Mobile GUI Use
Optimized for everyday tasks on a real phone, CoreAutomata’s hosted agent navigates changing Android and iOS apps — search, compare, schedule, check out — and asks before anything irreversible.
01 / 06
AI MOBILE QA + AUTOMATION · REAL DEVICES · ANDROID + iOS
TestersAutomation
Ask in natural language. We drive the phone.
Tell it to categorize the video content using AI
Write a test or describe a task. CoreAutomata’s hosted AI runs it on a real Android or iOS device you pair — autoverifies what the screen shows, self-heals when the UI moves, and adapts when a popup gets in the way.
200 free actions · hosted AI · real devices · no credit card
What it can do
Built to finish real work on a live phone.
One paired device and hosted AI cover natural-language tests, goal-driven tasks, and the repairs that keep both running.
What CoreAutomata does not require
Real devices
your phone or simulator, QR-paired
Hosted AI
our engineered vision models
No selectors
nothing to fix when the UI moves
Multiple apps
automate multi app flows
CoreAutomata at a glance
6500
agent actions run
28
real devices connected
600
test cases and agent tasks
90
users
Who it's for
One platform. Tests and tasks.
The same real-device connection and hosted AI power both — author natural-language cases or describe a goal.
Testers & QA
Test engineers · SDET · QA leads
Stop maintaining selector maps. Author cases in natural language against a build you captured once, keep them in a shared library, and let our AI re-find elements when the UI shifts — on real devices you already own.
- Auto discovery drafts your first cases from the build
- Three-tier self-healing replay on every rerun
- Intent checks on what the screen actually shows
- One shared case library across projects and folders
Automation teams
Developers · RPA · Data & ops
Automate workflows that only exist inside a mobile app — internal tools, vendor portals, apps with no public API — by describing the outcome. Our hosted agent drives a real phone; you never plug in or host a model.
- Works on any app already installed on the device
- Reasons through screens it has not seen before
- Recovers from popups, upsells, and consent banners
- AI included
How it works
Pair. Ask. Run.
One connection powers natural-language tests and goal-driven tasks — three steps, no scripting, AI included.
Pair a real device with a QR code
Install the CoreAutomata Portal on an Android phone, or run the iOS Bridge on a Mac, and scan the code in your browser. The device dials out — no cable, no adb, no Xcode on your machine.
Capture the build — or open any app
Upload an APK, XAPK, or IPA and CoreAutomata maps screens and drafts cases. Or skip the build and drive any app already on the phone by describing the outcome.
Ask, run, adapt
Accept discovery’s drafts, write your own checks, or name a goal. Every run executes on a live device; our AI repairs moved targets, handles unexpected screens, and keeps screenshots plus per-step outcomes.
Apps
Automate the apps your work lives in.
If it’s installed on the paired phone, CoreAutomata can drive it — messaging, social, commerce, work. No SDK in the app, no private API required.
Messaging
Chat and inbox flows on a real device.
WhatsApp
Confirmations, codes, and authorized business threads
Telegram
Channel checks and scheduled reads
Gmail
Triage, send, and follow-up in the mobile app
Discord
Post updates and monitor channels you belong to
Social
Publish, manage, and QA authorized accounts.
Instagram
Publish paths and DM triage on a real phone
TikTok
Schedule checks and pull what the app shows
X
Post, monitor mentions, and walk the feed
LinkedIn
Authorized outreach and profile flows
Facebook
Pages, Marketplace, and comment replies
YouTube
Upload paths and channel analytics screens
Commerce
Source, monitor, and QA on real shopper devices.
Amazon
Listing checks and checkout QA
Shopify / store apps
Promo paths and cart totals
Alibaba
Supplier inbox, quotes, and MOQ screens
Walmart
Catalog, pricing, and Marketplace flows
eBay
Price and stock watches on your listings
Uber
Book and confirm flows on your own account
Work
Maps, browsers, and field apps your agent can drive.
Chrome
Open URLs and sign into web apps on mobile
Google Maps
Local visibility checks from a real location
Expense & HR apps
Approvals and exports with no public API
Your internal app
Whatever your team already has installed
Why CoreAutomata
Stop maintaining selectors
Selector frameworks ask you to describe your app twice — once in code, once in the test. CoreAutomata works from the real-device screen, the same thing your users see, with AI that self-heals the suite and drives the next tap.
| CAPABILITY | CoreAutomata | Appium / Detox |
|---|---|---|
| Writing a step | Natural language | XPath / element IDs |
| Your first suite | Discovery drafts it | Hand-written, screen by screen |
| When the UI moves | AI re-finds the element | Locator breaks |
| Where it runs | Real device, QR-paired | adb / Xcode / driver farm |
| AI / model setup | Hosted — included | None (scripted) or DIY |
| Who can author | Anyone on the team | Whoever knows the framework |
Capabilities
Built for testers and operators
One capture or one paired phone becomes the shared context for natural-language cases and goal-driven tasks — on real devices, with our AI.
QA
Auto discovery from the build
Point CoreAutomata at a build and it explores on a real device — opening screens, filling forms, mapping flows — then drafts candidate test cases from paths it actually reached.
QA
Natural-language test cases
A step is a sentence: "tap Checkout", "confirm the shipping form is empty". No XPath, no accessibility IDs, no Appium driver. The step text is the test — anyone on the team can author.
QA
Three-tier self-healing replay
A rerun first replays the recorded step. If the target moved, our AI re-finds it. If the screen changed more than that, it works from the step’s intent. Only then does it fail — and the report says which tier saved it.
BOTH
Our AI reads the screen
CoreAutomata’s hosted vision stack reads each frame — controls, labels, state, what is tappable. No DOM access, no accessibility-tree assumptions, no changes to the app, and no model API for you to wire up.
AUTOMATION
Plans across many steps
For a goal, the agent breaks it into sub-goals, tracks which are done, and re-plans as the screen changes. Long flows stay on target instead of drifting after the third tap.
BOTH
Grounded gestures on real devices
Taps, swipes, scrolls, long-presses, and text entry resolve to real coordinates on the live screen and dispatch over the device’s own connection — nothing to install on a host machine.
AUTOMATION
Handles the screens nobody scripted
Consent dialogs, permission prompts, rating nags, upsell interstitials, an unexpected logout. The agent deals with whatever appears rather than needing a branch written for each one.
AUTOMATION
Any app on the device
If it renders, it can be driven. That includes apps you did not build — vendor portals, internal tools, third-party apps with no API and no chance of getting one.
BOTH
Evidence on every run
Each run keeps the screenshot, the action taken, and the outcome for every step. When something fails you get the frame it failed on — not a stack trace pointing at a locator.
QA
Vault for credentials
Account logins go in the project Vault, encrypted at rest and injected at run time. They never appear in step text, in run logs, or in the screenshots you share.
AUTOMATION
Guardrails and a replayable trail
Runs are scoped to the app you name, stopped before irreversible actions, and recorded step by step so you can see exactly what the agent did and replay it.
QA
Checks what the screen shows
Assertions read the live UI the way a tester does — labels, totals, empty states, clipped controls — not brittle pixel baselines. Catch visual and copy issues that selector checks miss.
What you actually write
Tests and tasks, same language
Natural-language cases for QA. Goal descriptions for automation. Both run on the same paired device with hosted AI.
A test case is a list of sentences
This is the whole artifact — no page objects, no locator file, no driver setup. Steps that act and steps that check sit side by side. The same case reruns on a real device against every later build, and our AI repairs what moved.
- Written in the Board, or drafted by auto discovery
- Credentials come from the Vault, never from step text
- Reruns self-heal steps whose targets moved
- Every step keeps its screenshot and outcome
checkout · guest purchaseQA
- →Open the app and dismiss the welcome sheet
- →Search for "running shoes" and open the first result
- ✓The product page shows a price and an Add to cart button
- →Add to cart, then open the cart
- ✓The cart total matches the price on the product page
- →Check out as a guest with the saved test address
- ✓The confirmation screen shows an order number
- rerun · 7 steps · step 4 re-grounded (tier 2)
A task is a description, not a script
You state the outcome. Our hosted agent decides what to tap on the real phone, deals with whatever the app puts in the way, and shows you what it did. Nothing here names a coordinate, an element ID, or a model you own.
- Runs against any app already on the paired device
- Re-plans when a screen is not what it expected
- Stops and asks before anything irreversible
- Every action is recorded and replayable
task · export last month's paymentsAutomation
- goal
- →In the payments app, open the transaction list,
- →filter to last month, and export it as CSV.
- run
- ✓Dismissed "Rate this app" dialog
- ✓Opened Transactions from the bottom tab bar
- ✓Set the range filter to Last month
- ✓Tapped Export, chose CSV, confirmed the share sheet
- done · 11 actions · 1 unexpected screen handled
Solutions
Find your workflow.
Teams that test and automate on real phones — and industries where the work lives in mobile apps.
Built for
Teams that run work on real phones
QA, ops, research, and product teams using the same Board.
QA Teams
Natural-language cases on real devices — self-healing when the UI moves.
Automation & RPA
Mobile-first process automation for apps with no API.
AI Agent Builders
A real phone for your agent — hosted vision, no model key to bring.
Ops & Support
Routine in-app tasks on the phone your team already uses.
Market Intelligence
Walk competitor apps and map screens, paywalls, and flows.
Product & UX
Replay onboarding and catch where the funnel stalls.
Industry
Verticals where the workflow lives in mobile
If the work only exists inside an app, CoreAutomata can drive it.
Supply Chain
Vendor portals and logistics tools that never ship an API.
Healthcare
Field notes and mobile EHR flows on a paired device.
Commerce
Checkout QA, promo paths, and catalog checks at release pace.
Finance & Finops
Expense approvals, payments exports, and reconciliation in-app.
Field Service
Dispatch and work-order apps your crews live in every day.
Internal Tools
Legacy vendor apps your ops team will never get an integration for.
Use cases
What teams run on CoreAutomata
REGRESSION
Regression pass on every build
Rerun the library against each new build. Steps whose targets shifted get repaired in place.
MOBILE RPA
Workflows that only exist in an app
Expense approvals, HR portals, dispatch tools — drive the app when there is no API.
EXTRACTION
Pull data off the screen
Prices, listings, statuses, inventory — structured from what the app displays.
SMOKE
Pre-release smoke checks
Launch, sign in, primary purchase path — run against a release candidate before it ships.
RESEARCH
Walk a competitor's app
Map onboarding, paywalls, and pricing into a graph without a person tapping through.
VISUAL INTENT
Catch what selectors never see
Empty states, error toasts, totals that match line items — grounded in the screen.
Simple pricing
CoreAutomata is in beta. Start free on a device you already own — early adopters get free upgrades when paid plans launch.
FREE
$0
200 agent actions a month and 2 projects on a real device you pair yourself. Hosted AI included. No card required. Early adopters get free upgrades as paid plans ship.
Start freeENTERPRISE
Talk to us
A dedicated device pool, SSO, custom limits, and support terms that fit your release cycle — available now while we’re in beta.
Contact usQuestions
The things people ask first
Does CoreAutomata run on real devices?
Yes. Pair a physical Android phone or iPhone (via the iOS Bridge on a Mac), or use a simulator. Tests and agent tasks execute against the live screen — not against a mocked UI tree alone.
Is this an alternative to Appium?
It covers the same job without the selector layer. Appium tests address elements by ID or XPath, so they break whenever the UI shifts. CoreAutomata works from what is on screen, so a moved or renamed element is something our AI can re-find rather than a broken locator.
What happens when the UI changes and a step no longer matches?
Reruns go through a three-tier repair ladder. First the exact recorded step is replayed. If the target moved, the step is re-grounded against the current screen. If the screen changed more substantially, the agent works from the step's intent instead. The rerun only fails after all three tiers fail, and the report shows which tier recovered it.
Where do my app builds and test credentials live?
Builds and run artifacts live in your project's storage. Test-account credentials go in the project Vault, encrypted at rest, and are injected into a run without appearing in step text, logs, or screenshots.
Is CoreAutomata open source?
No — it is a commercial product. The Portal app and the iOS Bridge are free companion apps you use to connect your own devices.
Do I bring my own LLM or API key?
No. Vision, planning, and grounding run on CoreAutomata’s own hosted AI backend. You describe test cases or tasks in natural language and pair a device — you do not plug in a model provider, host inference, or expose an LLM endpoint.
For app testing - or flow automation
200 free agent actions a month. Real devices. No card. Pair a phone or create a device farm and run your first case or goal today.