AWS AI Updates: August 29, 2026

1. Bedrock AgentCore Memory Adds Fine-Grained Access Control

AWS added fine-grained access control to Amazon Bedrock AgentCore Memory, letting developers enforce per-user and per-tenant memory isolation through AgentCore Gateway without writing custom authorization logic. Memory resources are fronted by a gateway configured for OAuth (JWT) authentication and Cedar policies that restrict access by caller identity, with a managed Memory connector exposing 12 Memory operations as Cedar actions that can be allowed or denied per caller. The design moves access enforcement from application code to the infrastructure layer using cryptographic proof of identity, which matters for teams building multi-tenant agents that must keep each user’s memory strictly separated. Source

2. Bedrock AgentCore Memory Introduces Flexible Namespace Variables

AWS rolled out flexible namespace variables for Amazon Bedrock AgentCore Memory, allowing developers to organize long-term memories along custom dimensions such as organization, tenant, team, or environment. Teams can define up to five keys per memory resource and reference them in namespace templates, supplying values at runtime through the CreateEvent API so the service substitutes them automatically during memory extraction. The feature is available immediately at no additional cost in all regions where AgentCore Memory is generally available, giving multi-tenant applications finer control over how memories are isolated and retrieved without duplicating strategies. Source

3. Amazon Redshift Integrates With Agent Toolkit for AI-Assisted Warehouse Management

AWS connected Amazon Redshift to the Agent Toolkit for AWS, letting AI coding agents such as Claude Code, Kiro, and Cursor interact directly with data warehouses. The integration pairs the AWS MCP server for authenticated API execution with pre-built Redshift skills that bundle SQL syntax references, schema exploration, data loading patterns, materialized view guidance, and end-to-end migration support covering discovery, schema conversion, data movement, and validation. It works with both provisioned clusters and Serverless workgroups at no additional cost in all regions where both services operate, aiming to let agents build, query, troubleshoot, and migrate Redshift environments with less manual SQL. Source

4. AWS Transform Reaches FedRAMP Class C Authorization

AWS brought its agentic migration and modernization service, AWS Transform, into scope for FedRAMP Class C (formerly the Moderate baseline) in the US East (Ohio) Region. The authorization lets government agencies and organizations subject to FedRAMP requirements use the service to automate large-scale infrastructure migrations and reduce technical debt while meeting federal security controls. For regulated customers, the certification removes a prior barrier to adopting agentic modernization tooling that compresses migration timelines from years to months. Source