes allow, deny, ask or rewrite
  4. Agents can also call jev_judge/jev_ask/jev_route via MCP

Strengths

  • Fewer costly subagent hops and file reads
  • Scope gate against unrequested refactors
  • Zero runtime npm dependencies

Caveats

  • Can block reads or writes you needed
  • Hooks fail open without a TypeSafe key
  • Little value on small edits

Getting started

git clone https://github.com/mywwave/cursor-jev.git, cd cursor-jev, node bin/cli.mjs install, then restart Cursor.

At a glance

Type
Plugin
Status
beta
Language
JavaScript
Runtime
Cursor plugin (hooks + MCP server)
Has code
yes

Use cases and integrations

How curated lists describe it

Frequently asked questions

Can I use Jev as a Cursor model?

No, Jev only classifies and scores; Cursor LLMs still write the code.

What does it save?

Extra subagents, unnecessary file reads, and out-of-scope writes.

What happens without a key?

Hooks fail open, so the agent runs as usual.

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-conductor-router

    MCP server and CLI where Jev decides if a coding task needs a Conductor cloud workspace, how hard it is, and which harness and model run it.

    ProjectbetaTypeScript
  • Jev-chooses-a-LLM

    Jev Router for Cursor: Jev picks a COST, BALANCED or INTELLIGENCE tier per task, with soft routing in chat and hard routing via the SDK.

    ProjectprototypeTypeScript
  • Gatekeeper

    Claude Code hook that decides which agent or skill should handle a request using YAML rulebooks plus one Jev Choice with confidence bands.

    HookbetaPython
  • 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-codex-router-skill

    Portable Codex skill that asks Jev to choose a model and reasoning tier for independent subtasks, with confidence gating.

    Agent skillbetaPython