AWS Agent Registry: enterprise catalog of AI agents now in preview
Why it matters
Amazon has released a preview of AWS Agent Registry, a centralized catalog of AI agents, tools and agent skills for enterprise organizations. The system indexes agents regardless of where they are hosted (AWS, other clouds, on-premises) and uses a combination of keyword and semantic search along with IAM-based access control.
On April 9, Amazon Web Services unveiled the preview version of AWS Agent Registry, part of the Bedrock AgentCore platform. The system addresses three key problems that enterprise organizations face when they begin scaling AI agents: a lack of visibility into which agents even exist in the organization, a lack of control over who can publish and use them, and duplicated development because teams are unaware that someone has already built similar functionality.
What the registry actually does
The registry indexes agents regardless of where they are hosted — whether on AWS, other cloud providers or on-premises systems. It uses standardized protocols such as MCP (Model Context Protocol) and A2A along with custom schema support. For each agent it records metadata: who the publisher is, how it is invoked, and which protocols it supports. Organizations can manually register entries via console or API, or point to an existing endpoint and have the system extract metadata on its own.
Search combines keyword and semantic matching, which means that when someone searches for “payment processing”, the registry also returns tools tagged as “billing” or “invoicing”. Without semantic search, such tools would be invisible to users who do not know the exact term the publisher used.
Governance features
For enterprise use, Amazon has added functions that reduce the risk of uncontrolled deployment: IAM-based access control for all operations, approval workflows that move entries from draft state to publicly discoverable, versioning with support for multiple active versions of the same agent, and deprecation tracking for retirement phases. The system integrates with existing organizational approval processes and supports custom metadata for compliance and ownership documentation.
For organizations with dozens or hundreds of teams experimenting with AI agents, this is the first real attempt by a major cloud provider to give AI agents the same treatment containerized software has received over the past 15 years — registry, versioning, governance.