Anthropic Lets Customers Self-Host Claude Managed Agents
What happened
On November 12, 2026, Anthropic released self-hosted Claude Managed Agents, allowing organizations to run the full agent service and sandboxes on their own cloud or on-premises infrastructure. Customers now operate the agent API, agent loop, and session storage themselves, with model inference routed to a provider of their choice, such as Anthropic's Claude API, Amazon Bedrock, Google Cloud, or Microsoft Foundry. All agent data and execution remain within the customer's environment, but model calls and conversation data are still sent to the selected inference provider. Customers are responsible for setup, security, and ongoing management of their deployments.
Key details
- Self-hosted Claude Managed Agents launched November 12, 2026.
- Customers can choose inference providers: Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry, or a self-operated gateway.
- All agent data, session history, and execution are stored and run on customer infrastructure.
- A persistent deployment requires PostgreSQL 16 or later; without it, data is not retained after server restarts.
- Some hosted features, like scheduled deployments and webhooks, are not available in the self-hosted version.
Why it matters
Enterprise customers and organizations that require more control over their data and infrastructure can now keep agent state and execution within their own environments, reducing reliance on Anthropic's hosted service. However, they must take on new responsibilities, including securing the server, managing data retention, and handling updates. This change gives these organizations more flexibility and privacy, but also increases their operational workload and limits access to certain features present in the hosted service.