What does jev-classifier do?

It connects your coding agent to Jev, a model that predicts which tool to use next. Start by watching its suggestions, then, where supported, let it choose the tool your agent calls. Every decision goes into a log you can review.

Which agents are supported, and how do they connect?

Codex, Claude Code, Grok Build and OpenCode connect through a proxy that passes requests through jev-classifier on their way to the model. Cursor and Antigravity use an MCP tool, calling Jev when they want a suggestion.

Install Node.js 22 or newer, then run npm install --global jev-classifier and check it with jev-classifier --help. The global install keeps desktop startup entries tied to a stable installation, and no repository checkout or build step is required.

Does it replace my agent's model or login?

No. The README says keep using your agent's model and login; you configure a separate key for Jev during setup. Connection settings apply to the current session, and your agent's configuration files stay as they are.

How do I try it without committing to enforcement?

Run jev-classifier setup and choose Observe to watch Jev's decisions first. If you do not have a key yet, choose Offline test to check the connection without calling Jev. Setup also picks your agent and saves your choices for all projects.

Does the gateway keep running after I close my agent?

Yes. For Codex, Claude Code, Grok Build and OpenCode the run command starts the gateway and opens your agent, but the gateway keeps running after you close the agent. Stop it with jev-classifier stop, or disable its optional desktop startup entry first.

Where did jev-classifier come from?

The original idea and inspiration come from jev-eval-agent by vinilana. This project brings that approach to a local gateway and MCP tools for coding agents. It is MIT-licensed and runs on Windows, macOS and Linux.

Jev directory

CommunityJev Integrations & Adapters

Jev Classifier — felpsdev

Local tool-routing classifier for coding agents: a gateway and MCP tool that ask Jev which tool to call next, keep your agent's own model and login, and log every decision so you can start by observing.

See highlights Open github.com

Stars: 4
Language: TypeScript
License: MIT
Reading time: 3 min
Published by: Community
Author: felpsdev
Last verified: Oct 2026

felpsdev/jev-classifier

User repository on GitHub

View on GitHub

Local tool-routing classifier for coding agents, with a gateway, MCP integrations, and decision logs.

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

Tagscommunity catalog-import logicrw routing gateway

Works forJev for Routing Jev for Tool calling Jev for Decision gating

Used inJev for Software & DevTools

Languages: TypeScript
Platforms: CLI MCP
Capabilities: Open sourceRequires API keyRuns locally

Highlights

Two connection styles: a proxy for Codex, Claude Code, Grok Build and OpenCode, and an MCP tool that Cursor and Antigravity call for suggestions.
Install with npm install --global jev-classifier; it needs Node.js 22+, with no repository checkout and no build step required.
jev-classifier setup walks through the Jev provider, its key and your agent; Observe mode shows decisions before anything is enforced.
Your agent keeps its model and login, and Jev is configured with a separate key.
The idea comes from vinilana's jev-eval-agent; this project brings it to a local gateway and MCP tools for coding agents.

Quickstart

bash

npm install --global jev-classifier
jev-classifier setup
jev-classifier run claude

Watch out

Copy reviewed against the repo README on 2026-10-15; the captured README was truncated mid-command in the Quick start section. Codex Responses Lite records suggestions without forcing a tool, and Cursor or Antigravity choose when to ask Jev and whether to follow it, so enforcement varies by agent.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

X: Jev 1.14 is now on OpenRouter

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

Other listings of this project

Forks and copies of this project in the directory. This entry is the canonical one.

Jev Classifier

A browser tool where you paste a Telegram channel JSON export and each post gets four Jev labels: category, quality, sentiment and reaction tone, written out as result.json.

Jev Guides & ArticlesArticle#community#awesomejev#catalog-import

Community

From the Jev guides

Original write-ups on what this entry is for.

Routing Models, Tools, and Skills with Jev

Sep 22, 2026

Every coding agent ships more than one model, and picking between them is a narrow decision. How the routers hand model choice, tool activation, and gateway policy to Jev without breaking the harness.

Jev Tools & CLIs#routing#models#coding-agent

Read article

Decision gates in agent frameworks: what each integration ships

Oct 16, 2026

How Pydantic AI, Mastra, TanStack AI, DSPy, Spring AI, and LiteLLM expose Jev: where the typed decision sits in the loop, what each integration returns, and which big frameworks are only adjacent.

Jev Integrations & Adapters#agents#frameworks#integrations