π©
A security-first sandboxing tool built by Anthropic for Claude Code. Enforces filesystem and network restrictions on AI agents and MCP servers using native OS primitives, enabling safer agentic systems without containers. Secure-by-default with minimal access and explicit allowlisting.