AI sandbox that runs on your homelab (github.com)

A new AI sandboxing solution has been announced, designed for home lab environments using TrueNAS and Incus. This platform allows users to create disposable Linux containers pre-loaded with various AI coding tools such as Claude Code, Codex, and OpenCode. Each container is equipped with SSH access, ZFS snapshot checkpoints, and customizable network egress policies, enabling developers to control what their AI agents can access. The system is managed entirely via command line interface (CLI) through the TrueNAS WebSocket API, making it user-friendly yet powerful. This development is significant for the AI/ML community as it facilitates rapid prototyping and experimentation with AI coding agents in a secure and isolated environment. By allowing multiple containers to be spun up from checkpoints and controlled with specific network permissions, developers can safeguard their systems while exploring new AI integrations and applications. The use of ZFS for snapshot management further enhances the flexibility of testing different approaches without permanent changes. This tool, leveraging Go programming and containerization principles, offers an innovative way to streamline AI development pipelines right from a personal setup.