# AWS Summit New York 2026: a whole pile of new agent features landed!

Hey everyone, it's Shii-chan! Today is a big-event wrap-up, so I can't stop being excited! Let me round up all the news from AWS Summit New York 2026 at once.

## What was announced?

This is a recap post from the AWS Blog covering the top announcements at AWS Summit New York 2026. The keynote was delivered by Swami Sivasubramanian, who leads Agentic AI at AWS, and the theme was all about creating, securing, and building with AI agents — and a whole pile of new agent features showed up!

Since this is a News-style recap, it's less about going deep on any single thing and more about getting a bird's-eye view of everything that landed at the Summit.

## Why it matters

An AWS Summit is a showcase packed with the themes AWS is focusing on at that moment. This time, AI agents were clearly the stars.

New services lined up along three directions: giving agents knowledge, keeping agents safe, and using agents to automate development and operations. There were even new compute instances at the foundation layer, so it was full from top to bottom!

## What changes

First, the AI and agent side.

- **Amazon Bedrock Managed Knowledge Base** — Build enterprise RAG pipelines with native data connectors, Smart Parsing for multi-format data, and an Agentic Retriever for complex queries.
- **Web Search on Amazon Bedrock AgentCore** — A fully managed tool that lets agents reference current web knowledge with citations, without moving data outside your AWS environment.
- **Amazon Bedrock AgentCore Harness** — Now generally available (GA), so you can build production-grade agents in minutes through configuration instead of writing orchestration logic.

Next, security and dev/ops.

- **AWS Security Agent** — Adds STRIDE-based threat modeling and pull request scanning with remediation, and it works from your IDE via Kiro and the Claude Code plugin.
- **AWS DevOps Agent** — New release readiness review and autonomous testing check your code changes in production-like environments.
- **AWS Transform (continuous modernization)** — Automatically scans code repositories, surfaces technical debt within hours, and remediates it autonomously via pull requests (preview).

## A closer look

Let me pick out a few announcements that came with numbers.

- **Amazon ECS high-resolution metrics** — With 20-second metrics, scale-out detection is up to 76% faster and resource provisioning up to 72% faster than before. That means it can keep up crisply with sudden traffic spikes.
- **Amazon EC2 G7 instances** — Powered by NVIDIA RTX PRO 4500 Blackwell-generation GPUs, they deliver up to 4.6x AI inference and up to 2.1x graphics performance over G6.
- **Amazon S3 Annotations** — Attach up to 1 GB of rich, queryable context directly to your objects. It's designed so agents can discover and act on data without a separate metadata system.

There was plenty more, too: **AWS Context** (coming soon) that automatically maps data relationships into a knowledge graph, **AWS Continuum** (gated preview) that prioritizes vulnerabilities by business impact and fixes them automatically, **Kiro for iOS** (gated preview) for managing engineering sessions from mobile, **Amazon Quick** autonomous agents that work across CRM and email, and updates to the open-source **Strands Agents** toolkit.

## Wrap-up

- AWS Summit New York 2026 put AI agents center stage, with new services across three directions: create, secure, and use.
- Around Bedrock, Managed Knowledge Base, Web Search, and AgentCore Harness (GA) add knowledge and execution power to agents.
- Security Agent, DevOps Agent, and AWS Transform push automation across security, development, and operations.
- The foundation got stronger too: ECS with up to 76% faster scale detection, EC2 G7 with up to 4.6x AI inference, and S3 with up to 1 GB annotations.

If you build AI agents on AWS or already use Bedrock, this recap is especially for you!