One ssh and a Firecracker microVM boots the instant you connect — a real Linux kernel, hardware-isolated, sub-second. Your SSH key is your identity. No signup, no console, no dashboard.
ssh box@box.hopbox.dev
Try it right now, no account: ssh demo@box.hopbox.dev
$ ssh box@box.hopbox.dev spawning microVM · firecracker · isolated ✓ booted in 0.4s — Linux 6.1 · 2 vCPU · 1 GiBbox@box:~$ uname -r && docker run hello-world 6.1.0 · Hello from Docker! # disconnect → auto-suspends to disk; reconnect → restoredbox@box:~$
No provisioning step, no waiting on a queue. The connection is the lifecycle.
Your public key is your identity. Per-owner box names: your myproject is yours alone.
A Firecracker microVM boots in well under a second — a real kernel, hardware-isolated, not a container.
Disconnect and the box auto-suspends: memory + state snapshot to disk. Nothing burning while you're away.
The next connection restores the snapshot instantly. Your durable disk is exactly where you left it.
Real virtualization, terminal-native ergonomics, secure by default.
Firecracker microVMs with their own Linux kernel. Run Docker, nested workloads, even load kernel modules — hardware isolation, not namespaces.
Idle boxes snapshot memory and state to disk and auto-suspend; the next SSH restores them instantly. You only pay attention when you're connected.
Your disk survives suspend/resume, daemon restarts, and host reboots. Pick up exactly where you left off.
Boot from a tooled base: ssh proj:ubuntu-22.04@box.hopbox.dev or proj:debian-12@…. List them with ssh images@box.hopbox.dev — git, vim, curl, tmux included.
A fresh throwaway box, reaped a couple of minutes after you disconnect. Register your key for the persistent tier when you want it to stick.
No web dashboard, no cloud console, no signup form. The SSH key in your agent is the whole auth story — per-owner box names keep them separate.
An in-box CLI — box-guest info, keep-alive, idle tuning — and an MCP server (box-guest mcp) so an AI agent inside the box can manage its own sandbox.
Per-box network isolation and an egress firewall out of the box. Each box is sealed off from the others and from the host.
One command on a KVM-capable Linux host stands up the boxd daemon. It's open source — bring your own hardware.
curl -fsSL https://box.hopbox.dev/install.sh | sudo sh
Source & issues on GitHub · docs at docs.hopbox.dev