Bridging AI Agents to Enterprise APIs with AgentCore Gateway

Amazon Bedrock’s AgentCore Gateway now supports Amazon API Gateway as a direct target, significantly simplifying the integration of existing enterprise REST APIs with AI agentic applications. Previously, organizations faced challenges in securely exposing their vast API estates to Large Language Models (LLMs) via the Model Context Protocol (MCP), often requiring manual translation and complex security configurations.

This new integration introduces an `API_GATEWAY` target type within AgentCore Gateway, eliminating the need to export API Gateway APIs as OpenAPI specifications. Developers can now expose new or existing REST API endpoints to AI agents with built-in security and observability, using a few console interactions or a single CLI command. Key features include `toolFilters` for precise control over exposed resources, and `toolOverrides` to customize tool names and descriptions for agents. The solution supports robust inbound and outbound authorization, including IAM-based SigV4 and API key authentication, ensuring secure interactions.

Benefits include streamlining the adoption of agentic applications by leveraging existing API infrastructure, standardizing secure data access for LLMs, and reducing integration complexity. The target audience comprises enterprises with extensive API ecosystems and developers building AI-powered workflows. For observability, the integration provides detailed logs via Amazon CloudWatch Logs, AWS CloudTrail, and AWS X-Ray, offering insights into tool calls, request parameters, responses, and performance metrics like latency and error rates. This empowers organizations to bring their established API landscape into the generative AI era efficiently and securely.

(Source: https://aws.amazon.com/blogs/machine-learning/streamline-ai-agent-tool-interactions-connect-api-gateway-to-agentcore-gateway-with-mcp/)