Typed decisions for agent workflows
Jev Playground
Try Jev on your own text. Classify, score, or ask a yes/no question and inspect the real API response.
Checking live access…
Independent playground by VerdictKit.
VerdictKit is a hands-on hub for System One models — decision models that return a probability instead of a paragraph.
Load an example response, or enter your own state and question.
Jev field guide
What is noul in Jev?
Jev’s yes/no question type and how the 0–1 probability works.
Self-hosted Jev: what runs locally?
Separate local application code from hosted Jev inference and independent local alternatives.
Jev as a classifier
Follow a support-queue example with JSON input and typed decisions out.
Jev limitations and known failure modes
Review nine documented failure modes and practical checks before using a decision.
Jev JSON output and response fields
Understand Choice, Score and Noul response shapes and how to parse their fields.
How to call the Jev API
Build a request with the official endpoint, API key setup and error reference.
/ INTRODUCTION
What Is the Jev Playground?
Jev is TypeSafe AI’s System One model. It evaluates the state you provide against questions with defined answer shapes. Instead of writing prose, it returns typed decisions and probabilities your code can use. VerdictKit lets you build a request without signing in and inspect clearly marked demo JSON; live evaluations require a configured server API key.
/ THE PRIMITIVES
Choice, Score and Noul
Three question types cover three different decision shapes. Pick the type before you write the question.
Choice
Pick exactly one option from your named set. Read the selected label, full probability distribution and confidence.
Score
Rate a state against ordered descriptions. Get a weighted score, legend, probabilities and confidence.
Noul
Ask a yes/no question and get one number: the probability that the answer is yes. Explore noul questions.
/ AGENT EXAMPLE RUNS
From context to decision.
Each card shows a preset response for a common agent task. Its probability values illustrate the response shape; they are not predictions from a Jev run. A live API result carries the LIVE RUN label.
Select an agent tool
Current public API pricing is requested; choose among internal docs, browser and calculator.
browser0.82Score action risk
An agent plans to send customer data to an unverified external endpoint.
high1.71 / 2Set ticket priority
A production outage blocks every workspace member from deploying.
P00.91All values shown here are preset illustrations, not results from live Jev evaluations. Live responses are marked LIVE RUN. More agent decision examples.
/ FOR DEVELOPERS
JSON Output
Every answer has a predictable shape. See the complete Jev JSON output schema, field reference and parsing pitfalls.
Explore Jev JSON output/ FOR DEVELOPERS
Use the Jev API
Start with the official endpoint, key setup, request examples, error codes and current pricing.
Read the Jev API guideCompare model costs and VerdictKit plans on the Jev pricing page.
/ FAQ
Questions about Jev
How much does it cost?
Demo previews are free. Live access offers two guest runs per UTC day without signup or five per UTC day with a Google account. Run Pack is $19 once, Pro is $12/month or $72/year, and Batch is $29/month or $174/year. Sign in on the account page to purchase a plan.
Do I need an account to try it?
No account is needed for demo previews or the two-run daily guest trial. Sign in with Google for five live runs per day on the Free account. Both allowances reset at 00:00 UTC.
Is this the official TypeSafe playground?
No. This site is operated by VerdictKit and is not affiliated with TypeSafe AI.
What is Jev?
Jev is TypeSafe AI’s System One model for typed decisions against a supplied state.
What’s the difference between choice, score and noul?
Choice selects an option, Score rates ordered levels, and Noul returns a yes/no probability.
Why does Jev return probabilities instead of a plain answer?
Probabilities let your code set thresholds and route uncertain cases for review.
How fast is Jev?
It is designed for fast decisions. Actual latency depends on state size, API load and network; benchmark your own use case.
What does the Jev API cost?
TypeSafe currently lists $0.042 per million input tokens for Jev 1.13, with free output tokens. Verify current pricing.
Can I use playground output in production?
Demo probabilities are prefilled illustrations and are not real predictions. Validate live output and thresholds before production use.
Where can I learn more about the Jev API?
Read our API guide and the official TypeSafe documentation.