Anthropic releases a self-hosted option for its Claude Managed Agents, allowing companies to run the agent harness and related infrastructure in their own cloud or data centre. The stated goal is to keep sensitive elements—such as agent credentials and agent histories—inside the customer environment, while model calls can still route to an outside provider like Anthropic or AWS.

The outlets differ mainly in emphasis. Help Net Security highlights operational fit for unattended use inside corporate systems, describing scenarios such as repairing broken builds, testing releases, triaging alerts, and working through software bug backlogs, and notes the option is available as a private beta starting November 12. The Next Web focuses on the deployment model: self-hosting for customer control of data and execution, paired with continued external handling of each model call.