Skip to content
Metatext Decide — typed decisions for your agents, with a probability on every answerGet early access
Metatext
Sign Up

jev-gateway

Local gateway for coding agents that asks Jev which tool to call next and passes everything else to your usual model.

Servicebeta
View on GitHub ↗

Overview

jev-gateway is a local LLM gateway for coding agents. When the agent is about to decide which tool to call, the gateway asks Jev instead of leaving the choice to the expensive reasoning model; all other traffic goes to the usual LLM untouched. It works out of the box with Codex, Claude Code, OpenCode, Kilo, Gemini CLI and Devin through launcher commands (jev-codex, jev-claude and so on), including on ChatGPT and claude.ai subscriptions, and with clients speaking OpenAI, Anthropic or Gemini APIs.

On first run a launcher asks whether to reach Jev through TypeSafe, OpenRouter, Vercel AI Gateway or OpenCode Zen, validates the key with one real call, and stores it in a user-only file shared by all launchers. A dashboard flag shows the gateway's decisions. It is an independent client of TypeSafe's public API. It requires Node 22.15+.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich tool should the agent call next?—

How it works

  1. Launch the agent through a jev- command
  2. Gateway detects tool-choice points in the conversation
  3. Ask Jev which tool to call
  4. Forward everything else to the usual model unchanged

Strengths

  • Drop-in launchers for many agents
  • Four provider options for Jev
  • Key stored once, user-only

Caveats

  • Independent project
  • Benefits not quantified in the excerpt

Getting started

`npm install -g jev-gateway`, then run `jev-codex` or `jev-claude` and answer the one-time provider and key prompts.

At a glance

Type
Service
Status
beta
Language
TypeScript
Runtime
Local gateway (Node 22.15+)
Has code
yes

Use cases and integrations

How curated lists describe it

  • An easy way to use jev with your coding agent for tool calling reasoning.

    wh000wh000/awesome-jev-live

  • An easy way to use jev with your coding agent for tool calling reasoning.

    hellogumbo/awesome-jev

  • Local gateway for Codex and Claude Code that asks Jev which tool to call next and passes everything else to your usual model.

    MrJev/awesome-jev

Frequently asked questions

What does Jev decide?

Which tool the coding agent should call next.

Which agents work?

Codex, Claude Code, OpenCode, Kilo, Gemini CLI and Devin via launchers.

Where can Jev be reached?

TypeSafe, OpenRouter, Vercel AI Gateway or OpenCode Zen.

Related projects

  • jev-routing

    Archived Go harness finding Jev tool steering does not reliably cut agent tokens in Claude Code or Codex; context trimming helped.

    MCP serverresearchGo
  • jev-orc

    Claude Code plugin that routes every subagent spawn through Jev to pick its model, effort and tool profile.

    Projectprototype
  • jev-classifier

    Local gateway and MCP tools that let Jev predict which tool a coding agent should call next, from suggestions to enforced choices.

    MCP serverbetaTypeScript
  • jevskill

    Agent skill that has Jev pick the few relevant lines from big logs before your coding agent reads them, with a reversible reduce and ledger.

    ProjectbetaPython
  • agent-fastpath

    MCP server giving coding agents fast typed decisions (ship gates, risk checks, file triage, page checks) as accept, review or escalate.

    MCP serverbetaTypeScript
  • azdaja

    Recursive language model layer for coding agents that keeps inputs in a local evaluator, with optional Jev judge_many for typed calls.

    ProjectbetaRust