jev-claude-router
Claude Code plugin that picks model tier and effort per turn with Jev and weighs each switch against cost, with Jev-based compaction.
Overview
jev-claude-router is a Claude Code plugin that picks the model and effort for every turn with Jev and weighs each switch against what it costs. For every prompt, Jev answers two questions in one request: which tier should handle it (haiku for trivial lookups and renames, sonnet for straightforward work, and higher tiers for harder tasks) and how hard that model should think. The plugin then checks the answer against the conversation's context window, Jev's confidence, price and a ceiling before any request goes out, and rewrites each model request in the turn accordingly; a status line shows the tier, effort, Jev's confidence, any capping, latency, cost and cache hit rate.
It is a fork of satviksinha/jev-model-router, adding cost-aware routing, compaction by Jev built on tamaratran/fast-jev-compaction, session persistence and other changes. The README covers how a turn is routed, compaction, reliability, commands, configuration and tuning. It is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which tier should handle this prompt? | haiku · sonnet · opus · fable |
| Choice | How hard should the model think? | — |
How it works
- User types a prompt
- Jev returns tier and effort
- Plugin checks window, confidence, price and ceiling
- Rewrite each model request in the turn
- Show routing and cost in the status line
Strengths
- Cost-aware checks before switching
- Transparent per-turn status line
- Includes Jev-based compaction
Caveats
- Switching models can still affect caching
- Fork; upstream changes may diverge
At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- Claude Code plugin
- Has code
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Model router for Claude Code using Jev
logicrw/awesome-jev-projects
Frequently asked questions
What does Jev decide?
The model tier and effort for each turn.
How is cost considered?
Each switch is checked against the window, confidence, price and a ceiling before use.
Is it original?
It is an extended fork of satviksinha/jev-model-router.
Related projects
Claude Code mod that picks reasoning effort per prompt with a Haiku judge, optionally Jev, and shows cache and context state.
ProjectbetaTypeScriptClaude Code and Codex plugin where Jev judges each turn for model routing, delegation, evidence search and compaction.
ProjectbetaTypeScriptClaude Code mod using Jev to pick model and effort per turn, size subagents, and prune stale context verbatim instead of summarising.
ProjectbetaTypeScriptOpen-source AI gateway for 100+ LLMs that can use Jev to classify requests for complexity routing and to compact tool results.
ProxyproductionPythonHermes Agent plugin that picks the model each turn with a direct TypeSafe Jev Choice and enforces trim-then-compress context handling.
PluginbetaPython- 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