News
AWS News Roundup Q1 2026: Key Updates and Announcements
Published on 03 Feb, 2026 by Jonathan
It’s been just over 3 months since the excitement of re:Invent, and hundreds of new AWS feature releases have occurred since. We want to share ten notable releases spanning agentic AI, security, data integration, IoT, storage, and productivity since the New Year.
The overarching theme is clear: AWS is investing heavily in the infrastructure needed to build, deploy, and govern AI agents at enterprise scale. Alongside this, there are meaningful improvements in security, visibility, and core platform capabilities that every organisation should be aware of.
Agentic AI & Amazon Bedrock AgentCore
Our biggest cluster of announcements centres on Amazon Bedrock AgentCore, which is rapidly becoming the backbone for deploying and managing AI agents on AWS. The following signals how seriously AWS is investing in this space.
Server-Side Tool Execution via AgentCore Gateway
In February, Amazon Bedrock introduced server-side tool execution through its integration with AgentCore Gateway and the Responses API. This is a significant simplification for teams building agentic workflows. Rather than writing and maintaining client-side orchestration loops, developers can now point a model at an AgentCore Gateway ARN and let Bedrock handle tool discovery, selection, execution, and result injection automatically within a single API call. Multiple tool calls per turn and real-time streaming of results are both supported.
Why it matters: This dramatically reduces the complexity of building agentic applications. If you are evaluating agent frameworks, the managed tool execution removes a significant piece of undifferentiated heavy lifting.
Stateful MCP Server Support in AgentCore Runtime
AgentCore Runtime now supports stateful Model Context Protocol (MCP) server features. This goes well beyond simple request-response: MCP servers deployed to AgentCore can now collect user input interactively during tool execution (elicitation), request LLM-generated content from clients (sampling), and send real-time progress updates for long-running operations. Each user session runs in a dedicated microVM with isolated resources and maintains context across interactions.
Why it matters: Stateful MCP opens the door to genuinely interactive agent experiences, think multi-step booking flows, guided troubleshooting, or progressive data collection, all running securely in isolated environments.
AG-UI Protocol Support for Real-Time Agent Frontends
Also, in March, AgentCore Runtime added support for the Agent-User Interaction (AG-UI) protocol. This open, event-based protocol standardises how AI agents communicate with user interfaces. It complements MCP (which gives agents tools) and A2A (which enables agent-to-agent communication) by focusing specifically on the user-facing layer. Capabilities include streaming text, reasoning steps, and tool results to frontends in real time, as well as state synchronisation for UI elements like progress bars. Both SSE and WebSocket transports are supported.
Why it matters: If you are building customer-facing or internal applications powered by agents, AG-UI provides a standardised way to deliver responsive, transparent experiences without custom plumbing.
Streaming Notifications for Long-Term Agent Memory
AgentCore Memory now supports push-based streaming notifications for long-term memory records via Amazon Kinesis. Previously, developers needed to poll for changes to memory; now updates stream automatically whenever records are created or modified. This enables event-driven downstream workflows, real-time application state refresh, and memory audit trails without custom polling logic.
Why it matters: Long-term memory is what allows agents to learn from past interactions and personalise future ones. Push-based streaming makes it practical to build reactive systems around this capability at scale.
AI-Powered Productivity & Self-Hosted AI
Amazon Quick Adds Third-Party AI Agents and Expanded Integrations
Amazon Quick, AWS’s AI-powered workspace, expanded significantly in January with the addition of third-party AI agents from Box, Canva, and PagerDuty, along with a broader library of built-in actions covering GitHub, Notion, HubSpot, Intercom, Linear, Monday.com, and more. The core value proposition is eliminating the context-switching overhead of working across multiple tools: users can now pull incident insights, generate presentations, query documents, create issues, and manage CRM records from a single conversational interface. Custom MCP and OpenAPI connectors extend Quick to thousands of additional applications.
Why it matters: For organisations already using AWS, Quick is becoming a compelling unified workspace. The third-party agent support and MCP connectivity are particularly interesting for teams looking to consolidate their tooling without replacing existing SaaS investments.
