Documentation menu

Docs/Core concepts/Billing model

Billing model

One prepaid balance, three meters. No subscriptions, no monthly fees, no auto-renewing card charges — when your agents stop, charges stop.

Prepaid balance

You top up a balance via Stripe (Alipay and WeChat Pay supported); larger top-ups earn bonus credit. Every product deducts from this single balance. Optional auto-top-up refills below a threshold you set — still prepaid in nature, never a subscription.

The three meters

  • Rotating traffic — per GB by pool type (residential $3.00, datacenter $1.20, mobile $6.00 — placeholder rates). Billed on actual bytes transferred, including failed requests.
  • Persona occupancy — per IP per month, priced by region (e.g. Japan $5.00/IP·mo), accrued daily from the balance rather than charged up front. Persona traffic is billed at rotating rates on top.
  • Managed Fetch (coming soon) — per successful request, tiered by target difficulty. Failures are free.

Budget caps and real-time enforcement

Every API key can carry daily and monthly budget caps. Enforcement happens in the request path: quota is pre-deducted before traffic flows, and long-lived tunnels are metered in 8 MB slices so a connection can be cut mid-stream within one slice of the cap. There is no after-the-fact surprise billing.

Attribution

Costs are attributed per API key and per agent_tag (a request-level label set in the gateway username, falling back to the key's default tag). The usage console answers "which agent spent what, where" — not just "what did the account spend".

Insufficient balance

At zero balance the account soft-stops: new traffic is rejected and the console becomes read-only. Personas enter a 7-day frozen grace period (the platform keeps the exit reserved) before release. Top up within the window and everything resumes untouched.

Rates shown here are launch placeholders pending the supplier benchmark; the pricing page is always the source of truth.

Next steps