Use Claude Code from your phone without a VPS

You start a run, walk away, and Claude Code stops on a single question you weren't there to answer. The usual advice is to move the whole thing to a cloud box you can SSH into. Sometimes that's right. Often it's a heavy fix for a small problem. Here's the honest menu, and what each one actually feels like at 9pm on a train.

Option 1 — SSH from a phone terminal (Termius, Blink)

Install a terminal app, add your key, ssh into the Mac or server, and reattach to a tmux session. It genuinely works and it's the most flexible: you get a real shell.

The catch: typing a paragraph-long prompt on a phone keyboard, inside tmux, over a flaky mobile connection, is miserable. And nothing pushes you when the agent stops. You still have to keep opening the app to check whether it's waiting on you. The tool that keeps the run alive doesn't tell you it needs a decision.

Option 2 — VPS + tmux (run it in the cloud)

Spin up a Linux box, run Claude Code there in tmux, SSH in from anywhere. The Mac can sleep; the run doesn't. Popular for exactly that reason.

The catch: your repo, your .env, and your credentials now live on a rented server instead of your laptop. There's a monthly bill. The dev environment on the VPS drifts from your real one. And you are still babysitting a terminal through a phone, just a remote one. For a lot of people the whole appeal of Claude Code was that it runs on their machine, against their local checkout. A VPS quietly gives that up.

Option 3 — TeamViewer / RDP / VNC (remote desktop)

Mirror the Mac's screen to your phone and tap around. Zero setup on the code side.

The catch: remote desktop on a 6-inch screen is a blunt instrument. You're pinching and zooming to hit a "2. Yes, and don't ask again" that a form could have shown as a button. It drains battery on both ends, the latency makes text selection painful, and it's overkill when all you needed to send back was "option B."

Option 4 — Tailscale + SSH (the nice-networking version)

Tailscale gives every device a stable private address, so you can reach your Mac from your phone without port-forwarding or a public IP. If you do want a real remote shell, this is the clean way to get one, and it's a great tool.

The catch: it fixes the networking, not the ergonomics. Once you're in, you're still hand-typing prompts and approving plans through a phone terminal. Tailscale gets you to the door; the awkward part is still on the other side of it.

Notice the pattern. Every option above is really about keeping a terminal reachable. But most of the time you don't want a terminal on your phone. You want the one thing the agent is stuck on: a plan to approve, a permission to grant, a multiple-choice question to answer. That's a much smaller surface than a full shell, and it deserves a much smaller tool.

Option 5 — Answer it locally, no VPS, no SSH (disclosure: this is us)

DevFob is our product, so weigh accordingly. It takes the opposite bet from everything above: instead of shipping a terminal to your phone, it ships the decision to your phone and leaves the run exactly where it is, on your Mac.

The honest boundary: if what you truly need is a full shell on the go — run arbitrary commands, edit files, kick off new work from scratch away from your desk — then SSH or Tailscale is the right tool and DevFob is not a replacement for it. DevFob is for the far more common case: the run is already going, and it just needs an answer from you to keep going. If that's your problem, you don't need a VPS to solve it.

Try DevFob free — 14 days $15 once · no account · no VPS · macOS 14+

Answering from the phone is Claude Code today; Codex and Grok are monitor-and-switch. Tool details accurate as of July 2026. Something wrong here? Tell us.