AWS Summit NYC 2026: 9 AgentCore features for production AI

Key Takeaways

• Amazon Bedrock AgentCore now includes managed Knowledge Base, web search with zero data egress, and a GA-ready harness for deploying production agents without coding
• AWS Continuum introduces machine-speed security that prioritizes vulnerabilities by business impact and proves exploitability before driving fixes
• New developer tools include Kiro for iOS (manage agent sessions from your phone) and S3 annotations that attach up to 1GB of queryable context per object

Amazon used its 2026 AWS Summit in New York to announce nine features centered on a single bet: that AI agents will soon operate as autonomous workers in production environments, not just chatbot assistants. Swami Sivasubramanian, AWS VP of Agentic AI, delivered the keynote and framed the shift bluntly.

The announcements cluster into three buckets: expanding what Amazon Bedrock AgentCore can do, adding AI-driven security tooling under the AWS Continuum brand, and giving developers new surfaces to build and monitor agents. Here's what each release actually changes.

What's new in Amazon Bedrock AgentCore?

AgentCore is AWS's framework-agnostic runtime for building autonomous agents. The summit added four capabilities that address the gaps teams hit once they move past prototypes: data access, real-time information, monetization, and deployment speed.

First, Amazon Bedrock Managed Knowledge Base lets enterprises wire up RAG pipelines without stitching together separate vector stores, parsers, and connectors. It ships with native data connectors, Smart Parsing for multi-format documents, and an Agentic Retriever that handles complex multi-step queries. Everything integrates with AgentCore Gateway, so you configure once and skip the glue code.

Second, Web Search on AgentCore gives agents access to live web data while keeping customer data inside AWS boundaries. AWS claims 0% data egress for this tool, a direct response to enterprise concerns about LLM providers scraping internal context. The search results come with citations, which matters for compliance teams that need to trace where an agent's answer came from.

Third, AWS WAF now includes an AI traffic monetization capability. Publishers and content owners can set a price for bot access, meter usage, accept payment through third-party providers, and grant scoped access at the edge. This flips the script on the current web scraping arms race. Instead of blocking AI crawlers, publishers can charge them.

Fourth, Amazon Bedrock AgentCore harness reached general availability. It lets teams deploy production-grade agents in minutes by defining model, tools, skills, and instructions in configuration files. No hand-rolled orchestration loops required. The harness handles retries, state management, and tool execution.

How does AWS Continuum change security workflows?

AWS Continuum is a new security suite available in gated preview. Its premise: security teams drown in vulnerability findings but lack the context to know which ones actually matter.

Continuum ingests findings from across your environment, prioritizes them by business impact, proves which vulnerabilities are exploitable, and drives fixes through your existing CI/CD process. That last step is key. It doesn't just file a ticket. It generates a fix and routes it through whatever workflow you already use.

The AWS Security Agent, now folded into Continuum, adds threat modeling in preview. It analyzes application context, identifies threats using the STRIDE framework, and recommends mitigations. Developers can trigger security reviews from their IDE via Kiro power, Claude Code plugin, or MCP integration. Pull request scanning with auto-remediation works across major Git platforms.

Kiro for iOS: useful or gimmick?

Kiro is AWS's spec-driven IDE that orchestrates agents to write code. The summit introduced a native iOS app, available in gated preview. You can start sessions, monitor progress, review diffs, and approve changes from your phone.

Reddit threads have been skeptical. The question is whether complex code reviews translate to a 6-inch screen. AWS's pitch: you're not reviewing 500-line diffs on your phone. You're checking that your overnight refactor finished, approving a small PR while commuting, or steering an agent when you're away from your desk. The app handles asynchronous workflows where your laptop doesn't need to stay open.

What else shipped at the summit?

AWS DevOps Agent gained release management capabilities. It can assess code changes against natural-language standards before production and run change-specific tests in production-like environments. This is AWS doubling down on autonomous deployment pipelines.

AWS Transform continuous modernization entered preview. It scans code repositories against configurable baselines, generates findings in hours, and can auto-generate pull requests for remediations. If you've been putting off that framework upgrade, this is AWS's pitch to automate it.

Amazon S3 annotations now let you attach up to 1GB of rich, mutable, queryable context directly to objects. For AI agents that need to discover and act on data at scale, this removes the need for a separate metadata catalog. The context lives with the object.

What's the strategic angle?

AWS is positioning itself as the governance layer for agentic AI. The message to enterprises: you can build agents anywhere, but AWS will be where you deploy, monitor, secure, and control them. The WAF monetization feature signals that AWS expects AI agents to become major consumers of web content, and it wants publishers transacting through AWS infrastructure.

Hacker News discussions have been mixed. Some see Continuum and the DevOps Agent as genuine progress toward autonomous software maintenance. Others argue these tools are sophisticated wrappers for existing CI/CD automation, repackaged with AI branding. The real test will be whether enterprises adopt them for mission-critical workflows or keep them sandboxed for experiments.

Frequently Asked Questions

When will AWS Continuum be generally available?

AWS Continuum is currently in gated preview with no announced GA date. Interested teams can request access through AWS.

Does Amazon Bedrock AgentCore work with non-AWS models?

Yes. AgentCore is framework-agnostic. You can configure it to work with models hosted on Bedrock or external providers, though integration depth varies.

What is the cost of AWS WAF AI traffic monetization?

Pricing wasn't disclosed at the summit. Content owners set their own prices for bot access. Payment processing goes through third-party providers, which will have separate fees.

Is Kiro for iOS free?

Kiro for iOS is in gated preview. Pricing will likely follow the existing Kiro subscription model, though AWS hasn't confirmed mobile-specific pricing.

How does S3 annotations differ from S3 object tags?

Object tags are limited to 10 key-value pairs with a 256-byte total size. Annotations support up to 1GB of rich, queryable context per object, designed for AI workloads.

Source: AWS News Blog