AgentCore Gateway Now Supports Rate Limiting for AI Traffic

Amazon Bedrock AgentCore gateway, a fully managed, serverless AI gateway providing a single secure entry point for AI traffic, now supports rate limiting. This allows fine-grained control over how much traffic individual users can consume through the gateway. Rate limiting gives per-user control over usage of tools, inference models, and agents, with OAuth or IAM-based rules for requests per minute, concurrent connections, and token throughput. The gateway supports three target types: MCP targets, inference targets, and HTTP passthrough targets. Supported rate limit metrics include request rate limits (RPS/RPM) for all targets, token rate limits (TPM) for inference targets, and connection rate limits (CPS) for all targets. Dimension keys define how traffic is grouped into rate buckets, including targetName, toolName, qualifiedModelId, and identity claims like JWT claims and IAM principals. Entries specify allowed throughput per bucket, with wildcard support. Two layers of rate limiting are enforced: customer-defined limits and service quotas. Customer-defined limits include user-level, target-level, and hybrid target-user level limits. An example configuration shows per-group and per-user limits for Basic, Advanced, and Beta user groups, with restricted models for Beta users.