Anthropic opens up full private deployment for its Agent: execution records stay within the enterprise, while Claude remains in the cloud.

Beating AI News Flash: Anthropic announces the opening of Claude Managed Agents self-hosted private beta. Enterprises can now deploy the AI Agent's runtime framework, execution sandbox, access credentials, and session history entirely on their own cloud servers or internal data centers, no longer needing to hand them over to Anthropic for hosting.

In May of this year, Anthropic already allowed enterprises to self-host sandboxes, enabling AI to run code on internal servers. However, the core systems responsible for scheduling Agents, calling models, and storing session records were still managed by Anthropic.

The new version moves these control systems into the enterprise intranet as well. For example, when letting Agents automatically fix code, test software, or troubleshoot system failures, enterprises can restrict which servers it can access, manage API keys, and record and audit every step of operations via OpenTelemetry.

However, the Claude model itself still runs in the cloud. During Agent inference, the relevant prompts, code, and tool results still need to be sent to the model service provider. Enterprises can choose to call Claude through Anthropic, Amazon, Google, or Microsoft cloud services, or route all model requests through their own servers first to uniformly control access permissions, invocation channels, and costs.

The new feature is currently only available in private beta to enterprises whose applications are approved, and they need to contact the Anthropic customer team to obtain deployment files.