InferenceDirect doesn't ask you to pick a cloud. Every model is served from a managed registry of infrastructure providers, each with its own regions and credentials, switched on centrally as they come online. Your integration never changes — you keep calling api.inferencedirect.com.
Served natively from two regions — eu-central-1 and us-east-1 — with automatic per-location routing built into the registry. Every model on the Models page is currently served through Bedrock.
The provider registry is already built to onboard more infrastructure providers without any change to your integration. These are configured and awaiting credentials — not yet serving traffic:
If your team already has a direct account with a provider — including several not yet live on the platform — Bring Your Own Key routes through it today, billed by you at that provider directly, with InferenceDirect still providing the single endpoint, budgets, and observability layer on top.