Claude Managed Agents now run inside company infrastructure

Claude Managed Agents can now run inside a company’s own infrastructure instead of Anthropic’s hosted environment. The update adds self-hosted sandboxes in public beta and MCP tunnels in research preview, letting agents execute code and reach internal tools without exposing those systems to the public internet.

This targets a core blocker for enterprise use: keeping files, credentials, and services inside an existing security perimeter. The setup supports things like zero-trust credential injection, custom egress controls, VPC access, and observability, with sandbox partners including Cloudflare, Daytona, Modal, and Vercel, which confirmed its integration through Vercel Sandbox.