Amazon Lightsail Launches OpenClaw: Private Self-Hosted AI Assistant
Amazon Lightsail now offers OpenClaw, a self-hosted AI assistant that can be deployed on your own cloud infrastructure with minimal effort. Each instance includes built-in security controls: sandboxed agent sessions, one-click HTTPS, device pairing authentication, and automatic configuration snapshots. Amazon Bedrock is the default model provider, and the assistant can be connected to Slack, Telegram, WhatsApp, and Discord.
Why it matters: OpenClaw provides a low-barrier path to running a private AI assistant under your own control. For organisations with data sovereignty requirements or those simply wanting to experiment with AI assistants without third-party data exposure, this is a compelling option.
Security & Governance
AWS WAF AI Activity Dashboard for Bot and Agent Traffic
AWS WAF introduced a new AI activity dashboard in February that gives centralised visibility into AI bot and agent traffic hitting your applications. The Bot Control detection catalogue has expanded to cover over 650 unique bots and agents across categories, including AI search crawlers, data collectors, AI assistants, and LLM training crawlers. The dashboard lets you visualise traffic trends, identify the most active bots, analyse volumes by category and verification status, and take action directly through WAF rules, for example, allowing verified search crawlers while rate-limiting unverified agents. The dashboard is available at no additional cost for WAF customers.
Why it matters: AI-driven traffic is growing rapidly and often goes unnoticed. This dashboard provides the visibility needed to understand how AI bots are interacting with your applications, manage infrastructure costs, and enforce content access policies.
Data Integration
AWS Glue Native REST API Connector
AWS Glue now includes a native REST-based connector that allows teams to read data from any source with a REST API, without building custom connectors or managing specialised JAR libraries. This extends Glue’s existing catalogue of 60+ native connectors and supplements the 100+ third-party sources already available. The connector is available across all commercial regions and can be used via the Glue APIs, CLI, or SDK.
Why it matters: For data engineering teams, this removes a persistent friction point. Connecting to new or niche REST APIs for ETL workloads no longer requires custom development and ongoing library maintenance; it’s now a configuration task.
IoT
AWS IoT Device Management: Wi-Fi Simple Setup
AWS IoT Device Management launched Wi-Fi Simple Setup (WSS) for managed integrations in January. WSS allows end users to onboard Wi-Fi-enabled IoT devices by simply scanning a QR code with a mobile app, creating a near-zero-touch setup experience. Once Wi-Fi credentials are stored in managed integrations, new devices can connect almost automatically. The capability is currently available in Canada (Central) and Europe (Ireland).
Why it matters: Device onboarding remains one of the biggest operational headaches in IoT deployments. QR-code-based provisioning significantly reduces setup time and support costs, particularly for consumer-facing or high-volume device rollouts.
Storage & Core Platform
Amazon S3 Account Regional Namespaces
Amazon S3 now supports account regional namespaces for general-purpose buckets. This means you can create buckets in a reserved namespace tied to your account and region, eliminating the longstanding challenge of finding globally unique bucket names. Names become predictable and guaranteed available across multiple regions. The feature is available in 37 AWS Regions at no additional cost. Importantly, cloud security teams can use Service Control Policies (SCPs) and IAM policies to enforce that all buckets are created within account regional namespaces, ensuring consistent naming governance.
Why it matters: This is one of those changes that resolves a longstanding pain point. For organisations running multi-account, multi-region environments or building platforms with per-customer buckets, predictable namespace management is a welcome improvement. The policy enforcement angle is equally valuable for governance.
Looking Ahead
Q1 2026 makes it clear that agentic AI is AWS’s strategic priority. The pace of Bedrock AgentCore releases, covering runtime, memory, protocols, and tool execution, suggests AWS is building toward a comprehensive platform for enterprise agent deployment. Alongside this, practical improvements in security visibility, data integration, and core services like S3 continue to strengthen the overall platform.
As always, the Green Custard team is here to help you evaluate these capabilities and understand how they fit into your cloud strategy. If any of these announcements are relevant to your organisation, get in touch: sales@green-custard.com (or) send a query (or) book a meeting with our AWS experts.
We’d be happy to discuss further.