Billing

Two ways to pay

Every tenant is billed one of two ways for POST /v1/ai/complete, decided automatically per request — you never choose it in the request itself:

• Metered billing — the platform's own Anthropic key is used, and the request is billed in KES against your prepaid balance (see Deposits for how to fund it).
• Bring your own key (BYOK) — you've stored your own Anthropic API key with the gateway (encrypted at rest with AES-256-GCM, never shown again once saved). Requests are sent using your key directly, so the platform isn't fronting Anthropic's cost at all — your prepaid balance is never checked and never moves for a BYOK request. You pay Anthropic their own rate, with zero platform markup.

A tenant can switch between the two at any time from their dashboard — saving a BYOK key switches you to BYOK immediately; clearing it switches you back to metered billing.

How metered cost is calculated

Metered requests are priced in Kenyan shillings (KES) from real token usage, not an estimate. The gateway reads the actual input_tokens and output_tokens Anthropic reports for your completion and prices them separately, because providers bill them very differently:

cost = input_tokens  / 1,000 × input rate
     + output_tokens / 1,000 × output rate

The current rates are on the Pricing page, which always shows the live numbers. For example, at KES 1.00 per 1,000 input tokens and KES 5.00 per 1,000 output tokens, a request with 1,000 input and 500 output tokens costs 1.00 + 2.50 = KES 3.50. Costs aren't rounded per request: a short prompt might cost "cost": "0.0457" (KES), and your balance reflects exactly that.

Rate changes

Rates can change, but never retroactively. A new rate is always scheduled with a future effective date (announced on Pricing), and a request is priced at the rate in effect when it started. Each charge in your ledger records the rate that priced it, so past charges never change.

The pre-request balance check

Before a metered request is sent to Anthropic at all, the gateway checks that your balance can cover at least the smallest plausible exchange (one input token, one output token) — if it can't, you get 402 Payment Required immediately (see Endpoint Reference), with no Anthropic call made. This is a floor, not an exact pre-authorization: the real cost isn't known until Anthropic's response comes back with actual usage numbers, so it's possible for a large response to leave your balance slightly negative afterward. A subsequent request is still checked against that same floor as normal.

BYOK, in practice

A BYOK-configured tenant's response always has "used_byok_key": true and "cost": "0", and "balance_after" simply reports your current balance, unchanged by the request — there's nothing to deduct, since you paid Anthropic directly with your own key. See Getting Started for a real example of both response shapes side by side.