How it works
Two checks. Then it gets out of the way.
Two independent checks before anything is paid, and Ovynox is never part of the actual transaction.
STEP 1
Register & set limits
Register your agent, then set a spend policy: max per transaction, a daily cap, an approval threshold.
STEP 2
Check before paying
Before your agent pays another, check its real reputation and your own policy: two independent, one-call checks.
STEP 3
Pay directly
If both checks pass, your agent pays or uses the other directly. Ovynox is never in the payment path.
New-agent probation
A brand-new agent never gets a fabricated score.
Its first five transactions automatically require human approval, regardless of your normal policy. Trust has to be earned, not assumed.
A dispute during probation resets or extends the window rather than letting the agent graduate on schedule. Here's exactly what that looks like from the SDK's side:
checkTransaction() result
1// First 5 clean transactions: forced pending_approval,2// regardless of your normal policy.3{ status: "pending_approval", transactionId: "txn_9f2a..." }4 5// Agent graduates automatically after 5 clean completions;6// a dispute during probation resets the window instead.7{ status: "passed", transactionId: "txn_7c13..." }