AWS has launched Agent Registry in preview — a centralized catalog inside Amazon Bedrock AgentCore that lets enterprises discover, approve, and govern AI agents across their entire organization. The launch comes at a moment when agent sprawl has become one of the loudest concerns in enterprise IT circles, and it addresses a problem that many organizations have started to feel acutely as their agent deployments move from pilots into production.

The new registry is available now across five AWS regions: US East (N. Virginia), US West (Oregon), Asia Pacific (Tokyo), Asia Pacific (Sydney), and Europe (Ireland).

The Agent Sprawl Problem Is Real

The context behind this launch matters. Businesses that started experimenting with AI agents in 2024 and 2025 often did so team by team, use case by use case. An operations team deployed a scheduling agent. A finance team built a reporting agent. Customer support stood up a ticket-routing agent. By the time anyone looked up, enterprises had dozens — sometimes hundreds — of agents running across the organization with no central inventory, inconsistent governance, and significant duplication of effort.

This is what the industry now calls agent sprawl, and it is becoming a genuine operational problem. CIOs interviewed by CIO Dive described three core challenges: they don’t know what agents exist across the business, they have no consistent mechanism for controlling what gets published and made discoverable, and teams keep rebuilding capabilities that already exist elsewhere in the organization.

The scale of the shift is significant. Searches related to multi-agent systems surged 1,445% between Q1 2024 and Q2 2025 according to industry data, and that interest has translated into real deployments in 2026. AWS designed the Agent Registry specifically to address what happens after the deployment wave hits.

What the Registry Actually Does

Agent Registry functions as a private, governed catalog for an organization’s agents, tools, skills, MCP servers, and custom resources. Teams can register resources manually through the console or API, or use URL-based discovery — where the registry automatically retrieves metadata like tool schemas and capability descriptions by pointing at a live MCP server or agent endpoint.

Once registered, resources go through an approval workflow: they start as drafts, move to pending approval, and only become discoverable across the organization once approved by a designated reviewer. Administrators control who can register resources and who can discover them using IAM policies, keeping governance consistent with the rest of an organization’s AWS access controls.

Every action in the registry is captured in CloudTrail audit trails, so compliance teams have the visibility they need for governance and regulatory requirements.

For builders, the registry is queryable as an MCP server — meaning developers can search it and invoke registered resources directly from their IDEs without switching context. That makes it practically useful at the point where agents are actually built, not just as a management dashboard that gets ignored once a team starts shipping.

Why This Matters Beyond the AWS Ecosystem

The strategic signal here goes beyond the specific AWS feature. Major cloud providers building centralized governance infrastructure for AI agents is an indicator of where enterprise AI is in its maturity curve.

When the market is early, governance feels like friction. When deployments are in production across dozens of teams, governance is what makes the whole thing sustainable. AWS building an approval workflow, a catalog, and CloudTrail integration into an agent management platform suggests the enterprise AI market has passed the point where “move fast” is the dominant instruction to customers.

The timing also reflects a broader industry pattern. Microsoft announced its own agent governance capabilities through Agent 365 and the Agent Governance Toolkit earlier this year. Okta positioned its identity platform specifically around AI agents. The governance layer for agentic AI is becoming infrastructure, not a nice-to-have.

What This Means for Business

If your organization is deploying AI agents across more than one team, you already have the conditions for agent sprawl. The AWS Agent Registry is a practical response to a problem that tends to go unnoticed until it creates operational risk — duplicate tools, inconsistent behavior, no audit trail, and no easy way to share working solutions across the business.

For AWS customers, the immediate takeaway is straightforward: Agent Registry is in preview now, and getting your governance model right before agent sprawl becomes unmanageable is significantly easier than retrofitting it afterward.

The broader lesson applies regardless of which cloud platform you’re on. As AI agent deployments scale from one or two pilots to organization-wide infrastructure, the governance layer matters as much as the models themselves. Visibility, approval workflows, and auditability are not bureaucratic overhead — they are what turns a collection of experiments into a reliable operating system for your business.

For organizations still in the planning stages of AI deployment, this is exactly the kind of infrastructure decision that is worth getting right early. Building a governed approach to agents from the start — knowing what you have, controlling what gets deployed, and reusing what works — is one of the key factors separating AI deployments that scale from those that create more problems than they solve.

---

If you’re building with Claude or Codex right now, grab the free Working With Claude field guide. Thirty-two pages on the full ecosystem, Claude Code in depth, and how to roll agents out properly. Get the free guide.