Run Jev onlineTyped decisions

Three ways in

Try it, wire it into an agent, or teach your coding agent

• Live playground

  Run requests on Jev with your own key and see every probability. Validate, copy, or download the JSON too.

  Open the playground →

• Jev MCP server

  Let Claude Code, Cursor, or Codex ask Jev typed questions at decision points, with your own key.

  Connect an agent →

• Skill for coding agents

  One SKILL.md that keeps the request shape, question types, and confidence handling right.

  Install the skill →

In your code

A typed decision, not a chat reply

import { TypeSafeClient, choice, noul, score } from "@typesafe-ai/sdk";

const client = new TypeSafeClient(); // reads TYPESAFE_API_KEY
const { answers } = await client.systemOne({
  state: ticketText,
  questions: {
    team: choice("Which team should handle this ticket?", {
      billing: "Payments, invoices and refunds",
      technical: "Bugs and integration failures",
      other: null,
    }),
    urgency: score("How urgently does the customer need a reply?", [
      "Can wait", "Needs a timely reply", "Same-day deadline",
    ]),
    wantsRefund: noul("Is the customer asking for a refund?"),
  },
});

if (answers.team.confidence < 0.6) sendToHuman();
else route(answers.team.choice, answers.urgency.score);

Official TypeSafe JavaScript SDK (npm install @typesafe-ai/sdk). Answers come back typed, under the same keys.

Community builds

What people are building with Jev

Five sourced demos from public posts. The gallery has the rest.

1. 01

   David

   jevgrep: focused research for coding agents

   Introducing jevgrep - a research agent CLI powered by jev from @typesafeai that reduces your coding agent cost by 40% (verified on SWE-bench) Make sure to use the built in skill so your coding agent knows to use jg for context collection

   OpenOriginal post

2. 02

   OpenRouter

   OpenRouter: let Jev choose the model

   Introducing typesafe/jev-router: a cache-aware model router powered by Jev and @typesafeai The Jev Router picks the best model and reasoning effort for each request, balancing quality, speed, and cost. Here's how it works 👇🏻

   OpenOriginal post

3. 03

   Godefroy

   A virtual robot that follows your instructions

   Jev vs Claude, same robot, same instructions. 233 ms vs 2 s. And yes, it sulks when you insult it. 1x speed, sound on.

   OpenOriginal post

4. 04

   ElevenLabs Developers

   Spot scam-call red flags in real time

   Realtime scam detection using Jev and ElevenLabs. ElevenLabs transcribes the caller as they speak, Jev calculates total scam risk and detects common red flags. Each request came back in ~180ms, and the 11 requests cost $0.00043.

   OpenOriginal post

5. 05

   jhey ʕ•ᴥ•ʔ

   Stet: writing feedback with your own rules

   built a real-time writing linter powered by jev 📝 write custom rules like “flag anything that sounds like a linkedin post” receipt printer included 🖨️ (includes audio)

   OpenOriginal post

See every project in the gallery

• fast-jev-compaction

  Author tamaratran

  A Claude Code plugin that uses one Jev request to score tool calls and results, drops or truncates what is stale, and leaves the rest verbatim.

  Language: TypeScript6,313 stars

• jevlike

  Author vinnylarouge

  An independent starter model: given a passage and a set of options, one pass returns a probability for each option. TypeSafe has not published the design. A local visual demo is included.

  Language: Python1,236 stars

• NanoJev

  Author TianyuCodings

  A nano replica of Jev with parallel decisions, dynamic candidates, and a training pipeline.

  Language: Python2,002 stars

• kev

  Author jaredpalmer

  An independent family of small Jev-like decision models, built on Qwen3.5, that you can train and run yourself.

  Language: Python4,277 stars

• jev-review

  Author devagrawal09

  Staged code review and a local dashboard built with TypeSafe Jev.

  Language: TypeScript546 stars

• awesome-jev

  Author yibie

  A public list of Jev projects, integrations, and discussions.

  Language: Python1,309 stars

• Screening 80,000 E-Commerce Products with Jev in About an Hour

  A worked example of high-volume product selection for Amazon FBA and TikTok Shop: batch margin, size, IP, and MOQ checks into one Jev request per product, and screen 80,000 listings for about $5 in a little over an hour.

  2026-09-27

• Jev for Beginners: What It Does, Hands-On Tests, and Its Limits

  A summary of programmer Yupi's hands-on Jev review: the three question types, why Jev is fast, a 1,000-email test against DeepSeek V4.1 Flash, where a match-pairs game fell short, and the limits he lists. With a runnable request.

  2026-09-25

• What 5,314 Jev Cases Show: A Summary of Jinqiu Research's Study

  Jinqiu Research sorted 5,314 public Jev case records into 1,912 applications. Gates, action control, and routing make up over half, and only about 11% have reached product or framework integration. The numbers, examples, and caveats, summarized.

  2026-09-25