AWS Agent Registry in Bedrock AgentCore: Scale AI Agent Management

What It Is: Introducing AWS Agent Registry in AgentCore

Managing a growing fleet of AI agents can quickly become a complex challenge for enterprises. Knowing which agents exist, controlling their use, and preventing teams from constantly rebuilding existing capabilities are critical hurdles. That's where the new AWS Agent Registry, now available in preview through Amazon Bedrock AgentCore, steps in. It provides a centralized hub to discover, share, and reuse AI agents, their tools, and skills across your entire organization.

The Agent Registry is designed to tackle these visibility, control, and reuse challenges head-on. It's built upon AgentCore, an open platform dedicated to building, connecting, and optimizing agents at scale, supporting any model, any framework, and any enterprise architecture. A key benefit is its ability to index agents regardless of where they're built or hosted – whether on AWS, other cloud providers, or even on-premises environments, ensuring comprehensive oversight.

Streamlining Agent Discovery and Reuse

The registry stores detailed metadata for every agent, tool, MCP server, agent skill, and custom resource as a structured record. This includes vital information like the publisher, implemented protocols (such as MCP or A2A), exposed functionalities, and invocation methods. You can register records manually via the AgentCore console, AWS SDK, or API, or automatically by pointing to an MCP or A2A endpoint, making it flexible for various deployment strategies.

Finding what you need is made simple through accessible channels like the AgentCore Console and APIs. It's even compatible with MCP-enabled clients like Kiro and Claude Code, and supports OAuth-based access for custom identity providers. A standout feature is its hybrid search capability, combining keyword and semantic matching. This means natural language queries, like searching for "payment processing," can intelligently surface related tools tagged as "billing" or "invoicing," dramatically improving discovery. Companies like Zuora, deploying 50 agents across various teams, are already leveraging the AWS Agent Registry in AgentCore for a unified view, promoting reuse and ensuring standardized metadata.

Enhancing Governance and Control

Beyond discovery, the AWS Agent Registry empowers organizations with robust governance capabilities. Admins can define who can register and discover agents, tools, and skills using IAM policies. Records follow a clear approval workflow, progressing from drafts to pending approval, and finally becoming discoverable enterprise-wide once approved. This ensures that only vetted and approved agents are made broadly available.

The registry tracks agents throughout their entire lifecycle, from initial development through deployment to eventual retirement. It also supports versioning to track changes over time and allows for deprecating records that are no longer in use. With the ability to add custom metadata, organizations can capture specific details like team ownership, compliance status, or deployment environment, providing a single source of truth for all your AI agents.

Read more: The Future of Managing Agents at Scale to explore how the AWS Agent Registry can transform your AI agent strategy.