How to use Claude to build an Android tracking app
The author reports a working Android MVP in four days, with background processing, push notifications, scraping, charts, and a ten-level alert system.
dev.to
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Help design [Android tracking app] for [data source] and [benchmark]. Produce scoped Markdown implementation steps for Claude Code. Investigate endpoint or build failures before fixing them, normalize datasets by [baseline date], and report builds for [emulator] and [physical device].
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- raio
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports a working Android MVP in four days, with background processing, push notifications, scraping, charts, and a ten-level alert system.
- Human involvement
- The author tested on real devices, made judgment calls, handled git, and directed changes after endpoint and database failures.
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
- Selected the Android stack and divided the project into seven steps.
- Confirmed builds on the emulator and a physical Android device.
- Replaced the unavailable CSV endpoint with dynamic HTML table scraping.
A control systems engineer with no Android development experience built ExitWatcher to examine an investment-linked life insurance product against a benchmark. The author reported a working MVP after four days, including background processing, push notifications, scraping, charts, and a ten-level alert system.
Their workflow separated design from implementation. Claude.ai handled requirements, system design, and scoped Markdown instructions. Claude Code wrote code, ran builds, and reported results. The author relayed instructions, tested on real devices, made decisions, and handled git.
On the first day, they chose Kotlin, Jetpack Compose, and Material3, then decomposed the work into seven steps. Builds passed on an emulator and a Xiaomi Mi 13T. During data-layer implementation, the planned insurance CSV endpoint returned 404, so they switched to HTML scraping with dynamic column detection. Room also failed to build; after attempting KSP, they replaced it with SharedPreferences and Gson.
The author then brought in benchmark CSV data encoded in Shift_JIS. A DataNormalizer rebased both datasets to 100 at the specified comparison date and joined them by date, making differently scaled values comparable.
Who it's for: Technically minded beginners willing to write specifications, test devices, and make architectural decisions between AI sessions.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: dev.to · raio
Frequently asked questions
How were design and coding separated?
Claude.ai handled requirements and design in Markdown instructions. Claude Code implemented scoped tasks, ran builds, and reported results; the author connected the two.
What replaced the original data and database plans?
The discontinued CSV endpoint was replaced with HTML scraping. After Room and KSP compatibility failures, the author used SharedPreferences with Gson.
More details
Remaining steps
- Replaced Room with SharedPreferences and Gson after build failures.
- Normalized insurance and benchmark datasets and joined them by date.
About this account
Claude.ai produced scoped implementation instructions; Claude Code implemented and built them while the author bridged the sessions.
Platform: dev.to. Attribution: https://dev.to/raio/i-built-an-android-app-in-4-days-with-zero-android-experience-using-claude-code-and-a-two-layer-2p44. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: raio on dev.to
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude.ai · Claude Code · Android Studio · Kotlin · Jetpack Compose · Gson
Android MVP · Two-layer workflow · Data normalization
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.