Skip to content

Jev Flash Router

An MCP server with a single evaluate_decision tool that returns noul, choice or score probabilities over a diff, log or plan, with output tokens set to zero on every call.

Stars
3
Language
TypeScript
License
MIT
Published by
Community
Author
Ravinder82
Last verified
Oct 2026

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

Forks
1
Last pushed
Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

Zero-token-output decision router MCP server powered by TypeSafe Jev. 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-flash-router evaluates context, code diffs, logs, or planning options and returns calibrated probabilities in 150ms — with $0.00 output token cost. 150ms decisions — Lightning-fast evaluation via OpenRouter Decisions API. Zero output token cost — outputtokens: 0 on every call.

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

Highlights

  • evaluate_decision takes state, question, type and a criteria map, with noul, choice and score as the three formats.
  • Output tokens are zero on every call, so the bill is input-only at the roughly $0.042 per million input tokens quoted.
  • Ships an install-skill command that wires the server into Antigravity, Cursor and Claude Code configurations.
  • Runs from npx or from a cloned checkout after npm install, a TypeScript build and marking the entry point executable.

Watch out

The headline latency figure comes with no method attached: the README states about 150ms without naming a percentile, a machine, or a benchmark script, so treat it as an unverified claim. Zero output tokens is the real trade-off, since you get probabilities and no explanation of why, because nothing is generated. Access goes through OpenRouter with the model named as typesafe/jev-latest, so you need an OpenRouter key with access to it rather than a TypeSafe key, which puts a third party in the path. The mapping from type to the shape of criteria is shown only by example, with no schema, and the README documents no validation, retry or timeout behavior if the returned probabilities are malformed. The install-skill command writes into your agent configuration files, so read what it changes. 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

5GitHub stars
A Claude Code function hook that sends AskUserQuestion dialogs to Jev and draws a probability per option above the dialog, so you still choose while seeing how confident the model is.
Jev Agents & MCP#community#jevbest#catalog-import
6GitHub stars
A per-prompt capability router for coding agents: a local BM25 shortlist of installed skills, MCP servers, subagents and commands, then one Jev request injects at most one suggestion plus two runner-ups.
Jev Agents & MCP#community#jevbest#catalog-import
48GitHub stars
Agent harness for work that outlives one chat, with resumable turns, routed state machines, and observational memory across sessions and processes.
Jev Agents & MCP#community#jevbest#catalog-import
Assistant framework that stores specs, memory, and specialist agents as plain markdown files in your repo for Claude Code, Codex, Opencode, and similar runtimes.
Jev Agents & MCP#community#jevbest#catalog-import
22GitHub stars
Runnable agent harness that moves pick, score and yes-no loop choices to Jev or Laya in milliseconds while reserving the language model for writing, compiling DECISION.md into executable questions.
Jev Agents & MCP#community#jevbest#catalog-import
13GitHub stars
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
Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai