Dock: Isolated Linux machines for AI agents, on your own hardware

Hey Product Hunt 👋 I built Dock because agents kept hitting the same wall: they need a real computer to work in, and most sandboxes vanish the moment a task ends. A dock is an isolated Linux machine with a persistent disk and a stable ID. You create it, run commands in it, stop it (compute is freed, the disk stays), and resume it later with the same files. My favourite part is branching: `dock branch api --count 3` gives you three copies of the same machine, so three agents can try three fixes at once and you keep the one that works. You don't need an account to try it. It runs on your own machine: curl -fsSL https://usedock.io/install | sh dock new demo dock exec demo -- uname -a (Linux with KVM, or Windows with Hypervisor Platform. `dock doctor` tells you what's missing.) If you want a fleet, you can self-host the control plane with one Docker Compose command. Hosts run docks as Firecracker microVMs or gVisor sandboxes and only connect outbound, so you don't open any inbound ports. Connectors (GitHub first) let agents work in your tools while the credentials stay outside the dock. It's open source (AGPL) and pre-release, so the interfaces can still change. Dock Cloud is coming soon. There are early SDKs for TypeScript, Python and Go. Code: https://github.com/usedock/dock and https://github.com/usedock/docklet Docs: https://docs.usedock.io I'd love to hear what you'd run in a dock, and what's missing. I'll be here all day answering questions.