How to use Claude to build an iOS app from your phone
A couple remotely built a Japanese phrase app with Claude Code from an iPhone, using persistent SSH sessions, TestFlight deployments, and feedback from restaurants and shops.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Help build [iOS phrase app] through this remote terminal session. Add [tourist phrases] and text to speech, refine [UI text], and use [existing TestFlight pipeline] for a test build. Revise the app based on [observations from real use].
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- lordokami
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reported shipping a phrase app that was stable, usable, and relied upon by the end of the trip, developed remotely without physically accessing the Mac.
- Human involvement
- The couple tested builds in real restaurants and shops, identified friction, and directed revisions throughout the trip.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Configured an always-awake Mac with VPN, SSH, persistent sessions, and TestFlight deployment.
- Connected from an iPhone and dictated prompts, interface revisions, and feature requests.
- Triggered remote builds and tested the phrase app in restaurants and shops.
Before traveling from Spain to Japan, the author prepared a Mac for remote development with Claude Code. They disabled sleep, configured VPN access for SSH from a phone, installed Zellij for persistent terminal sessions, and set up a TestFlight deployment pipeline. Remotely triggered builds could be tested about fifteen minutes later.
The couple chose to make a lightweight Japanese phrase app with tourist sentences and built-in text to speech. During the trip, the author's wife dictated changes into Terminus on an iPhone connected to the Mac at home. Voice input supplied prompts, revised interface text, and requested features.
Development happened in short sessions during travel stops and train rides. They shipped builds, tried the app in restaurants and shops, identified awkward phrasing or friction, and made further changes that evening. The next day provided another opportunity to test the revisions in actual use.
The author reported never opening Xcode locally or physically touching the Mac during the trip. By the end, they described the app as stable, usable, and something they relied on.
Who it's for: Developers interested in phone-based remote workflows with frequent real-world app testing.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · lordokami
Frequently asked questions
What kept the remote development sessions available?
The Mac was configured not to sleep, VPN and SSH provided access, and Zellij maintained persistent terminal sessions if the connection dropped.
How did the couple test changes during the trip?
They triggered TestFlight builds remotely, used the app in restaurants and shops, noticed friction, revised it later, and tested again the following day.
More details
Remaining steps
- Revised awkward phrasing and other friction, then redeployed for further use.
About this account
Claude Code supported remote app development through human-dictated prompts and repeated feature and wording requests.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=47011100. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: lordokami on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Mac · VPN · SSH · Zellij · Terminus · TestFlight
iOS · Remote development · Voice input · TestFlight
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.