User repository on GitHub
Route to the cheapest model in claude code for your task using jev-router
- Forks
- 91
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Router — gargpratyush
Wrappers for Claude Code and OpenAI Codex that ask Jev to choose a model tier for each fresh user turn while preserving each CLI's native interface, tools, sessions, permissions, and authentication. Ships as the jev-claude and jev-codex commands.
- Stars
- 602
- Language
- JavaScript
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- gargpratyush
- Last verified
- Sep 2026
About this project
Automatic per-turn model routing for Claude Code and OpenAI Codex. Jev sends simple work to the fast tier and difficult work to the strong tier, while preserving each CLI's native interface, tools, sessions, permissions, and authentication. Both commands launch the real upstream CLI. Jev only chooses the model for a fresh user turn. Requires Node.js 20.12+ and at least one supported CLI: Claude Code or OpenAI Codex. Get a key from TypeSafe. Then launch either interface from any repository:
Quoted from the project's README. Read the full README.
Works forJev for Routing
Used inJev for Software & DevTools
Highlights
- Both commands launch the real upstream CLI; Jev only chooses the model for a fresh user turn.
- No Anthropic or OpenAI API key is required when the CLI is already logged in with a subscription.
- Shows the last turn's model in the status line, for example haiku p=0.98 with 8% context.
- Bundles a /jev-explain skill that renders the factors behind the last routing decision.
- Keeps up to 20 recent routing exchanges per session in mode-700 temp files, deleted after 7 days.
Quickstart
bash
npm install -g jev-router
echo "JEV_API_KEY=YOUR_API_KEY" > ~/.jev-router.env
jev-claude # or jev-codexWatch out
MIT-licensed. Needs Node.js 20.12+ and a logged-in Claude Code or Codex CLI plus a JEV_API_KEY; prompt text is sent to TypeSafe, and documented compatibility testing is on Windows with specific CLI versions.
Questions about this entry
Short answers drawn from the linked project or page.
Your prompt text goes to TypeSafe for the routing decision, and the README says nothing else does. Each command starts a loopback proxy and forwards the CLI's own authorization headers without reading, storing or modifying them. Recent routing exchanges, containing the prompt and the exact request and response, are kept in OS temp directories at mode 700 with files at 600 and deleted after seven days.
No. Routing is fail-open in the safe direction: failure, timeout or an unrecognised answer keeps the current model; an unavailable tier steps upward rather than picking something weaker; and low confidence never downgrades, instead capping upgrades at the balanced tier.
Selecting a concrete model in the picker pauses routing and selecting Jev Router resumes it. jev-claude restores Claude's previous default model on exit, an existing custom status line is preserved, and JEV_NO_STATUSLINE=1 disables the injected one. Only the first request of a turn pays the Jev latency; tool-loop continuations reuse the tier and add none.
One call per fresh user turn picks an abstract tier (fast, balanced, strong, long) that policy maps to concrete models in the signed-in account's catalogue. Explicit asks such as use opus or use strong win. The long tier stays off unless JEV_ALLOW_FABLE=1, and /jev-explain renders the factors from the saved request and response without re-scoring the prompt.
Four abstract tiers per CLI. Claude Code maps fast to Haiku, balanced to Sonnet, strong to Opus and long to Fable. Codex maps fast to gpt-5.6-luna, balanced to gpt-5.6-terra, strong to gpt-5.6-sol and long to gpt-6-astra. The long tier stays disabled unless JEV_ALLOW_FABLE=1.
Run /jev-explain in jev-claude or $jev-explain in jev-codex. It renders locally from the saved prompt, System One request and response, showing factors such as task complexity, reasoning required and the recommended tier, and does not ask Jev to score the prompt again.
It needs Node.js 20.12+ and Claude Code or OpenAI Codex. Install globally with npm install -g jev-router, then write JEV_API_KEY to ~/.jev-router.env; TYPESAFE_API_KEY also works. For a local checkout, npm link installs both the jev-claude and jev-codex commands.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
X: A group chat bot with decision room routing
Other listings of this project
Forks and copies of this project in the directory. This entry is the canonical one.
A Pi plugin that has Jev pick one model and one reasoning-effort level from a configured list at the start of a session, then keeps both pinned for the rest of it.
Jev Tools & CLIs#community#javascript#routing
Community
mejiasd3v/pi-jev-routerJavaScript
mejiasd3v/pi-jev-routerJavaScriptLocal pass-through proxy that asks Jev which tier each human turn needs — fast, balanced, frontier, or max — then forwards Claude Code and Codex CLI traffic to that model while streaming responses back unchanged.
