Billing

SoxAI uses a prepaid credit model with atomic pre-consumption billing. You add funds to your balance, and usage is deducted as requests are processed.

How It Works

1. Add credits to your balance (Top Up)
2. Make API requests
3. SoxAI pre-consumes an estimated cost before sending to the upstream provider
4. After the response, SoxAI settles based on actual token usage
5. Underage is refunded; overage is charged
6. Errors result in a full refund

This approach prevents overspending while ensuring billing accuracy.

Pre-Consumption

Before forwarding your request to the upstream AI provider, SoxAI estimates the token cost and holds that amount from your balance. This requires your balance to be positive before making requests.

If your balance is insufficient to cover the estimated cost:

• The request is rejected with a 402 Insufficient Balance error
• No tokens are consumed
• You are redirected to top up

Settlement

When the upstream response is complete, SoxAI compares the actual token usage against the pre-consumed estimate:

All settlement operations are atomic PostgreSQL transactions — there is no risk of double-charging or lost refunds.

Error Handling

If a request fails (network error, 5xx from upstream, timeout):

• The pre-consumed amount is fully refunded
• No charges apply for failed requests
• The error is logged for diagnostics

Viewing Balance and Usage

Console: Billing → Overview

The billing overview shows:

• Current balance
• Month-to-date spending
• Spending breakdown by model and team
• Recent transactions

Sections

• Pricing — Model prices and markup details
• Top Up — Add credits to your balance
• Invoices — Download invoices and transaction history