Claude Code on the Go
A mobile-first development workflow: Claude Code agents running on a cloud VM, controlled entirely from an iPhone. The setup is a Vultr VM ($0.29/hour, on-demand), Tailscale for networking, Termius + mosh for resilient SSH, tmux for persistence, and Poke for push notifications. The result: "six agents, six features, one phone."
Key Quotes
"No laptop, no desktop -- just Termius on iOS and a cloud VM."
"The loop is: kick off a task, pocket the phone, get notified when Claude needs input."
"Without notifications, you'd constantly check the terminal. With them, you can walk away."
Key Themes
The push notification loop is the key insight. Without it, mobile development is constant terminal-checking. With it, you can treat agent runs as background jobs that interrupt you only when they need something. This is the same async pattern that makes Agent of Empires's multi-agent approach practical.
The cost model ($0.29/hour, start/stop on demand via iOS Shortcuts) makes this accessible rather than expensive. Combined with git worktrees for branch isolation, you get parallel development from a phone.
Critical Analysis
This is a beautiful hack, and the author clearly uses it. The question is whether "development from your phone" is solving a real problem or creating an impressive-looking workflow for a need that rarely arises. How often do you actually need to write software from your phone? The honest answer is probably "almost never" -- but the interesting reframe is that when agent runs take hours (as in Building low-level software with only coding agents), the phone becomes a monitoring interface, not a coding interface. You're not developing from your phone; you're supervising from your phone. That's a meaningfully different and more defensible use case.
The Tailscale + mosh + tmux stack is solid and well-chosen. Each component solves a specific problem (networking, connection resilience, session persistence) without overlap.
Sources: Claude Code on the Go Last updated: 2026-05-14