Anthropic launches secure sandboxes and private MCPs

TL;DR AI
2 min readKey summary
Anthropic expanded Claude Managed Agents with self-hosted secure sandboxes in public beta and private MCP tunnels in research preview.
The new setup lets enterprises run agents inside their own perimeter with stronger security, compliance, and network controls.
Companies can keep sensitive data and internal service access under tighter runtime controls, including zero-trust secrets and audit logging.
Anthropic is partnering with infrastructure providers such as Cloudflare, Daytona, Modal, and Vercel to support the rollout.
