Agent: Cursor, Claude CodeLLM: Claude 3.5, GPT-4#AI safety#Sandboxing#Code execution#Security#LLM tools
MXC is a cross-platform sandboxed execution system designed for running untrusted code—including AI model outputs and tool plugins—safely on Windows, Linux, and macOS. It offers multiple containment backends (from OS-native process sandboxes to full VMs) behind a unified JSON config and TypeScript SDK, with policy-driven security controls for filesystem, network, and UI access.