AWS adds an agent registry
AWS introduced a Bedrock Agent Registry to give companies central visibility and lifecycle control over proliferating AI agents. (infoworld.com) That product responds to the practical problem companies face when they must decide whether agents run as centrally managed services or as coworker‑style assistants — an operating decision that now maps directly to governance needs. (youtube.com)
A year ago, a company could count its artificial intelligence agents on one team’s whiteboard. In April 2026, Amazon Web Services launched a registry because those agents now sprawl across departments, tools, and model stacks fast enough to need a catalog of record. (aws.amazon.com) Amazon Web Services calls the new product AWS Agent Registry, and it put it into preview on April 9, 2026 inside Amazon Bedrock AgentCore. The service is a private catalog where a company can publish agents, tools, skills, Model Context Protocol servers, and custom resources in one place. (aws.amazon.com) The immediate problem is not building one agent. The immediate problem is finding the right one after 50 teams have each built their own billing bot, support bot, research bot, and internal assistant with different owners and permissions. (docs.aws.amazon.com) A registry works like an internal app store mixed with a company directory. Amazon says teams can search by keywords and by semantic search, which means the system can match by meaning instead of exact words. (docs.aws.amazon.com) Amazon also tied the catalog to approval workflows. That means an employee or another agent can discover a tool in the registry, but access can still be gated by a review step instead of becoming a free-for-all. (docs.aws.amazon.com) This fits a bigger Amazon Bedrock AgentCore push that started in 2025. AgentCore was introduced as a managed layer to deploy and operate agents built with any framework and any model, not just agents built the Amazon way. (aws.amazon.com) Once Amazon let companies run agents from different frameworks, it also inherited the mess that comes with that freedom. A shared registry becomes the control panel for mixed fleets, the same way a cloud console became the control panel for mixed servers and databases. (aws.amazon.com; docs.aws.amazon.com) The product is also broader than a list of chatbots. Amazon says the registry can hold Model Context Protocol servers, which are connectors that let agents reach outside systems, so the governance problem includes not just which agent exists but which outside tools it can touch. (docs.aws.amazon.com) That is why this story is really about operating models. Some companies want agents to run like central services managed by platform teams, while other companies want coworker-style assistants that individual employees can discover and use on demand, and those two setups need different approval and visibility rules. (infoworld.com; youtube.com) Amazon is not alone in seeing that layer as the next control point. Forbes reported on April 10, 2026 that Amazon Web Services, Microsoft, and Google are all building agent registries or adjacent governance layers, which suggests the fight is shifting from who has the best model to who manages the biggest fleet safely. (forbes.com) The quiet detail in Amazon’s launch note is “ownership and lifecycle control.” That means the hard part is no longer only creating an agent on day one, but knowing on day 180 who owns it, which version is live, what it can access, and whether anyone should still be using it. (infoworld.com; aws.amazon.com) So the registry is Amazon Web Services admitting that agent era chaos has arrived inside real companies. The next phase of artificial intelligence at work looks less like inventing one brilliant assistant and more like keeping hundreds of useful ones from turning into shadow information technology. (infoworld.com; forbes.com)
