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

jev-flash-router

MCP server exposing Jev Noul, Choice and Score via OpenRouter so coding agents can make quick decisions without reasoning tokens.

MCP serverbeta
View on GitHub ↗

Overview

jev-flash-router is an MCP server that gives AI coding agents fast typed decisions through Jev via OpenRouter's Decisions API. Its premise is that agents spend many reasoning tokens just deciding which file to edit, which route to take or whether a diff breaks tests; instead they can send context, diffs, logs or planning options and get calibrated probabilities back with zero output tokens.

It supports the three formats (noul, choice and score), runs via npx or from source, and works with Cursor, Windsurf, Claude Desktop and Claude Code. An install-skill command adds a skill that tells agents to use these decisions proactively. The README cites about 150 ms per decision; this is the author's figure.

What the decision model decides

TypeQuestionAnswers
ChoiceWhich file should be edited, or which route taken?—
Yes / noDoes this diff break tests?—
ScoreHow well does this option meet the rubric?(inferred)—

How it works

  1. Configure the MCP server with OPENROUTER_API_KEY
  2. Optionally install the agent skill
  3. Agent calls a noul, choice or score tool with context
  4. Agent acts on the returned probabilities

Strengths

  • Zero output tokens per decision
  • npx setup with no build
  • Works across several MCP clients

Caveats

  • Requires OpenRouter access to Jev
  • Latency figure is the author's claim

Getting started

Add an MCP server with command `npx -y jev-flash-router` and OPENROUTER_API_KEY; optionally run `npx -y jev-flash-router install-skill`.

At a glance

Type
MCP server
Status
beta
Language
TypeScript
Runtime
MCP server (Node)
Has code
yes

Use cases and integrations

How curated lists describe it

  • Open-sourced jev-flash-router: an MCP server for TypeSafe's new Jev model. AI coding agents waste hundreds of reasoning tokens just deciding which file to edit, which route to pick, or whether a diff breaks tests. Jev evaluates state and outputs calibrated probabilities. Works with Cursor, Windsurf, & Claude Code.

    hellogumbo/awesome-jev

Frequently asked questions

What does it provide?

MCP tools for Noul, Choice and Score decisions backed by Jev.

Which key is needed?

An OpenRouter API key with Jev access.

Which clients are supported?

Cursor, Windsurf, Claude Desktop and Claude Code.

Related projects

  • decisionkit

    CLI, MCP server and Claude Code plugin for closed-form yes/no, choice, score and rank decisions via Jev, OpenAI Decisions or Claude.

    MCP serverbetaTypeScript
  • jev-in-mcp

    MCP relay that adds use_jev to every server: Jev picks tool calls, the calling model writes free text, the relay executes.

    MCP serverbetaTypeScript
  • jev-skill-router

    Local Python MCP service that keeps the skill catalog out of the agent's context and lets Jev pick which SKILL.md files to load.

    MCP serverbetaPython
  • jev-mcp (burnigtm)

    Local MCP server putting Jev in the coding loop: route next steps, pick prepared tool calls, review patches and verify claims.

    MCP serverbetaTypeScript
  • Jev DSH Decision Engine

    Decision plugin for agent harnesses where Jev recommends tools, skills or owners and scores output, for DeepSeek Harness and iPolloWork.

    Agent skillbetaTypeScript
  • jev-code

    Gives Claude Code, Codex, Pi and OpenCode five typed Jev tools (classify, check, score, rank, ask) plus a skill on when to use them.

    MCP serverbetaTypeScript