Frequently asked questions

Do I pay for tokens through you?

No. Bedrock runs in your AWS account, and AWS bills the tokens to you at your own prices. AI Cloud Gateway bills only per routed request, against your plan's monthly allowance. See Plans and billing.

Do I need to change my code?

Usually only the base URL and API key. The API follows the OpenAI Chat Completions format, with streaming, tool calls, base64 images and JSON output. Embeddings, the Responses API and n greater than 1 are not supported.

Do I need to enable models in my own account?

Yes. The gateway picks from the models enabled in your AWS account. Amazon Nova, Meta Llama and OpenAI gpt-oss work on first use; Claude models need a one-time approval. See Your AWS account.

Which model will answer my request?

With auto, the cheapest model expected to answer it well. The same request, mode and catalog version always get the same model, and the x-acg-model header says which one answered. See How routing works.

What happens when a model is throttled?

The request moves to the next suitable model before any output is sent. If every model is throttled, you get 429 bedrock_throttled.

Are my prompts stored?

No. Prompts and completions pass through the gateway in memory, in AWS us-east-1, and are never written to a log, database or file. See Data handling.

Which regions are supported?

us-east-1, us-east-2 and us-west-2.

Is there a free trial?

No. Starter costs $19 a month for 50,000 routed requests, and you can upgrade at any time.

Is it available on Azure?

Not yet. An Azure AI Foundry edition is planned; see Azure Marketplace.

How do I get an invoice?

Charges appear in the AWS Billing console of the account that subscribed.

More answers: Troubleshooting, or the full guides on GitHub.