Jev Agents & MCP#community#github
Community
dirien/jev-routerJavaScript
dirien/jev-routerJavaScriptAn OpenAI-compatible LiteLLM proxy where clients send one model id and Jev picks which configured model serves each request, with capability filtering and a cheapest-eligible fallback.
Jev Tools & CLIs#community#python#routing
Community
prismhq/jev-routerPython
prismhq/jev-routerPythonNode router that sends each query to the cheapest capable LLM tier by asking Jev to classify demand against your tier descriptions, with no LLM call spent on routing.
Jev Agents & MCP#community#jevbest#catalog-import
Community
rajdhakad9826/jev-routerTypeScript
rajdhakad9826/jev-routerTypeScriptLaunchers for Claude Code, Codex, Grok CLI, and opencode that ask Jev which model-and-effort tier each human turn needs, rewriting only the model field while the real CLI runs unchanged.
Jev Agents & MCP#community#javascript#routing
Community
flaviusapop/jev-routerJavaScript
flaviusapop/jev-routerJavaScriptA Claude Code plugin that picks Haiku, Sonnet or Opus and a reasoning effort for every prompt, step and subagent, with an effort cap, a prompt-cache guard and a status band; the repo is named jev-router and tagged typesafe.
Jev Integrations & Adapters#community#github#typescript
Community
dominicrico/jev-routerTypeScript
dominicrico/jev-routerTypeScriptA Python CLI that asks Jev which of four coding agents should take a task, optionally waits for your approval, then opens that app or its website and pastes the prompt in.
Jev Agents & MCP#community#awesomejev#catalog-import
Community
AABBAASS1/jev-routerPython
AABBAASS1/jev-routerPythonZero-dependency Node ticket router that answers routing, priority, urgency, auto-reply, sentiment, and complexity for each ticket in one parallel Jev call.
Jev Integrations & Adapters#community#github
Community
hj01857655/jev-routerJavaScript
hj01857655/jev-routerJavaScriptA Claude Code and Codex skill with a CLI that answers narrow agent judgments - is this file relevant, which team owns this request, how severe is this - with Jev's calibrated probabilities instead of another LLM call. Zero dependencies, MIT.
Jev Agents & MCP#community#catalog-import#madewithjev
Community
heyman333/jev-routerJavaScript
heyman333/jev-routerJavaScriptSystem One routing for coding agents where Jev decides and the LLM generates: an OpenCode v2 plugin, Cursor hooks and a skill cover turn steering, tool-call redirects, permission gating, failure advice, subagent routing, code review and on-demand jev_ask decisions.
Jev Integrations & Adapters#community#github#opencode
Community
drmaas/jev-routerTypeScript
drmaas/jev-routerTypeScriptMore Jev resources like this
Local Codex Router extension that asks Jev for a model tier and thinking depth on each model call, including continuations after tools, then relays the native Responses request verbatim. Fails open on Jev errors and logs decisions locally.
Jev Tools & CLIs#community#javascript#codex
Community
0xNatoshi/jev-codex-routerJavaScript
0xNatoshi/jev-codex-routerJavaScriptA Claude Code plugin from alexei-led that picks the model and effort tier, micro to high, for each turn, advised by Jev, Cloudflare's Clef family, OpenAI, or a local Ollama classifier. It changes only model and effort, and it never proxies Anthropic traffic.
Jev Tools & CLIs#community#github#claude-code
Community
alexei-led/claude-routerJavaScript
alexei-led/claude-routerJavaScriptA Python toolkit that installs nine plain SKILL.md files for Hermes, Claude Code, and Codex, letting Jev route models, filter retrieved passages, select skills, and choose bounded computer or browser actions.
Jev Tools & CLIs#community#python#agents
Community
kerpopule/hermes-jev-skillsPython
kerpopule/hermes-jev-skillsPythonA local gateway for coding agents that asks Jev which tool to call each turn and steers the LLM when Jev is confident, while every other request goes to your usual provider untouched.
Jev Tools & CLIs#community#typescript#gateway
Community
vinilana/jev-gatewayTypeScript
vinilana/jev-gatewayTypeScriptA local-first Herdr CLI that filters your agent subscriptions with fixed eligibility rules, then asks TypeSafe Jev to rank the remaining Cursor, Claude Code, Codex, or OpenCode options and pick a reasoning effort.
Jev Tools & CLIs#community#typescript#cli
Community
nidhi-singh02/agent-routerTypeScript
nidhi-singh02/agent-routerTypeScriptA Claude Code mod that draws a live cockpit above the prompt in CLI and Desktop: plan bars, subagent strips, usage limits with reset countdowns, and the model route jev-model-router picked with its confidence. It watches and draws only.
Jev Tools & CLIs#community#github#javascript
Community
nvr0x5/claude-deckJavaScript
nvr0x5/claude-deckJavaScriptFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.