Is this the official Jev website?

No. JEV Gateway is an independent service that gives you immediate access to the Jev model while official signups are limited. We are not affiliated with TypeSafe.

How are credits counted?

One credit covers 1,000 input tokens. A request costs one credit for every 1,000 input tokens, rounded up, and never less than one. A short ticket with several questions is about 400 tokens, so it costs 1 credit. Output tokens are free. Failed requests are never charged.

No. Credits stay in your workspace until you use them.

Can my team share credits?

Yes. On the Team plan you can invite teammates to a workspace. Everyone creates their own API keys and draws from the same balance.

Is Jev accurate enough for production?

It depends on your task. Validate on your own data first and use the confidence score to route uncertain cases to a person, especially for high-stakes decisions.

No waitlist · Live now

Use Jev today. Skip the waitlist.

Typed decisions with calibrated probabilities for classification, routing, scoring and safety checks. Test in the browser, then call the same API from your code.

5 free credits on signup · Credits never expire · Pay only for what you use

Try Jev now

Playground

Define typed questions over any state and inspect the decision. Free, no signup required.

Examples

1. State

Text or JSON shared by every question.

2. Questions

Up to 8 questions, answered independently over the same state.

3. Results

Run the request to see typed answers with probabilities.

4. Call it from code

curl

curl https://jevai.website/api/v1/systemone \
  -H "Authorization: Bearer $JEV_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "state": "Help! My payouts have been failing for 3 days and my customers are asking where their money is.",
  "questions": {
    "is_urgent": {
      "type": "noul",
      "instructions": "Does this convey urgency?"
    },
    "department": {
      "type": "choice",
      "instructions": "Which team should handle this?",
      "criteria": {
        "billing": "Payments, payouts, invoicing, refunds",
        "technical": "Bugs, outages, integrations",
        "sales": "Pricing, upgrades, new accounts"
      }
    },
    "frustration": {
      "type": "score",
      "instructions": "How frustrated is the customer?",
      "criteria": [
        "Calm",
        "Frustrated",
        "Very angry"
      ]
    }
  }
}'

What Jev is

It does not write a paragraph. It returns a decision.

A chat model writes a sentence, and then your code has to parse it. Jev only answers the questions you bounded: one option, one score, or the probability that something is true. The answer stays inside the set you defined, so your code can branch on it.

Choice

Pick one option from the list you wrote, with a probability for each. Use it to route, classify, or choose the next step.

Score

Place the state on a scale you define. The score can fall between two levels. Use it for severity, quality, and risk.

Yes or no

Judge whether a statement holds, as a probability from 0 to 1. Use it to block, check, or allow.

Use it when the answer is a decision

Which tool or person should handle the next step
Which queue or model a request should go to
Intent, priority, or language
Whether a result meets a rubric before a person sees it
Whether something is on topic, finished, or safe to send

When you need an article, code, or a conversation, keep using a chat model. Most systems do both: Jev decides, a chat model writes.

Why developers use Jev

Defined answer space

You declare every possible answer up front, so the output always fits your types. No parsing free text.

Probabilities, not guesses

Every answer comes with a probability distribution and a confidence score you can threshold.

Fast and cheap

Built for decisions many times a second. A typical request costs 1 credit.

Many answers per call

Classify, score and check safety over the same state in a single round trip.

Your code keeps control. Jev makes the call.

Separate the state, the question and the action. Jev only handles the decision in the middle.

Give it state

A ticket, a message, form fields or the JSON your agent already sees.

Ask typed questions

Choice, score or yes/no. Up to eight questions answered in one request.

Act on the probability

Route, block, queue or ask a human based on the answer and its confidence.

Agent integration

Give your coding agent a typed decision layer

Set one API key and your agent can route, block, and check completion with a choice, a score, or a yes-or-no probability.

Read the integration guide

Install the skill

The command downloads SKILL.md into the agent. That file is what teaches it to call /v1/systemone.

Configure once

Set the API key and pick English or Simplified Chinese for the examples.

Ask for a judgment

Give the agent a bounded question. It gets a choice, a score, or a yes probability the code can use.

Install the skill, then set the key

English by default. Chinese is supported.

mkdir -p .agents/skills/jev-gateway .cursor/skills/jev-gateway
curl -fsSL https://jevai.website/skill/SKILL.md -o .agents/skills/jev-gateway/SKILL.md
cp .agents/skills/jev-gateway/SKILL.md .cursor/skills/jev-gateway/SKILL.md

export JEV_API_KEY="sk_your_key_here"
export JEV_BASE_URL="https://jevai.website/api/v1"
export JEV_LANGUAGE="en-US"

The agent still decides

Jev only returns the judgment. Whether to act still depends on your agent, app permissions, fixed rules, and a person when it matters.

Decisions already hiding in your code

Browse all use cases

Simple credit packs

Buy once, use whenever. Credits never expire.

Starter

$10

Validate one real workflow

100,000 credits
≈ 100,000 typical requests
Credits never expire
1 workspace
3 concurrent requests
120 requests / minute
Up to 5 API keys per workspace
Choice, score, and yes/no questions
Probability and confidence on every answer
Up to 8 questions in one request
Usage and request history
Free playground
Email support

Pro

$100

Put decisions in production

1,000,000 credits
≈ 1,000,000 typical requests
Credits never expire
50 workspaces
10 concurrent requests
600 requests / minute
Up to 20 API keys per workspace
Choice, score, and yes/no questions
Probability and confidence on every answer
Up to 8 questions in one request
Usage and request history
Free playground
Email support

Team

$1000

Shared credits for your whole team

11,000,000 credits
≈ 11,000,000 typical requests
Credits never expire
50 workspaces
30 concurrent requests
2000 requests / minute
Up to 50 API keys per workspace
Choice, score, and yes/no questions
Probability and confidence on every answer
Up to 8 questions in one request
Invite teammates with roles
Includes 10% bonus credits
Usage and request history
Free playground
Email support

A short ticket with 3 questions is about 400 input tokens, so it costs 1 credit. $10 covers about 100,000 calls like that.

Compare plans

Questions, answered

Is this the official Jev website?: No. JEV Gateway is an independent service that gives you immediate access to the Jev model while official signups are limited. We are not affiliated with TypeSafe.
How are credits counted?: One credit covers 1,000 input tokens. A request costs one credit for every 1,000 input tokens, rounded up, and never less than one. A short ticket with several questions is about 400 tokens, so it costs 1 credit. Output tokens are free. Failed requests are never charged.
Do credits expire?: No. Credits stay in your workspace until you use them.
Can my team share credits?: Yes. On the Team plan you can invite teammates to a workspace. Everyone creates their own API keys and draws from the same balance.
Is Jev accurate enough for production?: It depends on your task. Validate on your own data first and use the confidence score to route uncertain cases to a person, especially for high-stakes decisions.