AgentCore Gateway now supports Bedrock InvokeModel APIs and model fallback

Amazon Bedrock AgentCore Gateway supports the InvokeModel and InvokeModelWithResponseStream APIs for Amazon Bedrock at inference endpoints, also offering model fallback routing. This allows access to any model provider through a single managed endpoint without changing existing Amazon Bedrock SDK integrations. Fallback routing automatically redirects requests to alternate models or providers when the primary is throttled, lacks capacity, or is unreachable, keeping agents available. Gateway rules define alternates in order, switching on conditions like throttling, server errors, or response timeouts. Built-in circuit breakers skip currently failing targets, avoiding delays until the request reaches a healthy target. Handled internally, fallback routing requires only one request and one response per call, with no per-request trial and error. These features are available in all AWS regions where Amazon Bedrock AgentCore Gateway is offered, except China.

Why it matters

Developers can access multiple model providers without modifying existing Amazon Bedrock SDK, improving agent availability and simplifying retry logic implementation