Three things that determine whether an AI gateway holds up under load: provider routing, resilience, and billing. How SeaLink does each.
What customers actually ask
Most enterprise buyers we talk to ask three questions in the first call. They don't ask about benchmark scores or rate-limit numbers. They ask:
— Which providers are actually available through the API?
— What happens when one provider is degraded?
— Can you give me a billing record my finance team can reconcile?
This post is the long answer to those three. We thought it was worth writing down so customers can read it before the call.
1. Provider routing
SeaLink exposes a curated model catalog instead of passing through every model name from every provider. Each model page and /v1/models response shows what is available through the API.
On successful API calls, the X-SeaLink-Upstream response header identifies the provider family that served the request. X-SeaLink-Original-Model echoes the requested model ID, and X-SeaLink-Served-Model shows the model name returned by the routing layer.
2. Resilience
Provider APIs can degrade for many reasons: rate limits, regional incidents, provider-side maintenance, or temporary auth failures. SeaLink keeps the customer API stable by routing through a gateway that can isolate unhealthy upstream paths and return consistent OpenAI-compatible errors.
When a comparable fallback is used, the X-SeaLink-Fallback: true response header makes the substitution visible. /docs/error-codes documents the exact behavior on customer-visible failure modes.
3. Billing
Top-ups support cards through Stripe and local wallet payment links for GrabPay, PromptPay, TrueMoney, GoPay, and GCash. Token usage is metered per request, and the dashboard shows live usage broken down by model, key, and customer-supplied task_type metadata.
Per-key budget caps enforce ceilings — set a daily or monthly limit on a key, and SeaLink returns 429 when the budget is hit, no surprise bills. Billing records and receipts are available from the billing area for finance reconciliation.
We charge a 4% platform fee on top of upstream-published prices. There is no hidden currency-conversion margin, no tier-based token surcharge, and no markup on cached tokens.
How to verify
Three things you can check in 30 seconds:
1. /trust lists SeaLink's legal and operational transparency information.
2. /docs/error-codes documents customer-visible failure modes.
3. Make a test call with any SeaLink API key. Look at the response headers: X-SeaLink-Upstream tells you which provider family served the request, X-SeaLink-Original-Model echoes what you asked for, X-SeaLink-Served-Model says what answered.
If any of this turns out to be inaccurate, contact compliance@sealink.asia so we can correct it.