Dusk. The watch begins.
QA Beast learns your site, runs tests you describe in plain English, and messages you on Telegram when something breaks.
One eye on every page. The other on the flows that make you money.
First, it walks the territory.
QA Beast crawls every page it can reach and maps what's there: pages, forms, buttons, links, and the flows that matter, like sign up, log in, search and checkout.
On the way it writes down what's already wrong. That part's free on every plan.
Then it builds your starter suite: tests for the key pages, the main menu, links and images, the phone layout, and every form it found. Each one is tried in a real browser before you see it, so you start with tests that pass.
You tell it what to check.
New test Mobile, 390 × 844
On a phone, add the first plant that's in stock to the cart, go to checkout and make sure I can place the order.
The planner writes the steps and picks an engine for each one, and it tells you why. Playwright handles the exact parts. Jev, TypeSafe's fast decision model, handles the parts that need judgement.
Go to fernhouse.shop/shop
PlaywrightA fixed URL. Nothing to decide.
Pick the first plant that's in stock
JevStock labels differ card to card. This needs judgement.
Click "Add to cart"
PlaywrightKnown selector from the crawl.
Open the cart, click "Checkout"
PlaywrightStable link, same on every run.
Fill shipping with {{customer}}
Known fields, values from your test data.
Tap "Place order"
PlaywrightKnown selector: #place-order.
Does the page show an order confirmation?
JevA yes or no question about what's on screen.
A dry run, before you trust it.
The dry run plays the plan in a real browser and keeps a screenshot of every step, so you can see exactly what it saw.
Run it on desktop, mobile, or both. Schedules go as often as your plan allows.
Nothing moves.
Then it waits. Every fifteen minutes it looks again.
Something moved.
fernhouse
Checkout
1 bug caught
At 390px wide the promo bar is pinned to the bottom of the screen and sits on top of "Place order", so a shopper can't tap it. Desktop is fine.
QA Beast bot
Checkout (mobile) failed on fernhouse.shop. "Place order" is covered by the promo bar. Screenshot attached.
Re-runMute
The markup changed. So did it.
When a selector stops matching, Jev looks at the controls actually on the page and picks the one that does the job. The run is marked healed, and you get a patch you can apply with one click.
When a run really fails, you get the reason in plain language, not a stack trace.
Healed Search, step 3
Looked for#search-submit
Jev pickedbutton "Find plants"
- click('#search-submit')
+ click('button:has-text("Find plants")')
Apply patchExample
The rest of the kit.
When a step needs judgement, Jev reads the page, a local Qwen double-checks, and an image model looks at the screen. They vote.
Cookie walls, pop-ups, closed mobile menus, new tabs, frames and shadow DOM are handled before a step is called failed.
Standard tests built from your own pages and forms the moment the scan ends.
A test that needs a login waits for you to add one. It never makes one up.
A failed scheduled check runs again before anyone gets an alert.
Run any test at either viewport, or both.
Take any test with you as a Playwright spec file.
Start runs from your pipeline with scoped tokens.
SVG badges for a README or a dashboard.
One per site, for customers who want to check.
A DNS TXT record or a file, before any scheduled monitoring.
Every run, with a screenshot of every step.
Test data variables, and an encrypted vault for passwords.
Connect with one tap, in a private chat or your team's group. Failure and recovery alerts, a daily digest, Re-run and Mute buttons, /status and /run.
Alerts by email, or to any webhook. Slack-compatible.
What it costs.
| Free$0 | Pro$19/mo | Team$49/mo | |
|---|---|---|---|
| Sites | 1 | 5 | 20 |
| Pages per scan | 25 | 200 | 1,000 |
| Starter suite | Up to 5 tests | Up to 12 tests | Up to 20 tests |
| Your own tests | 5 | 50 | 250 |
| Runs a month | 150 | 3,000 | 15,000 |
| Schedule | Daily | Every 15 min | Every 5 min |
| History | 7 days | 30 days | 90 days |
| Alerts | Telegram, email | Telegram, email | Telegram, email |
| Also included | Mobile + desktop, self-healing, AI failure analysis, CI API + badges | Everything in Pro, status pages, priority queue | |
| Start free | Choose Pro | Choose Team |
Dawn.
Free covers one site, five tests and a daily check, with alerts on Telegram and email.
QA Beast botDaily digest, example
Good morning. Overnight on fernhouse.shop:
Checkout (mobile): 32 runs, 1 failed at 02:14. The promo bar covers "Place order".
Search: healed at 02:15. A patch is waiting for you.
Checkout (desktop): 32 runs, all passed.
