Releases sped up, QA did not
Features take days to build, and a manual checklist pass on Android and iOS takes the same days. Testing becomes the bottleneck.
claude code · appium · android + ios
Your own multi-agent QA conveyor for mobile apps
Claude Code agents write and run Appium tests on your real Android and iOS devices in parallel. Every change is checked before it is accepted, failures are analysed automatically, and found bugs go to your tracker.
- ✓Code and tests stay yours
- ✓Your phones, not a cloud farm
- ✓One-time purchase
// case
Five days of the conveyor
A mobile marketplace, Android + iOS. Anonymised at the client's request.
- autotests in 5 days (116 → 306)
- +190autotests in 5 days (116 → 306)
- devices working in parallel
- 4devices working in parallel
- bugs found and documented
- 49bugs found and documented
- of changes checked before acceptance
- 100%of changes checked before acceptance
Numbers come from the repository history and run reports; the app and client are not named.
One report for every device
// problem
Mobile regression is still done by hand
With coding agents, developers ship releases faster, yet the regression checklist is still walked by a person — on two platforms, on several phones, before every release.
Mobile automation is expensive to start
Appium on real devices means signing and environment setup, fragile locators and flaky runs. Months before the first payoff.
One agent for everything does not cut it
Ask an AI to "write tests" and you get false-green checks, a messed-up test account and tests that never ran on a phone.
// how it works
A team of agents, not one chat
Agents on your devices write and run tests in parallel. Every change is checked before it is accepted, failures are analysed automatically, and found bugs are documented and sent to your tracker. You get the result: new tests, a report and a bug list.
agents on devices
write and run tests
quality check
every change before acceptance
report
results and failure analysis
bugs → tracker
documented with evidence
Agents on devices
testsWrite tests from your regression checklist and run them on real Android and iOS devices — in parallel on several phones.
Quality check
controlNo change is accepted unchecked: false-green tests and side effects on the account do not reach your code.
Failure analysis
analysisRed results do not sit unattended: for every failure it is clear whether it is an app bug or a test problem.
Planning
prioritiesWork never stalls: checklist coverage grows, found bugs get rechecked, and you see the progress.
// what's inside
A ready repository you keep
The template is configured for your app and devices and then runs on your side.
A team of AI agents
write, run and check tests, analyse failures
Android and iOS test project
Kotlin + Appium, ready to grow with your app
Runs on your devices
in parallel on several real phones
One report
results from all devices in one place, with history
Bugs to your tracker
documented reports with evidence: Jira or manual
Safety rules
guard rails for the test account and irreversible actions
Your checklist as the base
tests are written from your regression checklist
Documentation
how to configure, run and grow the conveyor
// safety
Agents on a real account — only with guard rails
The conveyor runs on your devices and test accounts, so guard rails are built into the conveyor itself, not left to "we asked the agent to be careful".
01
Forbidden actions
You define what agents never do: delete the account, log out, make a real payment. Such checks stay manual.
02
Irreversible only in a sandbox
Real operations (an order, for example) are allowed only on a dedicated test object, and their cancellation is confirmed.
03
Devices don't get in each other's way
Parallel runs on different phones do not conflict or break each other's sessions.
04
Checked before acceptance
A change that could give a false-green result or mess up the account state does not reach your code.
05
The account stays clean
Tests clean up after themselves, and the account state is checked after the conveyor's work.
06
Everything stays with you
The conveyor runs on your machines and devices; code, tests and data do not go to a third-party service.
// comparison
How it differs from the alternatives
| SaaS platforms (QA Wolf, Maestro Cloud, etc.) | AI testing courses | This kit | |
|---|---|---|---|
| Where tests live | on the vendor's side | course examples | in your repository |
| Devices | the vendor's cloud devices and emulators | usually a browser | your real Android phones and iPhones |
| Platform | depends on the vendor | mostly web and Playwright | mobile: Android + iOS |
| Pricing model | subscription, often per test or device | one-time per course | one-time purchase + a year of updates |
| If you stop paying | access to runs ends | the knowledge stays | the code and the conveyor stay with you |
SaaS services are great when you want a ready device farm and a team on the vendor's side. The kit is for teams that want to keep tests and infrastructure in-house.
// pricing
Template, setup or training
The template is for teams ready to configure it themselves. Setup is for when you want a working conveyor on your devices.
Template
$1991 developer
$599team of up to 5
A ready multi-agent conveyor repository with documentation.
- +A team of agents for Android and iOS
- +Kotlin + Appium test project
- +One report for all devices
- +1 year of updates
Template + setup
from $2,500for your app
I configure the conveyor for your app and devices and run the first tests together with you.
- +Team template
- +Setup for your app, accounts and devices
- +Safety rules and sandbox
- +First tests on your phones
- +3 calls: kickoff, review, handover
Training
$3492-day workshop, per person
$150mentoring, per session
How to work with a multi-agent conveyor and run it yourselves.
- +Curriculum based on a real conveyor
- +Practice on your devices
- +Review of your tests and bugs
The Claude subscription and devices are yours and not included.
// training
Workshop curriculum
2 days online, small group, hands-on with your own devices. For QA engineers and mobile developers.
- 01
Claude Code for QA
where AI agents help in mobile testing and where their limits are
- 02
Starting on your app
what to prepare: devices, test accounts, a regression checklist
- 03
Appium on real devices
how to get stable tests on Android and iOS
- 04
Test quality
how to tell a real check from a false-green one
- 05
Working with bugs
from a failing test to a ready report in the tracker
- 06
Safety
how to give agents a test account without losing control
- 07
Reports
how to read results and show progress to the team
- 08
Growing it
how to extend coverage and maintain the tests
// faq
FAQ
Do I need a Mac for iOS?
Yes. Testing on an iPhone needs a Mac with Xcode. The kit targets macOS, so a Mac is needed for the conveyor as a whole.
How much does Claude cost?
You need a Claude plan with Claude Code access or an API key — paid by you directly to Anthropic. Parallel work on several devices uses a lot, so for 3–4 devices the higher tiers are more practical. Current prices are at claude.com/pricing.
Which phones work?
Android phones and emulators with USB or Wi-Fi debugging, and iPhones with Developer Mode. You can start with a single phone and add devices later.
What stack are the tests on?
Kotlin + Appium (Android and iOS) + JUnit 5 + Gradle. This is the stack supported out of the box; another one can be discussed as part of a setup.
Can I use my own tracker?
Out of the box: Jira or manual mode (bugs are prepared for filing and you file them). I can connect another tracker as part of the setup package.
What about account safety?
You define the forbidden actions, irreversible operations are allowed only in a sandbox, and every change is checked before it is accepted. Use test accounts.
// request
Let's talk about your conveyor
Tell me about your app and devices — I reply on Telegram or by email within a day.
Telegram @kolyall ↗Message me on Telegram
Tell me about your app and devices and what you need — template, setup or training.
Message on Telegram