AWS Agent Registry Tackles AI Agent Sprawl with Centralized Discovery and Governance

AWS has introduced Agent Registry in public preview as part of Amazon Bedrock AgentCore, offering enterprises a centralized solution for discovering, sharing, and governing AI agents, tools, MCP servers, and agent skills across their organizations. The registry addresses a growing challenge in enterprise AI adoption: the rapid proliferation of agents that creates governance gaps, duplicate development efforts, and compliance risks.

The problem Agent Registry solves is straightforward but increasingly critical. As platform teams watch agent counts climb across departments, they face a familiar scenario: nobody knows what exists, who owns it, whether it's approved, or whether another team has already built the same capability. This agent sprawl compounds quickly, creating compliance gaps and wasting development resources on duplicate work.

How Agent Registry Works

There are two primary ways to register agents in the system. Teams can manually provide metadata through the console, SDK, or API, specifying ownership details, capability descriptions, and compliance status. Alternatively, they can point to an MCP or A2A endpoint and let the registry automatically pull in the details.

Every registered record captures essential information: who published it, what protocols it implements, what capabilities it exposes, and how to invoke it. The system supports discovery through a hybrid search that combines keyword and semantic matching. For example, searching for "payment processing" can surface tools tagged as "billing" or "invoicing" even when the exact terms don't match.

The registry is accessible through multiple interfaces: the AgentCore console, APIs, and as an MCP server itself. This means any MCP-compatible client, including Kiro and Claude Code, can query it directly. For organizations with custom identity providers, OAuth-based access allows teams to build their own discovery interfaces without requiring IAM credentials.

Governance and Compliance Features

On the governance side, Agent Registry implements a structured approval workflow. Records begin as drafts, move to pending approval, and only become discoverable once approved. Administrators control registration and discovery permissions through IAM policies. The system tracks version history, allowing organizations to monitor changes over time, and supports deprecating records that are no longer needed.

Custom metadata fields enable teams to attach additional context like cost center, deployment environment, or security classification. This flexibility helps organizations maintain consistent governance standards while accommodating their specific operational requirements.

Real-World Testing and Limitations

Shinya Tahara, a solutions architect, conducted hands-on testing that revealed both strengths and limitations. Semantic search performed well with English queries but struggled with Japanese queries against English-only metadata. In testing, one out of three Japanese queries returned zero results. The solution was straightforward: adding bilingual descriptions resolved the issue, providing valuable insight for global organizations planning deployment.

Tahara also discovered that folding filter constraints into natural language queries significantly degraded precision, often returning all registered records instead of the intended target. Additionally, updating any record resets its status to DRAFT, requiring re-submission and re-approval. Teams managing frequently updated agents should factor this friction into their workflows.

Enterprise Adoption and Benefits

Early adopters are already seeing value. Justin Bundick, VP of AI and Intelligent Platforms at Southwest Airlines, emphasized the registry's role in addressing discoverability challenges: "AWS Agent Registry in AgentCore enables teams to find and reuse existing agents rather than rebuild capabilities from scratch. With managed governance across multiple platforms, every agent carries standardized ownership metadata and policy enforcement."

Zuora provides another compelling use case. The company deploys 50 agents across sales, finance, product, and developer teams. Pete Hirsch, Zuora's Chief Product and Technology Officer, noted that the registry gives principal architects a unified view for discovering, managing, and cataloging every agent, tool, and skill in use, with standardized metadata ensuring consistent ownership and capability details across the ecosystem.

Future Roadmap and Integration Plans

AWS has outlined an ambitious roadmap for Agent Registry's evolution. Planned features include automatic indexing of agents the moment they deploy, cross-registry federation for searching across multiple registries as one, custom categories and taxonomies, and integration of operational data from AgentCore Observability.

The observability integration is particularly noteworthy. Future versions will incorporate invocation counts, latency, uptime, and usage patterns directly into registry records, providing teams with operational context alongside governance information. AWS has also mentioned early partner interest in connecting external catalogs to enable centralized discovery across technology landscapes.

Competitive Landscape

AWS is not alone in addressing agent sprawl. Microsoft offers Entra Agent Registry and Azure Agent Registry, while Google Cloud has its own Agent Registry. Protocol-level efforts like the Agent Client Protocol (ACP) Registry also exist.

What distinguishes AWS's offering is its provider-agnostic indexing combined with native support for both MCP and A2A protocols. This allows it to catalog agents built outside the AWS ecosystem, making it particularly valuable for organizations with multi-cloud or hybrid deployments.

Availability and Access

Agent Registry is currently available in preview across five AWS Regions: US East (N. Virginia), US West (Oregon), Asia Pacific (Sydney), Asia Pacific (Tokyo), and Europe (Ireland). This geographic coverage supports enterprise deployments across major business regions while maintaining the governance controls necessary for regulated industries.

The introduction of Agent Registry represents a significant step toward addressing the governance challenges that emerge as AI agents become increasingly central to enterprise operations. By providing a centralized catalog with built-in approval workflows, semantic search, and multi-protocol support, AWS is tackling the agent sprawl problem head-on while laying groundwork for more sophisticated governance and observability capabilities in the future.