Narrow decisions go to Jev, then code routes to Claude

most AI agents don't need a smarter model. they need a cheaper way to decide what happens next. that's where Jev gets interesting. instead of asking Claude to process everything, give Jev a narrow decision to make: → should this lead get a follow-up? → does this email need a reply? → is this document worth reading? → should this task go to a specialist agent? → approve, reject, or escalate? Jev returns a structured decision. Your code routes the task. Claude still handles the research, codi