Claude Managed Agents Can Be Self-Hosted in Private Beta
Anthropic says eligible organizations can now run the Claude Managed Agents service in their own cloud or on-premises in a private beta, including the harness, sandboxes, credentials, and session history. The article also covers the earlier self-hosted-sandbox and MCP-tunnel announcements.
Anthropic says Claude Managed Agents can now be self-hosted in a private beta. Eligible organizations can run the service in their own cloud or on-premises and choose where model requests are sent. This is broader than Anthropic's earlier self-hosted-sandbox option, which moved tool execution to a customer-controlled environment while keeping the agent loop on Anthropic's infrastructure. (Anthropic's self-hosted Managed Agents article; Anthropic's May 2026 announcement)
November 12, 2026 update: Full-service self-hosting (private beta)
Anthropic says every part of the Managed Agents service can run in a customer's own cloud or on-premises environment, including the harness, sandboxes, credentials, and session history. Customers operate the Managed Agents server, its database, and the workers where agents run shell and file tools. (Anthropic's self-hosted Managed Agents article)
The private beta gives organizations control over where model requests are routed. Anthropic lists direct requests to Anthropic, a cloud provider such as Amazon Bedrock, Google's Agent Platform, or Microsoft Foundry, and a customer-operated gateway. Developers build agents with the same core APIs as the hosted Managed Agents service. (Anthropic's self-hosted Managed Agents article)
Anthropic says the beta is intended for developers building agents that work directly with company code and internal systems. Eligible organizations can ask their Anthropic account team for access. (Anthropic's self-hosted Managed Agents article; ClaudeDevs on X)
This is distinct from the earlier self-hosted-sandbox option: that feature moves tool execution into a customer-controlled sandbox, while the agent loop stays on Anthropic's infrastructure. The November beta lets customers host the broader Managed Agents service itself. (Anthropic's May 2026 announcement; Anthropic's self-hosted Managed Agents article)
May 2026: Self-hosted sandboxes (public beta)
Anthropic says self-hosted sandboxes let Claude Managed Agents run tool execution inside an environment you control.
Key details Anthropic describes:
- The sandbox can run on your own infrastructure, or through managed providers like Cloudflare, Daytona, Modal, or Vercel.
- The agent loop for orchestration, context management, and error recovery stays on Anthropic's infrastructure, while tool execution moves to your configured environment.
- Anthropic says this helps keep files and repositories within your perimeter and lets you apply existing network policies, audit logging, and security tooling.
- You control compute characteristics like resource sizing and the runtime image.
May 2026: MCP tunnels (research preview)
Anthropic says MCP tunnels let agents connect to private Model Context Protocol (MCP) servers inside your network without exposing them to the public internet.
Key details Anthropic describes:
- A lightweight gateway you deploy makes a single outbound connection.
- Anthropic says there are no inbound firewall rules required and no public endpoints, and that traffic is encrypted end-to-end.
- Anthropic says MCP tunnels work with both Managed Agents and the Messages API.
- Anthropic says MCP tunnels are managed from workspace settings in the Claude Console by organization admins.
Availability stated by each announcement
| Feature | Status announced | Notes |
|---|---|---|
| Self-hosted Managed Agents service | Private beta (November 12, 2026) | Contact the Anthropic account team to request access |
| Self-hosted sandboxes | Public beta (May 19, 2026) | Available on the Claude Platform |
| MCP tunnels | Research preview (May 19, 2026) | Request access required |
Read next
Related coverage- Claude
Claude Opus 4.7 Launch: Stronger Coding, xhigh Effort, and a New Cyber Safeguards Tier
Anthropic launched Claude Opus 4.7 on April 16, 2026, with gains over Opus 4.6 in advanced software engineering at the same $5/$25 pricing, a new xhigh effort level, /ultrareview, higher-resolution vision and the first cyber safeguards. It is now a legacy model, succeeded by Opus 4.8, Opus 5 and Opus 5.5.
- Claude
Claude Design Launches: Anthropic Labs Turns Opus 4.7 Into a Prototype, Deck, and Wireframe Surface
Anthropic launched Claude Design on April 17, 2026, an Anthropic Labs tool on Claude Opus 4.7 that turns a prompt, image, or codebase into prototypes, decks, and mockups. As of October 2026 it is out of beta and on every plan, Free included, as a template in any Claude chat. The standalone claude.ai/design closes December 14, 2026.
- OpenAI
OpenAI's Secure MCP Tunnel: Connect Private MCP Servers Over Outbound-Only HTTPS
OpenAI's Secure MCP Tunnel connects private and on-prem MCP servers to ChatGPT, Codex, and the Responses API without opening inbound firewall ports. You need a tunnel_id, an API key with Tunnels Read plus Use, and -- separately -- ChatGPT developer-mode access. That permission split is the most common setup blocker.
- Claude
Claude Enterprise Auth: A Beta Claude-Issued Path for MCP Connectors
An August 24 update to Anthropic's June post made identity-provider-managed auth generally available. Current documentation also describes Claude-issued assertions in beta, with important differences in user identity, SSO eligibility, and access revocation.
Sources
- ClaudeDevs on X, “Self-hosted Claude Managed Agents is now in private beta,” November 12, 2026: https://x.com/ClaudeDevs/status/2120938988165091693
- ClaudeDevs on X, beta eligibility and access details, November 12, 2026: https://x.com/ClaudeDevs/status/2120938990518116578
- Anthropic, “Operate Claude Managed Agents inside your security perimeter,” November 12, 2026: https://claude.com/resources/articles/self-hosted-claude-managed-agents
- ClaudeDevs announcement post (May 19, 2026): https://x.com/ClaudeDevs/status/2056740346529468717
- Claude blog post, "New in Claude Managed Agents: self-hosted sandboxes and MCP tunnels" (May 19, 2026): https://claude.com/blog/claude-managed-agents-updates
Frequently Asked Questions
What are self-hosted sandboxes in Claude Managed Agents?
Anthropic says self-hosted sandboxes let Claude Managed Agents operate in a sandbox you control, running on your own infrastructure or with managed providers like Cloudflare, Daytona, Modal, or Vercel.
What are MCP tunnels?
Anthropic says MCP tunnels let agents reach private Model Context Protocol (MCP) servers inside your private network without exposing them to the public internet, using a lightweight gateway that makes a single outbound connection.
What can an organization self-host in the new Managed Agents beta?
Anthropic says the private beta lets organizations run the Managed Agents service in their own cloud or on-premises, including the harness, sandboxes, credentials, and session history. Customers operate the Managed Agents server, its database, and the workers where agents run shell and file tools. ([Anthropic's self-hosted Managed Agents article](https://claude.com/resources/articles/self-hosted-claude-managed-agents))
How can an organization request access to self-hosted Managed Agents?
Anthropic says the private beta is available to eligible organizations. Developers can contact their Anthropic account team to request access. ([ClaudeDevs on X](https://x.com/ClaudeDevs/status/2120938990518116578); [Anthropic's self-hosted Managed Agents article](https://claude.com/resources/articles/self-hosted-claude-managed-agents))
What availability statuses did Anthropic announce for these features?
Anthropic's November 12, 2026 announcement places full-service self-hosted Managed Agents in private beta. Its May 19, 2026 announcement listed self-hosted sandboxes as public beta and MCP tunnels as a research preview requiring an access request.