Anthropic launches Claude Managed Agents in public beta
Anthropic has launched Claude Managed Agents, a new product designed to help companies build and deploy AI agents without spending months setting up infrastructure.
The service is now in public beta on the Claude Platform and is aimed at developers building AI tools that need to run for long periods, manage tasks independently and connect with external software.
Claude Managed Agents handles much of the technical work that teams would normally need to build themselves. That includes infrastructure, orchestration, secure sandboxing, state management, permissions and long-running execution.
Anthropic says developers can define an agent’s tasks, tools and guardrails, while Claude takes care of the rest through its own infrastructure. The company says this could cut development time from months to days for teams trying to move from a prototype to a production product.
The launch reflects a wider trend across the AI sector, where companies are trying to make agents easier to deploy at scale. Businesses have often struggled with the cost and complexity of running agents that need secure access to tools, persistent memory and the ability to recover from errors during long tasks.
Claude Managed Agents includes long-running sessions that can continue operating for hours, even if a user disconnects. Anthropic is also testing multi-agent coordination, where one agent can direct other agents to complete separate tasks in parallel. That feature is still in research preview.
Anthropic says early users are already building practical tools with the system. Notion is using it for open-ended workplace tasks such as coding, generating slides and handling workflows inside its platform. Rakuten has deployed specialist agents across departments including sales, finance and marketing, while Sentry is using it to move from bug detection to code fixes within one workflow. Asana is also building collaborative AI teammates with the platform.
There are still questions around whether businesses will be comfortable relying on Anthropic’s infrastructure rather than running their own systems. Some developers may welcome the convenience, while others may worry about becoming too dependent on one provider for hosting, orchestration and governance.
Reaction from online communities has been mixed. Some developers see the product as a practical way to reduce engineering work and speed up deployment. Others have questioned whether companies should trust AI systems to manage complex workflows autonomously, particularly while the technology is still in beta.