The Mac mini for your agents, without the Mac mini.
Buying a Mac mini to run agents around the clock is a reasonable idea. Here’s what it involves, and what Cube does instead.
Why people buy a Mac mini for agents
Agents are most useful on long tasks, and long tasks need a computer that stays on. A Mac mini on a shelf is always on, costs little to run, and keeps your laptop free. It’s a popular plan.
What the Mac mini plan involves
- Buying the hardware up front, and replacing it when you outgrow it.
- Keeping it online: power cuts, router restarts and macOS updates all stop your agents.
- Reaching it from outside: a VPN, Tailscale or port forwarding, set up and kept working.
- Setting it up: repos, credentials, SSH keys, a tmux or session manager so agents survive disconnects.
- macOS for servers: some developers find it a poor fit for lots of agents in parallel.
What Cube does instead
Cube is a full Linux computer in the cloud with an app on your Mac.
- Always on, with nothing to plug in.
- Reachable anywhere, from the Mac app, from your phone, or over SSH (
ssh cube), set up for you. - Ready for agents: sign in with GitHub, install Claude Code, Codex and others from a list.
- Worktrees for parallel work, and a sidebar for every session.
- Easy to leave: download your files or reset the machine from the app.
Questions
Is a Mac mini a good machine for running coding agents?
It can be. It’s quiet and efficient. But it’s hardware you have to buy, keep powered and online, reach from outside your home network, and maintain. Many agent tools also run more smoothly on Linux than on macOS.
Do I need macOS to run Claude Code or Codex?
No. Claude Code, Codex and most coding agents run on Linux, which is what Cube’s cloud computer runs. You drive it from the Cube app on your Mac.
What happens to my work if I stop using Cube?
You can download your files from the app at any time: your repos, worktrees, Git history and hidden files.