QA is the bottleneck. Meet your QA workers.

Your team ships more than anyone can check. QA workers are AI testers in your Slack, each with its own computer. Ask the lead to check a flow. It shares the work with its specialists and tells you what broke.

Ask for a first look
  1. Monday
  2. Dana Reyes4:18 PM

    @Milo please check sign-up after the new email step.

  3. MiloAPP4:51 PM

    Done. Sign-up works. 4 tests, all passed.

    check mark reaction:2

  4. Yesterday
  5. Owen Pratt5:40 PM

    The new checkout is on staging.

  6. Dana Reyes5:52 PM

    Nice. I’ll have it checked in the morning.

    thumbs up reaction:1

  7. Today
  8. Owen Pratt9:02 AM

    Promo codes are back on checkout for Thursday’s release.

    party reaction:2

  9. Dana Reyes9:12 AM

    @Milo can you check guest checkout on staging before Thursday’s release?

    eyes reaction:1

  10. MiloAPP9:13 AM

    Sure. My plan, riskiest first:

    1. Pay as a guest by card
    2. Use a promo code
    3. @Ivy takes the declined card

    I’ll leave out gift cards. They’re lower risk.

    OK to start?

  11. Dana Reyes9:14 AM

    Approved.

    Plan approved · 1 of 2

  12. MiloAPP10:02 AM

    The 3 tests are ready. OK to run them?

  13. Dana Reyes10:05 AM

    Approved, run them.

    Tests approved · 2 of 2

  14. MiloAPP10:14 AM

    Done. I tested guest checkout: 3 tests, 2 passed.

    Found 1 problem (Ivy): after a declined card, the page still says “Order placed”. Steps and a recording are in the thread.

    Not tested: saved addresses. They need a sign-in.

    thank you reaction:1

It looks closely.

It uses your app like a careful tester, then checks what the screen does not show.

  1. Walks the checkoutTypes a promo code and picks shipping.
  2. Tries a phone widthEverything fits and is easy to tap.
  3. Tests the API behind the pageEach call answers, and answers fast.
  4. Checks accessibilityNames, contrast and the keyboard.
  5. Reads your last CI runAll green. No test checks the total.
  6. Finds the wrong totalIt says $74.61. The lines add up to $68.21.

When it finds something, you get proof.

The lines add up to $68.21.

$64.00 − $6.40 + $6.00 + $4.61 = $68.21

Every finding comes with a screenshot and the test that shows it.

staging.pellwick.example/checkout

Review your order

Pour-Over Kettle, 1 LQty 1$64.00
Promo code WELCOME10Applied
Subtotal
$64.00
Discount
−$6.40
Shipping
$6.00
Tax
$4.61
Total$74.61
A screenshot from the run.

It tells you what it didn’t test.

Every summary ends with what it left out, and why. You always know where the checking stopped.

MiloAPP

Guest checkout: 3 tests, 2 passed, 1 failed.

Found: the review step total ignores the promo code. It shows $74.61. The lines add up to $68.21.

Left out: gift cards, lower risk.

Saved addresses need a sign-in. I can’t open those on my computer yet.

The tests are ready for review: open them.

It waits for your yes.

It asks twice: before it writes the tests, and before it runs them. Everything else, it writes down.

  1. Monday, October 12th
  2. Owen Pratt5:40 PM

    The new checkout is on staging.

  3. Dana Reyes5:52 PM

    Nice. I’ll have it checked in the morning.

    thumbs up reaction:1

  4. Tuesday, October 13th
  5. Owen Pratt9:02 AM

    Promo codes are back on checkout. They ship on Thursday.

    party popper reaction:2

    2 repliesLast reply today at 9:05 AM

  6. Dana Reyes9:12 AM

    @Milo can you plan tests for guest checkout? It ships Thursday.

    eyes reaction:1

  7. MiloAPP9:13 AM

    Plan ready. 3 tests, riskiest first:

    1. A guest pays by card and sees the order confirmation
    2. A promo code lowers the total by its amount
    3. A declined card shows an error and places no order

    Left out: gift cards, because they are lower risk.

    Do you approve the plan?

  8. Dana Reyes9:14 AM

    Approved.

  9. MiloAPP9:14 AM

    Thanks. I’m writing the tests now.

    Plan approved by Dana Reyes · approval 1 of 2

  10. MiloAPP10:02 AM

    The 3 tests are ready, one for each step of the plan. I walked each one in my browser first.

    Do you approve them? Then I run them.

  11. Dana Reyes10:05 AM

    Approved, run them.

  12. MiloAPP10:05 AM

    Running them now.

    Tests approved by Dana Reyes · approval 2 of 2

    raised hands reaction:1

Ship with confidence. Your QA team is always on.

Your workers keep an eye on your app all day. They remember how it works, know the paths your customers take, and check what changed before anyone asks. While everyone sleeps, they keep going.

Overnight, Milo tests the store checkout on desktop, Kai the same store at phone width, Nora the orders API and Tomo the admin app. In the morning, Milo posts a summary in Slack.

  1. Today
  2. MiloAPP1:58 AM

    Starting tonight’s run on staging. Kai, Nora and Tomo are on it with me.

  3. MiloAPP7:30 AM

    Morning. Last night’s run on staging is done. Here’s what we checked:

    • Kai, the store at phone width: 1 new problem
    • Milo, checkout on desktop, with a keyboard too: passed
    • Nora, the orders API, 5 checks: passed
    • Tomo, the admin app: 1 flaky test, cause found

    New: the cookie banner covers Place order

    On a phone, a customer has to close the banner before they can pay.

