How to manage your VPS with AI from any device

Most developers keep a terminal window open to their VPS all day — checking disk space, tailing logs, restarting services after a deploy. With an MCP relay you can do all of that in plain English from ChatGPT or Claude, including from your phone.

The traditional way vs the AI way

The traditional way: open a terminal, SSH in, type commands, copy output, repeat. The AI way: type "Is nginx running and are there any errors in the last 100 log lines?" and get a clean summary back.

Both do the same thing on your server. The difference is that the AI approach works from any device, composes multiple checks into one answer, and lets you describe the problem rather than remember the exact commands.

What you need

• A Linux VPS (Ubuntu 20.04+ or Debian 11+).
• Node.js 22+ on the server (node --version).
• A Shellway account — mcp.getshellway.com/dashboard.
• ChatGPT Plus/Team, or Claude Pro/Team.

Set up in five minutes

1. Sign in to the Shellway dashboard and click Add computer. You get a pairing code like SW-XXXX.

2. SSH into your VPS and run:

npx shellway pair SW-XXXX
npx shellway start

The agent makes an outbound HTTPS connection to Shellway's relay. No firewall changes, no open ports.

3. To start automatically after a reboot:

npx shellway service install

4. Add https://mcp.getshellway.com/mcp as a connector in ChatGPT (Settings → Apps & Connectors) or in Claude (Settings → Connectors). Sign in with your Shellway account.

Things you can ask from your phone

• "Is my server healthy? Check disk, memory and any failed systemd services."
• "Show the last 50 nginx error log lines and tell me what the most common error is."
• "Restart the shellway service and confirm it came back up."
• "How long has this server been running and what's the current load?"
• "What's using port 3000 and is it the process I expect?"

Run a deploy from chat

You can run multi-step tasks in one message:

"Go to ~/myapp, pull the latest code from git, run npm install, restart the myapp service and tell me if it came back up healthy."

The AI runs each step in sequence, reports what it did and flags anything that failed.

Safety controls

• Run the Shellway agent as a dedicated non-root user with only the permissions it needs.
• Use --readonly mode when you only want to read logs or check status.
• Remove a computer from the dashboard at any time to cut access immediately.
• Every action is logged in the Shellway dashboard with the tool called and the result.

Ready to try? Create a free account — your trial starts when you run your first command, not when you sign up.