Independent Jev AI guide · TypeSafe System One · September 2026

Jev AI: TypeSafe’s model for typed decisions.

Jev is not a chatbot. Send it application state and it returns a choice, a score, or a yes/no probability your code can branch on — currently billed at $0.042 per million input tokens, with output free.

jev-1.13.0Current public alias
$0.042 / 1MInput tokens, output free
choice · score · noulThree typed outputs
Hosted APINot open weights
software state typed output
state

"The customer has been billed twice and wants a refund today."

Jev / System One
decision
choicebilling0.98
noulescalate0.92
The contract

Messy text in.
Typed Jev decision out.

Jev’s value is not another chat window. It is a fast, typed branch your application can consume. The panel below is a local shape demo so you can see the interface before you call the hosted model.

shape demo · not the hosted Jev API local example
state0c
questions3 typed decisions
choicetopicbilling, bug, account, feature
scoreseverity0 to 3 risk level
noulescalateyes/no probability

This panel shows the contract: messy state in, typed values out. It is a local example, not TypeSafe Jev. For a live call, use the Jev API guide or the official console.

Typed output will appear here.
Where Jev fits

Put the Jev decision layer
between intent and action.

Jev is most interesting when a workflow already has an LLM, an API, or an agent, but needs a small reliable branch before the next step.

See all Jev use cases
Start with a branch

Call Jev for the fuzzy if
your agent was missing.

How to use the Jev API