Jev AI
Ask questions about any text.
Get probabilities, not prose.
Ask questions.
Get probabilities.
Jev AI turns any text into typed, calibrated answers. Ask many questions in one call and get the same predictable shape every time—without parsing prose or maintaining a classifier.
Yes / No
Binary decisions with a probability you can act on.
Choice
Route text into one of your own labeled options.
Score
Return a calibrated score across a defined scale.
Write your questions
Describe the decision in plain language.
Send your text
One document, or thousands of rows in a batch.
Use the answer
Build routing, scoring, and guardrails on structured output.
One request, many decisions
Keep related checks together and receive a stable answer shape your application can validate.
Fast to test, simple to scale
Start in the playground, reuse the same rules in batch jobs, then move to the API when the workflow is ready.
Spend stays visible
Every playground and API run shows its credit cost. Failed provider calls are reversed server-side.
Questions before you ship?
What can Jev classify?
Any text or JSON where you need a typed decision: support triage, moderation, routing, review scoring, lead qualification, guardrails, and citation checks.
Can I ask more than one question?
Yes. The playground and API accept up to 64 questions in one request, including Yes/No, Choice, and Score questions.
How does billing work?
A successful playground or API run costs one credit in this deployment. The interface shows that cost before you run, and failed upstream calls are reversed.
Can I move from testing to production?
Yes. Start with a saved decision in the playground, process a file in Batch, or use the API endpoint with a server-side API key.
Do you charge for output tokens?
No output-token charge is exposed by the Jev AI interface. Your balance is charged per successful run in the current deployment.
Where can I read the full contract?
See the API guide, Batch guide, and pricing page before integrating.