Back to list

Blog

From Desk to Couch: The Mobile Developer Workflow

The phone isn't replacing your laptop. It's filling the gaps between desk sessions — and those gaps are where AI coding agents add the most value.

Published Tags: workflow / productivity / mobile / seo
Blog

Follow product and engineering updates from this channel.

Browse category

The "mobile developer workflow" sounds like a pitch for working from a beach with an umbrella drink. The reality is more mundane and more useful: you're on your couch after dinner, your Mac is in the office, and you want to kick off that test refactoring before tomorrow morning.

This article describes how developers are actually using phone-based coding tools — not as a laptop replacement, but as an extension of their desk setup that fills specific gaps in the workday.

The Gaps in a Developer's Day

Most developers don't code for 8 straight hours. The day has natural breaks:

  • Commute (30-60 min) — can't code, could monitor or delegate
  • Meetings (intermittent) — can't code, could check status
  • Lunch (30-60 min) — could review results or start the next task
  • Evening (variable) — might want to check on something or start a low-priority task
  • Waiting for builds/tests (variable) — time exists, screen access doesn't

These gaps are when phone-based coding tools are genuinely useful. You're not trying to write complex algorithms on a 6-inch screen. You're delegating to an AI agent, monitoring progress, and making decisions.

Three Workflow Modes

Mode 1: Delegate and Forget

When: Commute, meetings, any time you can't actively monitor

How it works:

  1. Send a clear, self-contained task to Claude Code from your phone
  2. Put the phone away
  3. Get a push notification when Claude finishes
  4. Review the result when you have time

Good tasks for this mode:

  • "Run the full test suite and fix any failures"
  • "Update all dependencies to their latest stable versions"
  • "Add TypeScript types to the utility functions in src/utils/"
  • "Generate unit tests for the PaymentService class"

Bad tasks for this mode:

  • Anything requiring design decisions Claude can't make alone
  • Tasks where you need to see intermediate output to guide direction
  • Work involving credentials or environment-specific setup

Mode 2: Supervise and Steer

When: Couch time, lunch break, any low-focus period where you can glance at your phone

How it works:

  1. Send a task that might need guidance
  2. Keep your phone nearby
  3. When Claude asks for input or approval, respond
  4. Iterate until the task is done

Good tasks for this mode:

  • "Refactor the authentication flow — I'll review each change"
  • "Write the README for the new module — let me approve the structure first"
  • "Set up the CI pipeline — I'll confirm the configuration choices"

This mode works because approval prompts on Tactic Remote have dedicated Y/N buttons on the keyboard toolbar. You don't need to type "yes" — it's a single tap. Quick enough to do while watching TV.

Mode 3: Quick Burst

When: You have 2-5 minutes and want to accomplish something specific

How it works:

  1. Open Tactic Remote
  2. Do one specific thing
  3. Close the app

Good tasks for this mode:

  • Check if the build is passing
  • Review Claude's output from an earlier task
  • Send a quick commit message
  • Start a background task for later review
  • Check token costs across sessions

Snippets are essential for this mode. One tap for "git status", one tap for "run tests", one tap for "/cost". No typing required for routine checks.

The Daily Rhythm

Here's what a typical day looks like for a developer using this workflow:

8:00 AM — At desk Start the Mac app, review yesterday's work, plan today's tasks. This is full-desk mode — you have your keyboard, monitor, and focus.

8:30 AM — Send first task Before diving into your own work, delegate a task to Claude: "Fix the three bugs from yesterday's triage." Claude starts working in a session.

9:00 AM - 12:00 PM — Desk work + monitoring You're working on your own code at your desk. Occasionally your phone buzzes:

  • "Task completed" — glance at it, send the next task
  • "Approval needed" — review, tap Y, back to work
  • "Error" — check the output, send a correction

12:30 PM — Lunch On the couch with your phone. Check session status: Claude finished 2 tasks, 1 in progress. Send a new task: "Add integration tests for the endpoints we changed this morning." Back to eating.

1:00 PM - 3:00 PM — Meetings Phone on silent. Claude is working. Two notifications during the meeting — you'll check them after.

3:15 PM — Between meetings Two tasks completed. Review output for 3 minutes. One looks good, commit it. One needs a tweak — send a correction. Start a documentation task.

5:30 PM — Commute On the train. Check status: everything pushed, CI is green. No action needed. Good day.

8:00 PM — Couch Low-priority task: "Clean up the test fixtures and remove unused data." Delegate and forget. Notification will come whenever — no rush.

10:00 PM — Before bed Glance at phone: task completed, 3 files modified, tests pass. Don't even open the app — tomorrow-morning review.

What Changes in Your Workflow

You Think in Tasks, Not Keystrokes

When your interface is a phone, you naturally compose higher-level instructions. Instead of manually typing commands and editing files, you describe what you want done. This is how AI coding agents work best — clear intent, not micromanagement.

You Batch Decisions

Instead of approving each step as Claude works, you might let a few notifications accumulate and review them together during a break. This is more efficient than context-switching every few minutes.

You Work More Asynchronously

The phone workflow is inherently asynchronous. You send work, you get notified, you respond. This maps well to how AI agents actually operate — they're not waiting for your next keystroke, they're executing a task.

You Code More Hours (Sort Of)

You're not coding more at your desk. But the total hours of productive coding output increase because Claude is working during your gaps. The 30-minute commute that was dead time now produces a completed refactoring task.

Honest Limitations

Small Screen, Limited Review

Reviewing code changes on a phone screen is viable for small diffs but painful for large ones. If Claude modified 15 files, you probably want to review that at your desk.

Workaround: Use Claude to summarize its own changes: "List all files you modified and summarize what changed in each." This gives you a phone-friendly overview.

Typing is Slow

Composing detailed prompts on a phone keyboard is slower than on a real keyboard. Keep prompts concise when on your phone.

Workaround: Custom snippets for common prompts. If you frequently start sessions with similar instructions, save them.

No Visual Context

You can't see your IDE, browser, or design tools from your phone. Tasks that require visual verification ("make the button blue", "check if the layout looks right") need your desk.

Network Dependency

The phone workflow requires an internet connection (for Cloudflare Tunnel) or being on the same WiFi (for LAN mode). No connection means no coding.

Mitigation: Claude Code keeps running inside its tmux session regardless of your connection. If you lose connectivity, you lose monitoring — not work.

The Right Mental Model

The phone isn't a tiny laptop. It's a remote control with notifications.

You wouldn't try to edit a movie on a TV remote. But you'd absolutely use it to start playback, skip tracks, and adjust volume. That's the role of the phone in a mobile developer workflow — control, monitor, and decide. The heavy processing happens on your Mac.

Once you internalize this, the workflow stops feeling like a compromise and starts feeling like an upgrade: your Mac works while you don't.

Try Tactic Remote

Control your coding Agents from your phone

Connect to Claude Code, Codex, and other Agents on your Mac, Windows, or Linux computer. Check progress and send the next instruction from iPhone or iPad.