At today’s AWS Summit held in New York City, Swami Sivasubramanian, AWS VP of Agentic AI, delivered the keynote address.
Below is a summary of the most significant announcements from the event:
Agent Creation Updates
Amazon Bedrock AgentCore is rolling out new features: integrating AI agents with organizational, online, and paid knowledge sources, assisting teams with troubleshooting production issues, and applying controls that grow alongside increasingly capable agents.
These combined capabilities enable you to develop more powerful agents rapidly, manage them with controls that scale effectively, and refine them on an ongoing basis. For all the details, check out our blog post covering the full list of new features.
- Launch of Amazon Bedrock Managed Knowledge Base for faster, more precise enterprise AI applications — Enterprise RAG pipelines can now be constructed using Bedrock’s managed Knowledge Base. It features built-in data connectors, Smart Parsing for hands-free multi-format data processing, and an Agentic Retriever for handling complex multi-step queries — all seamlessly connected with AgentCore Gateway, allowing developers to concentrate on business results rather than infrastructure upkeep.
- Web Search on Amazon Bedrock AgentCore is now available: Give your AI agents access to up-to-date, well-sourced web information — A fully managed web search tool lets agents base their responses on current, properly cited web content with no data leaving your secure AWS environment. Rather than manually adding web search functionality to agents on Bedrock AgentCore and handling the underlying infrastructure, you can concentrate purely on agent development.
- AWS WAF introduces AI traffic monetization to help content owners bill AI bots for content access — A new Bot Control feature lets publishers and content providers set pricing, track usage, and accept payments from AI bots and agents that access their content and APIs. With AWS WAF, you can define access pricing, accept payments through third-party services, and grant scoped permissions directly at the edge.
- Amazon Bedrock AgentCore harness is now generally available — Production-grade AI agents can be built and deployed in minutes — without having to write orchestration logic. Simply define your agent’s model, tools, skills, and instructions via configuration within Bedrock AgentCore harness.
- Coming soon: AWS Context — This upcoming service will automatically map relationships across your existing data into a knowledge graph and deliver agentic search, enabling AI agents across the organization to tap into governed data relationships, business rules, and domain knowledge as they operate.
Agent Security Enhancements
- AWS Continuum introduced: Security moving at machine speed — AWS Continuum for code vulnerabilities, offered in a gated preview, aggregates findings from across your environment, ranks them by business impact, verifies which are actually exploitable, and drives remediation through your established workflow.
- AWS Security Agent (now folded into AWS Continuum) brings threat modeling, Kiro power, a Claude Code plugin, and more — A new threat modeling capability (in preview) lets you assess the full context of your application surface and uncover threats with recommended countermeasures using the STRIDE framework. Additionally, you can take advantage of pull request code scanning with automated remediation across major Git platforms, along with IDE integrations through Kiro power, the Claude Code plugin, and MCP — enabling developers to perform security reviews and apply fixes without ever leaving their workflow.
Agent Development Innovations
- Kiro for iOS is here — Kiro has launched a native iOS app, currently in a gated preview, designed for genuine engineering work. This gives developers a new way to kick off, track, direct, and interact with their Kiro sessions straight from their phone. Now you can initiate sessions, check results when they’re complete, review diffs, and approve changes — all while staying plugged into your work without needing a laptop.
- AWS DevOps Agent gains release management features to evaluate code changes before they hit production — A new release readiness review and autonomous release testing let you verify each change against whatever natural language criteria you’ve defined for the DevOps Agent. These features also execute change-specific tests in environments that closely mirror production.
- Proactively cut down on tech debt with AWS Transform – continuous modernization — Continuous analysis (in preview) can automatically scan your code repositories against configurable baselines and deliver findings in hours instead of weeks. Once you’ve identified and ranked the findings, you can set up autonomous remediations that automatically create pull requests for all affected repositories.
Agent Productivity Upgrades
With the debut of Amazon Quick’s new autonomous agents, you can build agents that run in the background with defined expertise, communication style, and tool access. You might set up a finance agent to handle incoming orders automatically, or a sales agent that watches activity across your CRM, email, and Slack channels to proactively draft follow-ups, flag potential risks, or suggest next actions.
A new activity feed has also been released, designed around how you actually work. It merges email, messaging, calendar events, and tasks into one prioritized feed — and it learns which messages you tend to reply to quickly, which conversations you tend to skip, and what topics dominate your week.
To see it in action, watch the demo of Amazon Quick – AI Assistant.
Beyond the keynote announcements, several other important launches have also been made this week:
