The AI QA agent for mobile and web apps

Ship FAST. Break nothing.

mini writes, runs and maintains your regression tests.
Even the tricky ones.

Book a demo Watch mini work

Train with every class

200+ classes and routines
Progress log and streaks
Offline downloads

Yearly €59.99Monthly €9.99

Start free trial

Renews automatically. Cancel anytime in Settings.

3/10

Type it in English¿Dónde está la estación?Where is the sta

Check

CancelNew noteDone

Filed under Offsite

Ideas for the offsite: book the venue by Friday, ask Sam about the budget, and a walk on day two

0:07

Mia’s phone

MMiaAt home · Just now

Home
School

For youFollowing

NBlock Drop@noa · 2.4k plays1.2k86

Lisbon12–16 May · 4 travelersAJRS

Flights€612
Stay€548
Tours€88

Total€1,248Each€352Settle up

PayoutsSeptember · 4 partnersExport CSV

Partner	Clicks	Sales	Payout
Halden Studio	18,204	412	€6,180
Fernway	9,876	233	€3,495
Kestrel Media	7,431	161	€2,415
Marlow & Pike	4,012	88	€1,320

Train with every class

200+ classes and routines
Progress log and streaks
Offline downloads

Yearly €59.99Monthly €9.99

Start free trial

Renews automatically. Cancel anytime in Settings.

Lisbon12–16 May · 4 travelersAJRS

Flights€612
Stay€548
Tours€88

Total€1,248Each€352Settle up

3/10

Type it in English¿Dónde está la estación?Where is the sta

Check

PayoutsSeptember · 4 partnersExport CSV

Partner	Clicks	Sales	Payout
Halden Studio	18,204	412	€6,180
Fernway	9,876	233	€3,495
Kestrel Media	7,431	161	€2,415
Marlow & Pike	4,012	88	€1,320

CancelNew noteDone

Filed under Offsite

Ideas for the offsite: book the venue by Friday, ask Sam about the budget, and a walk on day two

0:07

Mia’s phone

MMiaAt home · Just now

Home
School

For youFollowing

NBlock Drop@noa · 2.4k plays1.2k86

iOS
Android
Web
Real devices and simulators

Full regression

From 2 days to 1 hour

Time to full coverage

48 hours

Release cadence

From Fortnightly to Weekly

Link a Child Device to the Parent Account

Acceptance criteria

ParentOn device one, the child card shows a Phone row with a ‘+ Connect’ button; tapping it opens the pairing screen showing a QR code and a 6 digit code displayed as plain text, with the three setup steps described below it.
ChildTyping the 6 digit code and tapping the ‘Link Device’ button accepts the code and opens ‘Let’s set up protection’ (an ‘Invalid or expired pairing code’ alert for a freshly displayed code is a failure).
ParentOn device one, returning to the Family tab shows the phone row of the child card as Connected with the device name, without a manual refresh.
+6 more criteria

Running9 of 9 criteria passed

Enforce the Daily Limit on the Child Device

Acceptance criteria

ChildAfter the parent sets today’s limit, the child Home ring on device two shows the new budget within ten minutes (reopen the app if needed).
ChildOnce the budget is exhausted (zero budget, or after using Chrome until the ring reaches zero), opening Chrome on device two shows a Helmit time-is-up block screen or Chrome is closed immediately, and the child Home banner reflects the limit reached.
BothGranting bonus time on device one (from the request or the Rules tab) updates the child ring with a bonus badge and Chrome opens again on device two within ten minutes.
+2 more criteria

Running5 of 5 criteria passed

Create a Safe Zone and Verify Enter and Exit Alerts

Acceptance criteria

ParentSafe zone creation lets the parent place the zone with the current mocked parent position or by searching a place name and tapping a suggestion (raw coordinates are not accepted in the search), give it a name and an icon, and save it.
ParentThe saved zone appears on the map as a marker or circular overlay with its name and icon, centred near 48.1374, 11.5755 (Munich), not at a default location.
BothMoving device two to 48.1374, 11.5755 and holding it there for at least four minutes produces a zone ARRIVAL notification on device one naming the child and the zone: check the Android notification shade and the in-app notifications screen, never the Alerts tab.
+3 more criteria

Running6 of 6 criteria passed

Offline Notice and Back Online Toast

Acceptance criteria

Offline (wifi off): within about 20 seconds of turning wifi off, a white card appears at the top of the app with the title ‘You are offline.’ and the line ‘You can still play anything you’ve saved, or reconnect to keep scrolling.’
After turning wifi back on and returning to the app, a card titled ‘Back online’ with the line ‘Fresh games are loaded - keep scrolling.’ appears at the top of the app within about 15 seconds.
The ‘Back online’ card disappears on its own within about 15 seconds without being tapped.
+7 more criteria

Running10 of 10 criteria passed

Email OTP Sign Up and Onboarding

Acceptance criteria

The welcome screen shows a single ‘Continue with Email’ button (there is no separate sign-up link or password field)
Entering a new email and tapping ‘Send code’ shows the 6-digit code entry screen
Tapping ‘Continue’ on the genre screen leaves onboarding and opens the app’s main screen (the game feed, which may show a ‘make a game’ prompt over it)
+5 more criteria

Running8 of 8 criteria passed

Track Bet Lifecycle Through Win, Loss, Awaiting and Refund

Acceptance criteria

A market the persona bet on and that resolved in their favor shows as Won in Bets, with the payout in Points displayed
A market the persona bet on and that resolved against them shows as Lost in Bets, with the lost stake shown
A market the persona bet on that was voided shows as Refunded in Bets, with the refunded credits reflected in the balance
+12 more criteria

Running15 of 15 criteria passed

Parent device

Child device

Customer runs, shared with permission. Waits cut or sped up.

mini does the homework first.

It explores your app and reads your specs, tickets, code, bug reports and existing tests. Then it writes your suite.

How mini builds a suite

Product specs, tickets, code, bug reports and existing tests flow into mini. mini then tests what changed on each pull request and runs the full regression before a release.

Product specs
Tickets
Codebase
Bug reports
Existing tests

Tests what each PR changes.
Only the journeys it puts at risk.
Runs the full regression.
Before each release, on iOS, Android and web.

Tests that keep up with your code.

mini tracks your code, specs and tickets. It adds, edits and deletes tests to match, and your team reviews every change.

Move a button. The test still passes.

mini’s tests have no selectors to break. It reads the screen on every run, finds the new path and checks the same outcome.

How mini handles UI changes

mini speaks up before you ask.

It flags bugs no test asked about, suggests UI/UX improvements and asks your team in Slack when context is missing. Accessibility audits come with every plan. On request, mini checks your iOS and Android apps against WCAG 2.2 too.

mini in Slack

Sign in1Catalog23Checkout9Settings8Profile

Conformance

52 criteria

Non conform34
To review1817
Conform77
Not checked2222
Not applicable22

CheckoutAudited

Non conform

Contrast (minimum)1.4.3 · AA3.2:1 measured, 4.5:1 required

Target size (minimum)2.5.8 · AA20×20 dp measured, 24×24 dp required

To review

Use of color1.4.1 · ADelivery status shown by color aloneIssueDismissConfirmed

Copy for agent

Fits into the way you ship.

All integrations

Reads your code and specs in GitHub and Notion; Jira and Linear on request
Runs from your CI, CLI or coding agent
Sends findings to Linear, Slack and GitHub

miniCritical

Cart subtotal does not match the sum of line prices

Android · Checkout

Expected: $82.48
Actual: $85.48

Recording
Steps to reproduce
Prompt to fix

Slack#qa-mobilenow
miniAPPCritical on Android: subtotal $85.48, items add up to $82.48
GitHubRelease PR #1290now
Minitest (Shop) · 411 passed, 1 failed
Some checks were not successful
LinearMOB-418now
Cart subtotal does not match the sum of line prices
UrgentBug

Any other system with an API

mini writes the tests. Then runs them.

mini learns your app, writes each test as a scenario with acceptance criteria, and keeps them current. No steps, no scripts.

12 scenarios ready to test

RegistrationRegister a New Account
SearchSearch for Restaurants and Dishes
CheckoutPlace an Order and Complete Payment
LoginSign In to Existing Account
Order TrackingTrack an Active Order in Real Time
Cart ManagementManage Cart Contents
Cart ManagementView Item Details and Add to Cart with Customization
NavigationBrowse Home Screen and Discover Restaurants
ProfileView Profile and Order History
Ratings & ReviewsRate and Review a Restaurant
Support ChatContact Support via AI Chat
FavouritesFavourite and Unfavourite Restaurants

Manage Cart Contents

A customer fills a cart and checks the total.

Acceptance criteria

Passed: Two different items are in the cart
Passed: Each line shows its extras, quantity and price
Passed: Changing a quantity updates the totals
Failed: The price summary reflects every item

Android failed
iOS passed

The subtotal is $3.00 too high Critical

Expected Items add up to

$82.48

Actual Subtotal shown

$85.48

The subtotal adds $1.50 per cart line, so the grand total is overstated.

Recording0:18
Steps to reproduce6 steps
Prompt to fixCopy for Cursor / Claude

Create ticketNot a bug · refine mini

Open the full report

mini writes the tests
mini runs them
mini calls it out
Your team decides

One setup. Every release.

Your team: Open a PR
mini tests the change
Only the journeys it puts at risk.
Your team: Merge to staging
mini runs the full regression
Your agreed suite, end to end.
Example: build 2.14 failed Checkout on Android. After your fix, build 2.14.1 passed all 412 tests.
If mini finds a bug, your team: Apply the fix
mini sends a recording, the steps to reproduce and a prompt to fix. Your team copies the prompt into its coding agent, such as Cursor or Claude Code.
Your team: Release to production
You decide when to ship.
Example: build 2.14.1 passed all 412 tests. Someone on your team approves the release in your own release process.

Full regression22 testsQueuedRunning21 passed1 finding

Hosted in the EU or the US, your choice.
On Enterprise, in your cloud or on-premise.
Parallel runs, sized to your plan.

Teams that ship with mini.

**Mikalai Dzmitryieu**Senior Software Engineer

Your team decides.
mini runs.

mini writes your first suite, and we review it with your team. We connect it to your release process and work through your security review.

You choose the deployment, who has access and the support you need.

Enterprise controls

Your enterprise setup

Trust Center Status page

Access: SAML SSO, SCIM provisioning and workspace roles.
Auditability: Exportable audit logs and traceable test changes.
Data: Zero data retention on model calls, on request.
Deployment: On-premise, your cloud, a dedicated instance or our managed cloud in the EU or the US.
Support: Hands-on onboarding, a named owner in your Slack, a 99.5% uptime SLA and a first response to a P1 in 2 business hours.
Compliance: SOC 2 Type II audit in progress. GDPR in progress.

Less repetition.
More exploration.

QA teams

More releases.
Same team.

Engineering leaders

The work behind mini.

All articles

Startup Minitap tops DeepMind's mobile AI benchmark, raises $4.1 million seed round (opens in a new tab)

Announcing minitest: the agent owns the suite now

An earlier Artemis package file lists the three Minitap authors.

I expected better from Google

Good questions.

Security and enterprise Compare plans

We already have automated tests. Now what?

Keep them. Hand mini the gaps and the manual checks. It writes, runs and maintains those tests.

Can mini meet our security requirements?

Enterprise plans offer on-premise or private deployment, SSO, SCIM and audit logs, with zero data retention on model calls on request. Our SOC 2 Type II audit is in progress, and we work through your security review with you.

How does mini decide whether a test passed?

Against the test’s acceptance criteria. Each verdict links to the exact moment in the recording, so your team can check the call.

What about false positives and flaky results?

Before mini fails a criterion, it rules out its own input, the network and the starting state. If our infrastructure hiccups, the test re-runs on a fresh worker. Every failure comes with the recording and a screenshot of the moment it broke, next to what mini expected. Mark it “Not a bug” once and mini won’t raise it again.

What happens when my UI changes?

A moved button doesn’t break a test, because mini’s tests have no selectors. When behavior changes, mini picks it up from your code, tickets or specs and updates the test for your team to review.

Does mini run in our CI?

Yes. Trigger runs from CI, a pull request or Slack. Results come back to your release workflow with recordings and steps to reproduce.

What if the requirements are unclear?

mini proactively asks your team in Slack, then reuses the answer.

Can we chat with mini?

Yes. In the app, ask mini about any report or finding, or ask it to cover a new flow. It answers with your app’s context.

Does mini test accessibility?

Yes, on every plan, for iOS and Android apps. Ask and we switch it on. mini audits every screen a run reaches against WCAG 2.2 A and AA, with the evidence behind each verdict.

Which devices and browsers does mini use?

iOS simulators and Android emulators in the cloud, plus real iPhones and Android phones on request. Web runs in Chrome and Firefox at desktop, tablet and mobile sizes, and in Safari on iPhone and Chrome on Android. Your plan sets how many run in parallel.

What does getting started involve?

Bring your app, test accounts and any existing tests or specs. mini writes your first suite, your team reviews it, and we connect it to your release process.

How is mini priced?

Plans depend on your coverage, platforms and run volume. We’ll scope the right plan around the journeys you want tested.

Ready to
meet mini?

Bring your hardest flow. Watch mini run it.

Book a demo See a real report

Ship FAST. Break nothing.

mini does the homework first.

Fits into the way you ship.

Reads your code and specs in GitHub and Notion; Jira and Linear on request

Runs from your CI, CLI or coding agent

Sends findings to Linear, Slack and GitHub

Your team: Open a PR

mini tests the change

Your team: Merge to staging

mini runs the full regression

If mini finds a bug, your team: Apply the fix

Your team: Release to production

Teams that ship with mini.

Memrise

Your team decides. mini runs.

Less repetition. More exploration.

More releases. Same team.

The work behind mini.

Startup Minitap tops DeepMind's mobile AI benchmark, raises $4.1 million seed round (opens in a new tab)

Announcing minitest: the agent owns the suite now

I expected better from Google

Good questions.

Ready to meet mini?

mini does the homework first.

Fits into the way you ship.

Reads your code and specs in GitHub and Notion; Jira and Linear on request

Runs from your CI, CLI or coding agent

Sends findings to Linear, Slack and GitHub

Your team: Open a PR

mini tests the change

Your team: Merge to staging

mini runs the full regression

If mini finds a bug, your team: Apply the fix

Your team: Release to production

Teams that ship with mini.

Memrise

Your team decides. mini runs.

Ready to meet mini?

Your team decides.
mini runs.

Less repetition.
More exploration.

More releases.
Same team.

Ready to
meet mini?

Your team decides.
mini runs.

Ready to
meet mini?