6/5/2026, 12:00:00 AM ~ 6/8/2026, 12:00:00 AM (UTC)

Recent Announcements

Amazon Bedrock AgentCore Runtime introduces interactive shells for terminal access into agent sessions

Amazon Bedrock AgentCore Runtime now supports interactive shells through a new InvokeAgentRuntimeCommandShell API, opening a persistent, PTY-backed terminal directly into a running agent session over WebSocket. This complements the existing InvokeAgentRuntimeCommand API for one-shot execution, giving developers a full terminal experience inside an isolated microVM with colors, tab completion, Ctrl+C, terminal resize, and automatic reconnect on network drop.\n This is particularly important for developers hosting coding agents such as Claude Code, OpenAI Codex, Amazon Kiro on AgentCore Runtime. In addition to the asynchronous command execution they already had, they can now authenticate, drop into the microVM hosting their coding agent, and interact with it like a local terminal: interact with the agent, inspect files, run ad-hoc commands, or debug the environment state. The shell carries persistent state across commands within the same session, so environment variables, working directory, and command history all behave as expected. Each interactive session is identified by a runtime session ID and a shell ID. Passing both back when reconnecting lands you in the exact same shell. Brief network drops reconnect automatically, and longer disconnects can be resumed manually using the same IDs. A single agent runtime supports up to 10 concurrent shells, allowing developers to open multiple terminals against the same or multiple microVMs and watch agents work different branches in parallel. To get started using the AgentCore CLI: agentcore exec --it --runtime <runtime-arn>. To learn more, see Interactive Shells (Terminals) and Shell execution in AgentCore Runtime for a comparison of both shell modes.

Simplified permissions for Amazon S3 Tables and Iceberg materialized views are now available in AWS GovCloud (US) Regions

AWS Glue Data Catalog now supports AWS IAM-based authorization for Amazon S3 Tables and Apache Iceberg materialized views. With IAM-based authorization, you can define all necessary permissions across storage, catalog, and query engines in a single IAM policy.\n This capability simplifies the integration of S3 Tables or materialized views with any AWS Analytics service, including Amazon Athena, Amazon EMR, Amazon Redshift, and AWS Glue. You can also opt in to AWS Lake Formation at any time to manage fine-grained access controls using the AWS Management Console, AWS CLI, API, and AWS CloudFormation. This feature is now available in AWS GovCloud (US-East) and AWS GovCloud (US-West) Regions. To learn more, visit the S3 Tables documentation and the AWS Glue Data Catalog documentation.

Amazon OpenSearch UI is now available in GovCloud regions

Amazon OpenSearch Service expands its modernized operational analytics experience to GovCloud regions, including AWS GovCloud (US-East) and AWS GovCloud (US-West), enabling users to gain insights across data spanning managed domains and serverless collections from a single endpoint. The expansion includes Workspaces to enhance collaboration and productivity, allowing teams to create dedicated spaces. Discover is revamped to provide a unified log exploration experience supporting languages such as Piped-Processing-Language (PPL) and SQL, in addition to DQL and Lucene. Discover now features a data selector to support multiple sources, new visual design and query autocomplete for improved usability. This experience ensures users can access the latest UI enhancements, regardless of version of underlying managed cluster or collection.\n The expanded OpenSearch analytics helps users gain insights from their operational data by providing purpose-built features for observability, security analytics, and search use cases. With the enhanced Discover interface, users can now analyze data from multiple sources without switching tools, improving efficiency. Workspaces enable better collaboration by creating dedicated environments for teams to work on dashboards, saved queries, and other relevant content. Availability of the latest UI updates across all versions ensures uninterrupted access to the newest features and tools. OpenSearch UI can connect to OpenSearch domains (above version 1.3) and OpenSearch serverless collections. To get started, create an OpenSearch application in AWS Management Console. Learn more at Amazon OpenSearch Service Developer Guide.

Amazon ECS with AWS Fargate now supports 32vCPU compute configurations

Amazon Elastic Container Service (Amazon ECS) with AWS Fargate now supports 32vCPU compute configurations, enabling customers to run more demanding applications with greater flexibility and performance. AWS Fargate offers 32vCPU tasks with the following memory configurations: 60 GiB, 120 GiB, or 244 GiB, for both x86-based and ARM-based workloads on Linux. These new task sizes extend Amazon ECS’s capability to support high-performance computing use-cases, large-scale data processing, AI inference, and other compute-intensive workloads. With 32vCPUs and up to 244 GiB of memory, Amazon ECS customers can now deploy larger containers and scale applications beyond previous limits, all while leveraging the reliability, security, and scalability of AWS Fargate.\n To use the new 32vCPU task sizes, simply configure your task definitions to specify 32 as the vCPU value and select one of the new memory options (60, 120, or 244 GiB), then deploy your Amazon ECS services or tasks as usual via the AWS Management Console, CLI, or your infrastructure-as-code of choice. The new vCPU and memory configurations are available on both Fargate and Fargate Spot capacity providers, and existing Compute Savings Plans apply automatically. For pricing details, refer to AWS Fargate pricing page. The 32vCPU tasks are available with Amazon ECS and AWS Fargate in all AWS commercial and AWS GovCloud (US) Regions. To learn more, refer to the Amazon ECS documentation.

The AWS MCP Server now supports cross-account and cross-role access

Today, AWS announced cross-account and cross-role access for the AWS Model Context Protocol (MCP) Server, part of the Agent Toolkit for AWS. This feature allows developers using AI coding agents like Kiro, Claude Code, or Codex to work across multiple AWS accounts and AWS Identity and Access Management (IAM) roles within a single session, with no restarts required. Previously, switching profiles required stopping the AI coding session, updating local AWS credentials, and restarting the MCP server for every account change. Now, AI agents using the AWS MCP Server can specify a profile on each command, allowing users to switch between accounts and roles seamlessly.\n

Cross-account access helps developers move faster across multi-account environments. For example, a DevOps engineer can query CloudWatch logs across production and staging accounts to diagnose a performance issue, or an application developer can update a Lambda configuration in one account and adjust an S3 bucket policy in another, all within the same conversation. Each request specifies which profile to use, so there is no risk of commands reaching the wrong account.

To get started, see Multi-profile support in the Agent Toolkit for AWS user guide. The AWS MCP Server is available in the US East (N. Virginia) and Europe (Frankfurt) Regions.

AWS Blogs

AWS Japan Blog (Japanese)

AWS News Blog

AWS Messaging Blog

AWS Security Blog

Open Source Project

AWS CLI