Jev Model Router
Route subagent models and main-loop effort with Jev.
Illustration. Actual results depend on the mod, terminal and environment.
Install
Installation is unusual; follow the original author’s instructions
Read the author’s instructions ↗What it accesses
Based on the author’s documentation, checked when listed; use claude plugin validate to check for yourself.
Reads
YesReads prompts and subagent descriptions for classification.
Writes files
NoneRuns commands
NoneNetwork
YesWith a key, calls TypeSafe or Vercel AI Gateway; with none, no network.
Introduction
jev-model-router is a Claude Code function-hooks Mod that uses TypeSafe Jev (or Vercel AI Gateway / builtin classifier) to score task tier, effort, and risk, routing subagent models at agent.spawn and main-loop effort at turn.step (main-model routing off by default to protect cache). It can show decisions on the status line. With a key, prompt text goes to that backend; with none it uses the builtin classifier offline. Install via npx claude-code-templates --mod productivity/jev-model-router. Needs Claude Code 2.1.287+. Scores are document-calibrated, not runtime-tested.
Who it is for
- People who want automatic cheaper/stronger model and effort routing per task
Requirements
- Claude Code ≥ 2.1.287
- Optional TypeSafe or Vercel AI Gateway API key; builtin classifier without a key.