Anthropic opens the self-hosted Claude Managed Agents beta: the agent runs in your cloud
Direct answer
On November 12, 2026, Anthropic opened a private beta of self-hosted Claude Managed Agents. Companies run the agent server, its database with session history, the workers that run shell and file tools, credentials and sandboxes in their own cloud or on premises. Model calls go where the company chooses: Anthropic, Amazon Bedrock, Google's agent platform, Microsoft Foundry or its own gateway. The agent control plane lives in the customer's infrastructure, and the execution sandbox does too.
What stays inside your cloud
- The agent server and its database: session history lives in the database running in your environment, with the workers that run the agent's tools.
- The credentials: the keys the agent uses to touch your systems are stored and managed in your infrastructure, under your access controls.
- The sandboxes: the isolated environments where the agent runs code and files run in your cloud or on your premises.
- The model gateway: you choose the provider for model calls, among Anthropic, Amazon Bedrock, Google, Microsoft Foundry or your own gateway.
Why this matters for your business
- Data residency becomes your decision: session history and credentials stay where your policy says they must stay, under the controls your team already audits.
- The agent stops being an opaque service: when the control plane runs in your cloud, your team sees sessions, permissions and logs with its own tools.
- Dependence on a single provider loosens: the model gateway is interchangeable. The company using one provider today can move calls to another with the same control plane.
- The beta is private and by application: access is limited for now. The interested company applies and Anthropic decides entry.
What to do in your business this week
- List your agents and where each piece runs: server, history, credentials and sandboxes. The inventory is the first step to deciding what moves into your cloud.
- Set your data residency policy for agents: which session data may leave, which stays inside and who audits the history. The written policy comes before the tool.
- Ask your current provider about the control plane: where session history lives and who holds the credentials. Today's answer is your baseline.
- Evaluate the beta if your industry demands full control: banking, health care and regulated services are the first candidates. Applying is the concrete step.
Note: this article summarizes Anthropic's announcement of November 12, 2026 and coverage in SXF Brief and The Operators Pulse, with reporting from The Register. The beta is private and by application. Pricing details and general availability have not been published.
Want AI that works for your business?
Message us on WhatsApp and let's talk about automation and AI agents for your business.
WhatsApp inicIARelated articles
- Claude becomes your workspace: Whiteboards and Sheets arrive in beta
- ChatGPT Enterprise comes to AWS Marketplace
- The 7 types of AI agents for businesses (and which one yours needs)
Frequently asked questions
What are self-hosted Claude Managed Agents?
It is a private Anthropic beta, opened on November 12, 2026, that lets a company run the agent server, its database with session history, the workers, credentials and sandboxes in its own cloud or on its premises.
Where do model calls go?
Where the company chooses: Anthropic, Amazon Bedrock, Google's agent platform, Microsoft Foundry or its own gateway.
What changes compared with using Claude in Anthropic's cloud?
The agent control plane, with session history and credentials, lives in the customer's infrastructure, under its own access controls and audit.
How do you join the beta?
It is a private beta by application. Anthropic decides entry. Pricing and the general availability date have not been published.