    For this morning

    • The banner fix needs an owner.
    • Tomo has a fix for the flaky test. It waits for your yes.

    Full report saved to your repo

  1. Tuesday
  2. Owen Pratt11:20 AM

    @Milo the promo code fix is on staging.

  3. MiloAPP11:42 AM

    Checked. The total is right now. 3 tests passed.

  4. Dana Reyes11:48 AM

    Great, thanks.

  5. Yesterday
  6. Dana Reyes4:05 PM

    Release is Thursday. Let’s keep checkout green until then.

  7. Owen Pratt6:12 PM

    New product page is on staging. Small copy changes too.

    thumbs up reaction:2

  8. Owen Pratt6:13 PM

    The bag button reads better now.

  9. Today
  10. Dana Reyes9:31 AM

    @Milo last night’s report has two checkout tests in red. Can you look?

    eyes reaction:1

  11. MiloAPP9:38 AM

    Both stop at the same step. The Add to bag button now says Add to cart.

    I updated the two tests, and they pass. Was the rename intended?

  12. Dana Reyes9:40 AM

    Yes, intended. Owen renamed it.

  13. MiloAPP9:40 AM

    Thanks. I saved it as a decision, so I won’t ask again.

    Remembered: the button is now Add to cart

    raised hands reaction:1

When your app changes, it asks.

A test breaks because a button got a new name. The worker fixes the test, then asks if the change was intended. It remembers your answer.

Saved to its memory“Add to cart” is the new name. Intended, says Dana.

Each one works on its own computer.

Its own browser, its own test runner and a copy of your code, set up for that worker alone. It plugs into the tools your team already uses.

How access works

Plugs into

GitHub and Slack work today. The others are coming.

One worker, or a QA team.

Start with one worker. When you ship more, move to a team. A lead takes each request, shares it with up to five specialists and reports back in one thread. Each one learns its part of your product and remembers it.

You give each specialist a domain, a name and a shape.

  1. Dana Reyes9:12 AM

    @Milo can you check the release before Thursday?

  2. MiloAPP9:12 AM

    Yes. I split it in four:

    • Security: sign-in and sessions
    • Performance: checkout on a slow phone
    • API and backend: orders API and webhooks
    • Functional flows: guest checkout, start to end

    4 repliesLast reply today at 9:40 AM

  3. MiloAPP9:41 AM

    Done. Three parts passed. One issue: the orders API accepts an empty cart. Steps and a recording are in the thread.

  • Solo

    One QA worker for your app.

    $300USD a month

    • One QA worker with its own computer
    • In your Slack, answers at any hour
    • Always on, nights included
    • Learns your product and remembers it
    • You pick its name and shape
    Ask for a first look about the Solo plan
  • Team

    A lead and up to five specialists.

    $1,200USD a month

    • One lead plus up to 5 specialists
    • Each with its own computer and memory
    • You give each one a domain
    • They work in parallel
    • One summary back in your thread
    Ask for a first look about the Team plan

Prices are planned. More than one team? Talk to us.

Start with a first look.

Type your web address. A QA worker opens your public pages, read-only, and sends you what it saw and the three flows it would test first. It’s early access, so requests run in the order they arrive.

Questions

Why more than one worker?

One worker is enough to start. On the Team plan, a lead takes each request and shares the work with up to five specialists. You give each specialist a domain, like security, accessibility or the API, and you can change it at any time. Each one has its own computer and memory, so they test in parallel and the lead reports back in one thread.

Can we name them?

Yes. You pick each worker’s name and shape. Milo, Ivy and the others are just the names on this page.

What do the workers need from us?

A staging site they can reach from the internet, access to your GitHub repo and one flow to start with. Slack is optional. You can also write to them in the web app.

How do we talk to them?

Mention the lead in Slack, or write to it in the web app. Each worker has its own Slack app. Today, only the account owner can give them work. Everyone else in the channel can read the replies.

Are they really always on?

Yes. They keep an eye on your app through the day and keep working while your team sleeps. When something breaks or your app changes, the lead tells you in Slack. In the morning, it posts one summary of the night’s work.

Can they test pages behind a sign-in?

Not on their own computers yet. Signed-in flows run in your own CI with a test user, or on a person’s machine. The sign-in is created on that machine and stays there. A worker never asks for a password.

What if a worker is wrong?

Every finding comes with the steps, what it expected, what it saw and the test that shows it. You decide what to do with it. If your app changed on purpose, the worker asks, and it remembers your answer.

We already have flaky tests.

A worker runs your suite and proposes turning retries off. It waits for your yes. After that, a test that passes only on a second try counts as flaky, with its cause and how often it fails.

Which browsers and screens do they test?

Chromium, on each worker’s own computer. The tests it writes run at desktop size, and it also walks your pages at phone widths there. Safari, Firefox and native mobile apps are not supported.

Where do they run, and who can see them?

Each worker runs on its own computer, in a space only your organization uses. The account owner can watch its browser live.

How much does it cost?

Solo is $300 a month for one worker. Team is $1,200 a month for a lead and up to five specialists. For more than one team, talk to us.

What do we keep if we stop?

The tests. They are standard Playwright files in your repo.

Is it SOC 2 certified?

Not yet. Ask us and we’ll tell you who processes your data and how access works.