Understand this story
On November 12, 2026, Anthropic launched a self-hosted version of Claude Managed Agents, currently in private beta. This version can run agents in an enterprise's own cloud or on-premises environment, keeping the harness, sandbox, credentials, and session history entirely within the organization's own environment. Model requests can be sent to Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry, or a self-built gateway, and session history is stored in the organization's own database. The company also described four security design features: separation of the harness from the sandbox, credentials not entering the sandbox, network access restricted by an allowlist of hosts, and export of complete session records via OpenTelemetry. The next day, Anthropic's Claude Devs further explained that this version is aimed at tasks requiring access to code and internal systems, such as fixing build failures or investigating alerts, and is currently open only to eligible organizations, which must apply through Anthropic's customer team.
AI-generated from public reports
Latest update · Discussion
Ado Kukic joined this discussionDevelopments
1 confirmed developments
Anthropic launches private beta of self-hosted Claude Managed Agents
Source:Claude Blog
Reports and discussion timeline
Browse reports and discussion by time. Their evidence levels remain distinct.
Nov 13
Ado KukicRelated discussion
Ado Kukic joined this discussion
This post contributes to heat; it is not a formal report or confirmed fact.
Claude DevsFirst-party
Anthropic's Self-Hosted Claude Managed Agents Enter Private Beta
Anthropic announced that self-hosted Claude Managed Agents have entered private beta. Users can deploy the agent runtime framework, sandboxes, and session history in their own cloud or on-premises environment, and choose where Claude requests are sent. This feature is aimed at tasks that require access to code and internal systems, such as fixing build failures or triaging alerts. It is currently available only to eligible organizations and requires an application through the Anthropic customer team.
Claude DevsFirst-party
Claude Managed Agents Launches Self-Hosted Private Beta
Anthropic's Claude Devs announced that the self-hosted version of Claude Managed Agents has entered private beta. Users can deploy the agent runtime, sandboxes, and session history in their own cloud or on-premises, and choose for themselves where requests are sent to Claude. This version is aimed at tasks that require access to code and internal systems, such as fixing build failures or investigating alerts. It is currently open only to eligible organizations, which must apply through the Anthropic customer team.
Unite.AI
Anthropic Launches Self-Hosted Claude Managed Agents
On November 12, 2026, Anthropic launched self-hosted Claude Managed Agents, allowing customers to run the complete service and its sandboxes in their own cloud accounts or on-premises data centers, with agents, sessions, and session history all stored on that infrastructure. The core component cma-server provides the Managed Agents API, runs the agent loop, and calls the chosen inference provider; model calls go through the Claude API by default, but can alternatively use Claude models on Amazon Bedrock or Google Cloud, Microsoft Foundry, or a self-built model gateway. Migration does not copy any data from the managed service; customers must rebuild agents, skills, credentials, and memory storage themselves, and take on server security, tenant isolation, and version installation, while features such as scheduled deployments, dreams, and webhooks are unavailable.
Nov 12
Claude BlogFirst-partySelected
Anthropic Launches Self-Hosted Claude Managed Agents Private Beta
Anthropic has launched self-hosted Claude Managed Agents, currently in private beta, which can run agents in an enterprise's own cloud or on-premises environment. This version places the harness, sandbox, credentials, and session history entirely in the customer's own environment; model requests can be sent to Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry, or a self-built gateway, and session history is stored in the customer's own database. The company also described four security design points, including separation of the harness and sandbox, keeping credentials out of the sandbox, restricting network access based on an allowed-host list, and exporting complete session records via OpenTelemetry.
Heat trend for this story
- Current
- 8
- Period peak
- 29 Nov 13 03:00
- 24-hour change
- -50%
Move the pointer or select the chart to inspect hourly heat. Use the left and right arrow keys with a keyboard.
AI Trends