Your AI app costs money per call. Make it earn per call.

Declare billing.enabledand a price in your app manifest. Usage events, credit math, refund-on-failure, idempotency, and settlement are the platform's job — your code just does the work it charges for.

When enabled by the live server, signed-in users can spend from their own Aev balance, including model calls the app makes on their behalf. A quote is not a charge or settlement result.

Everything is cost-plus with a published markup. Ask for a quote before an expensive job, read every charge back from the ledger, and cap blast radius with per-app allowances.

Deploy on SettleMesh with billing declared in the manifest. When the live server enables delegated billing, signed-in users can pay per use from their Aev balance; otherwise keep the payment path unavailable. Read quotes, delivery status, and the ledger before claiming a charge or settlement.

Usage events, aggregation, credit balances, funding flows, refund-on-failure, idempotency, and reconciliation are separate responsibilities. A manifest flag declares intent; live server/preflight availability and readback decide what can actually run.

With end-user-pays, the signed-in user's balance funds the calls your app makes on their behalf — your margin no longer disappears into API bills for free riders.