Anthropic Open-Sources OS-Level Sandbox Runtime to Secure AI Agents Without Containers
Anthropic has released sandbox-runtime, an open-source security tool that powers Claude Code. Utilizing Seatbelt on macOS and bubblewrap on Linux, the utility restricts file and network access for AI agents, CLI tools, and Model Context Protocol servers at the operating system level, eliminating the overhead of traditional containerization.
Source: MarkTechPost