Your Laptop Goes to Sleep. Your AI Coding Agent Shouldn't. / xShellz

Picture an AI coding agent that lives on your laptop. It starts a long test suite. You close the lid. The tests die. You can fix that by giving the agent its own computer in the cloud, one it truly owns. An xshellz.box is a hardened Ubuntu sandbox that stays on 24/7, with real root access faked safely so you can't break anything real. And you can wire it directly into Claude Code using the Model Context Protocol. One line gets it done.

The one-liner

Once you have a box (grab one from xshellz.box, free tier never idles), copy its host and port. Then run this on your local machine:

claude mcp add xshellz -- ssh -p <port> agent@<host> xshellz-mcp

Replace <port> and <host> with your connection details. That's it. Claude Code now sees a remote tool called xshellz that can run commands, read files, and install packages inside your always-on sandbox. The box comes preloaded with git, Node 22, Python 3, ripgrep, tmux, build‑essential, and the agents Claude Code, Codex, Gemini CLI, and borg. It's ready to build anything immediately.

Why this matters for long‑running work

Before, you might keep your laptop awake or configure a VPS just to let an agent finish a task. Now you delegate the heavy lifting:

claude -p "use the xshellz sandbox to run the full test suite and store the results in /home/results.txt"

Close the lid, walk away, or lose your network. The box keeps running. You can check back later via SSH (ssh -p <port> root@<host>) or the browser web terminal. The /home directory survives reboots, so the agent can leave artifacts across sessions. If you need to reproduce the exact environment every boot, drop an xshellz.box manifest file in the home directory (one line per package, like pip: pytest); the box reinstalls it on startup.

One honest limit: MCP over SSH is a request‑response protocol. The agent won't wake you up at 2 a.m. with a build failure. If you intentionally reboot the box, running processes die, though the manifest can restore dependencies. Still, for offloading long tasks from your laptop, it's a clean way to give your agent a persistent home without babysitting.