Claude agents can now run in 1 firm's cloud

Anthropic now lets companies run its Claude Managed Agents service on their own servers, the company said on the 12th of November 2026. The option is in private beta, a limited early test.

Managed Agents is Anthropic's service for building AI agents, programs that use tools and work through tasks on their own.

What stays inside the company

Agents, their saved histories and passwords now stay on the company's own systems, Anthropic said. Firms can run the service in their own cloud account or data center.

Companies must ask their Anthropic account team for access, the release notes say.

The catch

Each time an agent asks Claude a question, the conversation still goes to an outside model provider, The Next Web reported. Firms can pick Anthropic, Amazon Bedrock, Google Cloud or Microsoft Foundry.

Some features are missing, including scheduled runs and webhooks, which are automatic alerts, Anthropic's guide says. Nothing moves over from the hosted service, so agents must be set up again.

Below we break down the choices

Firms can send questions to 4 named providers or to their own gateway, Anthropic said. That makes 5 routes, by our arithmetic.

Only 1 of those 5 routes is run by the firm itself.

What is confirmed and what is not

Anthropic's release notes confirm the launch and the private beta. Its pages give no date for wider access.

Help Net Security first reported the launch, The Next Web said.

Key facts

Anthropic launched self-hosted Claude Managed Agents on the 12th of November 2026.

It is in private beta.

Agent data and passwords stay on company systems.

Questions to Claude still go to an outside provider.