A lightweight, cross-platform Virtual Machine Monitor (VMM) designed for securely running AI agents on personal computers. It addresses the need for isolated execution environments, protecting user data by strictly controlling file access. The VMM enables full Linux desktop environments with hardware-accelerated virtualization, GPU, audio, and seamless host-guest integration, benefiting users who require sandboxed AI workloads without compromising privacy or performance.

How It Works

TenBox employs a shared C++ runtime VMM architecture, abstracting underlying hypervisors. It supports hardware-accelerated virtualization via native backends: WHVP on Windows, Apple's Hypervisor Framework on macOS (Intel/Apple Silicon), and KVM on Linux (x86_64/arm64). This approach ensures efficient, near-native performance for guest OSes. Key integrations include virtio-gpu for display, virtio-snd for audio, virtiofs for shared folders, and SPICE protocol for clipboard sharing and display resizing, facilitating a rich user experience within the isolated VM.

Quick Start & Requirements

• Linux: Execute curl -fsSL https://tenbox.ai/install.sh | sudo sh. Requires Debian 11+/Ubuntu 20.04+/Raspberry Pi OS 11+ (amd64/arm64), glibc 2.31+, and /dev/kvm. The installer configures the system, adds the user to the tenbox group, and enables the tenboxd service.