Operate Claude Managed Agents inside your security perimeter (claude.com)

Anthropic has announced the private beta launch of self-hosted Claude Managed Agents, allowing developers to implement autonomous agents directly within their company’s infrastructure. This new capability is significant as it offers organizations greater control over their sensitive data and security provisions, enabling agents to operate more effectively in environments where internal systems and code access are essential. With self-hosted options, teams can manage everything from session history to credential storage, which alleviates previous concerns regarding data privacy and compliance. The architecture of Managed Agents emphasizes security, dividing the "brain" (agent control) from the "hands" (where code execution occurs) to prevent any single point of failure. Key security features include the separation of session history from the sandbox, controlled credential management through an external proxy system, and customizable network access restrictions. Organizations can maintain detailed activity logs through OpenTelemetry, enabling seamless integration with existing monitoring systems. This innovative approach positions Claude Managed Agents as a compelling solution for organizations looking to leverage AI capabilities while safeguarding their internal environments.