AWS Launches Bedrock AgentCore at NYC Summit: Production-Grade Infra for Enterprise AI Agents

While everyone's shouting "AI agents," AWS showed up in New York and said: models alone aren't enough — you still need plumbing.

Three key facts

The AgentCore trio. At the AWS Summit NYC on June 17-18, AWS VP of Agentic AI Swami Sivasubramanian unveiled three new capabilities for Amazon Bedrock AgentCore: a Managed Knowledge Base with built-in data ingestion and native connectors to S3, SharePoint, Confluence, and Google Drive for end-to-end RAG retrieval; a web search tool that lets agents cite real-time web information without data leaving the customer's secured AWS environment; and production monitoring to help teams find and fix what's going wrong with agents in production.

Agent guardrails ship. AgentCore also introduced governance controls that scale as agents grow more capable — addressing the number-one enterprise fear: "what if my AI agent goes rogue?"

Kiro IDE unveiled. The summit also showcased Kiro, AWS's agentic IDE built on the VS Code open-source codebase, launched in May 2026 as the formal successor to Amazon Q Developer, featuring aerospace-grade spec standards for AI-assisted coding.

WangDou's Take

AWS didn't ship a model — it shipped the pipes and the water meter. What AgentCore does, stripped to essentials: your AI agent needs to connect to databases, search the web, and run in production without crashing — AWS handles all that dirty work. The clever part is that models get replaced and commoditized, but infrastructure is sticky. Once a company wires its knowledge base, data connectors, and monitoring into AWS, try migrating to Azure three years later — the migration cost alone will keep the CFO up at night. That's the cloud vendor's favorite business: don't sell the most expensive part, sell the glue that's hardest to peel off.

Source: AWS Blog — Top announcements of the AWS Summit in New York, 2026