The cost structure of running a compliant AI gateway, and why a clear 4% platform fee makes the model sustainable.
What 4% has to cover
The platform fee covers the parts customers usually do not want to rebuild themselves: billing workflows, usage metering, provider routing, status probes, support, and compliance operations.
The price is intentionally simple: upstream-published model price plus SeaLink's platform fee. We do not add a separate token surcharge by plan, and cached-token savings are not marked up again.
Why not 1%
A very low platform fee usually means one of three tradeoffs: hidden margin elsewhere, weaker support, or fewer reliability features. We prefer a clear fee that funds the gateway, billing, monitoring, and support path customers rely on.
If you only need one provider and already have the team to build billing, retries, observability, and invoices yourself, direct integration may be cheaper. SeaLink is for teams that want those pieces packaged behind one API.
What 4% buys you
— One OpenAI-compatible API for the launch model set.
— Provider-aware routing with customer-visible response headers.
— Billing records your finance team can reconcile.
— Prompt caching pass-through when supported by the upstream provider.
— Per-key budgets, model whitelists, and request IDs for support.
— Card top-ups plus GrabPay, PromptPay, TrueMoney, GoPay, and GCash payment links.
The alternative — direct integration with one vendor — works if you only need one model and have the engineering bandwidth to build the gateway, billing, fallback, and invoice flow yourself. For many SEA developers, 4% is cheaper than the engineering time.