Are the cost savings real, or just list-price arithmetic?

Mostly arithmetic, and the README says so. The 42 labelled tasks are scored against live OpenRouter prices with a script in the repository, so every percentage moves whenever prices do. The stronger claim is that cheap one-model habits miss the quality bar on 12–29% of tasks while balanced routing met the bar on all of them — and that is a judgement about 42 tasks, not a benchmark suite.

What happens if I do not give it an API key?

It still answers, using offline keyword analysis instead of a Jev call, and the README describes that path as less accurate without quantifying it. For the real behaviour, set OPENROUTER_API_KEY or TYPESAFE_API_KEY; the repository ships no keys, and the Claude Code plugin stores yours in your system keychain rather than in the repo.

Can it actually switch my model, or do I have to copy a command?

You copy it. The router returns the exact invocation — /model and /effort in Claude Code, -m and model_reasoning_effort for Codex, or an OpenRouter request body — and a separate delegate command routes and then runs the task in a subagent on the chosen model. Every MCP tool is read-only, so nothing in your session changes until you run what it printed.

How much does asking the router cost, and how slow is it?

The README quotes about $0.08 per 1,000 decisions and roughly 0.4 seconds for the Jev analysis, before any model call runs. That is small next to model spend, but it is not free and it sits on the critical path of every question you ask the router.

Which clients can host the plugin?

Claude Code, the Claude desktop app, and Codex, plus any other MCP client such as Cursor, VS Code, or Windsurf through the bundled no-dependency server.

What do the slash commands do?

In Claude Code, route gives the best model and effort for a task with alternatives, compare-models compares two to eight models, delegate routes then runs the task in a subagent, and setup checks keys and configuration.

How do the routing priorities work?

balanced is the default best trade-off; intelligence prefers the smartest option, speed the fastest good-enough one, and cost the cheapest good-enough one, and priority words stated in the task are picked up automatically.

CommunityJev Agents & MCP

Jev Model Router — Mandrilsquad1441

Plugin for Claude Code, the Claude desktop app and Codex that has Jev read the task, then ranks live OpenRouter models by benchmark quality, price, speed and reasoning effort, returning a pick plus cheaper and faster alternatives with the exact command to switch.

See highlights

Open github.com

Stars: 2
Language: TypeScript
License: MIT
Published by: Community
Author: Mandrilsquad1441
Last verified: Oct 2026

Mandrilsquad1441/jev-model-router

User repository on GitHub

View on GitHub

Pick the best AI model and reasoning effort for any task in ~1s. Plugin for Claude Code, Claude Desktop and Codex, powered by TypeSafe's Jev decision model and live OpenRouter pricing. Balance intelligence, speed and cost, or choose your priority.

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

This is one of several Jev listings for the same project. The canonical entry is Jev Model Router — satviksinha.

About this project

Pick the best AI model and reasoning effort for any task in about a second. A plugin for Claude Code , the Claude desktop app and Codex , powered by TypeSafe's Jev decision model. Every coding session starts with the same guess: which model, and how hard should it think? Flagship models on every task burn money and time; small models on hard tasks burn hours of rework. Jev Model Router makes that call for you.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification Jev for Routing

Used inJev for Software & DevTools

Languages: TypeScript

Highlights

One Jev call classifies the task's type, difficulty, reasoning need, and whether it wants tools, vision, latency or volume.
Returns a best-balance pick plus smartest, fastest and cheapest good-enough alternatives, each with the command that switches to it.
Priorities — balanced, intelligence, speed, cost — are picked up from how you phrase the task, with custom weights and hard constraints on top.
All four MCP tools are read-only, so the router can recommend and compare without changing your session.

Watch out

You need your own OpenRouter or TypeSafe key and Node 20+, and routing itself costs roughly $0.08 per 1,000 decisions on top of whatever the recommended model charges. The savings table rests on 42 labelled developer tasks priced at live list rates with a script you can re-run, so it measures the routing arithmetic rather than a controlled benchmark of model quality, and cheap one-model habits are reported as missing the quality bar on 12–29% of those tasks. Model quality scores come from benchmark data fetched from a URL that defaults to this repository, so they can lag a model release. Without a key it degrades to offline keyword analysis, which the README calls less accurate without saying by how much. Claude Code needs a session restart, and Codex hooks stay inert until you trust them. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev-code

TypeScript SDK and CLI coding agent where Jev only picks options while your code runs effects, building programs as syntax trees.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

rhighs/jev-codeTypeScript

pi-jev-router

Pi extension that classifies each task with Jev over OpenRouter and routes it to a Pareto-optimal model balancing measured quality, cost, and latency under a role policy.

Jev Agents & MCP#community#jevbest#catalog-import

Community

philippdubach/pi-jev-routerTypeScript

jevcache

Local OpenAI-compatible proxy that skips repeat model calls on same-intent paraphrases and retried agent turns, judged by Jev or a local System One model.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

kushals256/jevcacheTypeScript

jev-router

Node 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

jev-guard

Claude Code plugin that scores pending tool calls and finished answers with Jev noul probabilities, denying the call with the correction attached when a rule threshold is met.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

muratcakmak/jev-guardTypeScript

pijev

Terminal coding agent on Pi where Jev shortlists skills, ranks retrieved excerpts, and triages tool failures while the coding model handles reasoning and edits.

Jev Agents & MCP#community#awesomejev#catalog-import

Community

tonyzdev/pijevTypeScript

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai