Anthropic said self-hosted Claude Managed Agents is now in private beta. Managed Agents is a set of tools for building agents that work on their own. Until now, customers could host only the sandbox, the isolated space where agent code runs. Anthropic said the new option lets a company run the whole service in its own cloud or on its own servers. That includes the agent loop, credentials and session history. Help Net Security reported that the beta opened on November 12.
Help Net Security said the customer runs the server, its PostgreSQL database and the workers that run shell and file tools. Model requests can go to Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry or the customer's own gateway. Help Net Security noted that Claude itself still runs outside the customer's network. Anthropic said a proxy outside the sandbox adds credentials for MCP tools and supported API calls. Neither source lists which API calls count as supported. Neither gives a price or a date for wider access. Customers must ask their Anthropic account team.
For teams that run agents unattended, credentials, logs and network limits now sit under their own control. Help Net Security pointed out that the customer must also patch the server and database and set the firewall allowlist correctly. Anthropic said each agent step can be sent over OpenTelemetry to monitoring tools a company already uses.