# Claude Managed Agents get self-hosted sandboxes and MCP tunnels

By Steven Van · 2026-05-20

The additions let agents run in a customer's own infrastructure and reach internal services without exposing them publicly.

Claude Managed Agents now support self-hosted sandboxes and MCP tunnels, two security additions from Anthropic. Self-hosted sandboxes let an agent's execution environment run in a customer's own infrastructure or with a managed sandbox provider, rather than only in Anthropic's environment. MCP tunnels let an agent reach services sitting inside a customer's security perimeter.

Together these give teams running [Claude](<https://creatorstoolbox.com/tools/claude>) agents on internal systems more control over where code executes and how it reaches internal services, without exposing those services publicly. Full details are in the [announcement](<https://x.com/i/status/2056740346529468717>).

![](<https://pbs.twimg.com/media/HIr-GW0bUAA8VvA.jpg>)

![Claude](<https://claude.com/images/icon-512x512.png>)

Claude

AI assistant by Anthropic — write, analyze, code, and create with one of the most capable large language models available.

[View Claude →](<https://creatorstoolbox.com/tools/claude>)

[Original source](<https://claude.com/blog/claude-managed-agents-updates>)

[Read Claude Managed Agents get self-hosted sandboxes and MCP tunnels on Creators Toolbox](<https://creatorstoolbox.com/blog/claude-self-hosted-sandboxes-mcp-tunnels-for-managed-agents>)

---
Canonical source: https://creatorstoolbox.com/blog/claude-self-hosted-sandboxes-mcp-tunnels-for-managed-agents