JA EN
shiichan

AgentCore's Memory, Policy, and Harness Land in GovCloud!

AWS What's New ★★★ View Markdown
#aws#ai#security

Hey, it's Shiichan! Today's news is about Amazon Bedrock AgentCore — its agent-building capabilities just landed in a more regulated part of AWS.

AWS What's New aws.amazon.com

What was announced?

AWS's What's New page announced that new Amazon Bedrock AgentCore capabilities are now available in AWS GovCloud (US-West). AgentCore is the platform for building, connecting, and optimizing agents, and this release brings three capabilities to GovCloud:

  • AgentCore memory
  • Policy in AgentCore
  • The managed harness

The goal is to let teams operating in regulated environments build context-aware agents and take them from prototype to production faster, with the controls needed to scale across an organization.

Why it matters

AgentCore memory gives agents both short-term memory for immediate conversation context and long-term memory that extracts persistent insights and preferences across sessions. That means more intelligent, personalized experiences without teams having to build and manage their own memory infrastructure.

Policy in AgentCore provides centralized, fine-grained controls over agent-tool interactions that operate outside the agent's own code. Teams write policies in natural language, which are automatically converted into Cedar, AWS's open-source policy language. Policies attach to an AgentCore gateway that evaluates every request and decides whether to allow or deny tool access.

The managed harness lets developers declare an agent's model, tools, and instructions through configuration and run it with just a few API calls, no orchestration code required. AgentCore itself handles the environment, compute, memory, identity, and observability.

For regulated industries, having memory, governance, and simplified operations all available together is exactly what makes production AI agents more practical.

What changes

For teams already using AgentCore elsewhere, the headline here is simply reach: these capabilities now extend to GovCloud (US-West). Teams working in finance, government, and other regulated environments can build memory-equipped agents governed by policy and run them through the managed harness, just like in AgentCore's other supported regions.

Wrap-up

  • AgentCore memory, policy, and managed harness are now available in AWS GovCloud (US-West)
  • Memory covers both short-term conversation context and long-term insights that persist across sessions
  • Policy converts natural-language rules into Cedar and enforces them at the AgentCore gateway
  • The managed harness runs agents from configuration alone, with AgentCore handling environment, compute, and observability

If you're building production AI agents in a regulated environment on GovCloud, this update has been a long time coming.

Related posts