Multi-OS sandbox to run AI agents with better constraints (it is not 100% secure, but enough) - View it on GitHub