Claude Code On-the-Go

Why It Matters

The setup proves that full‑stack AI‑assisted development can be performed from a mobile device, cutting infrastructure costs and freeing developers from a fixed desk. It showcases a secure, pay‑per‑use model that can reshape remote coding workflows.

Key Takeaways

• Run Claude Code agents entirely from a mobile phone
• Use Tailscale VPN and mosh for resilient, secure VM access
• Pay‑per‑use Vultr VM costs $0.29 per hour, $7 daily
• Push notifications via webhook let developers respond asynchronously
• Parallel agents with tmux and git worktrees enable multitasking

Pulse Analysis

Mobile‑first development is no longer a niche experiment; it is becoming a practical reality for AI‑augmented coding. By leveraging a lightweight iOS SSH client (Termius) paired with mosh, developers maintain persistent shell sessions across network changes, while Tailscale provides a zero‑trust private network that isolates the cloud VM from the public internet. This combination delivers a resilient, secure bridge between a phone and a disposable compute instance, dramatically reducing the hardware footprint required for Claude Code workflows.

The architecture hinges on a pay‑per‑use Vultr VM that can be spun up or shut down with a single tap via an iOS Shortcut. At $0.29 per hour, the cost remains bounded even if an agent runs unchecked, and the VM’s firewall plus fail2ban add layers of defense. Inside the VM, tmux preserves session state, and git worktrees allow multiple feature branches to run concurrently, each paired with its own Claude agent. A custom PreToolUse hook posts questions to Poke’s webhook, turning Claude’s prompts into push notifications that the developer can answer on the go, eliminating the need for constant terminal polling.

The broader impact is a shift toward asynchronous, pocket‑sized development cycles. Engineers can initiate long‑running AI‑driven tasks during idle moments—on a train, in a coffee shop, or while watching TV—and resume with a single tap when notified. This model not only slashes idle compute costs but also democratizes access to powerful AI coding assistants, potentially accelerating adoption across startups and enterprises that value flexibility and cost efficiency.