Describe your ideal customer and their goal. An AI stranger tries your web app in your own browser, and shows you where they get stuck.
The task vanished after refresh.
Choose who to simulate and what they should try to accomplish.
Say what your product is and who your customer is. Give that customer one goal, the one you want a stranger to reach.
Organize my next launch.
When your invite arrives you install once, sign in once with au login, and start each run with au run. It opens your own Chrome on your own machine and your own IP. No source code access, no plan of yours is touched.
macOS, your own Chrome, and a terminalau login
Signed in. 5 sessions.› au run
Chrome opened. Watching the stranger work.Every finding carries the screenshot taken at the moment of the claim and the steps to reproduce it. Fix it, then run it again.
From things that break to things that make your product harder to use.
Actions that fail or produce the wrong result.
Confusing choices and obstacles that make simple tasks harder.
A finding that has no screenshot of its own does not reach you.
Claims the screenshot contradicts are removed before you read them, and the counts are printed on the report, whatever they say.
Know what happened, why it matters,
and how to reproduce it.
Goal: buy a few items and place a second order.
Could not continue past checkout, because the cart total stopped matching the quantities.
Pricing · Sign up · Empty states
Every claim checked against its own screenshot. The removed ones are counted on the report.
What leaves it: the screenshots, the page text, the values the stranger typed, and the event log, redacted before they are sent, for the check, the audit and the report. Kept 90 days, deleted on request.
Read the privacy policyFirst run free for accepted beta teams, on your real product. Full report.
Founding pack: $29 for five sessions. Sessions never expire. Another five for $29 whenever you need them.
A session is one run: one stranger, one motivation, one report. A run that fails, or never gets past your front door, does not use a session.
Runs on your machine, in your own browser. No terminal? Join the waitlist and tick the box; you hear first if a version without one ships.
Get your first run freeInvites go out in small batches.Founding beta pricing. If it changes, everyone on the waitlist hears first.Each session is one run: one stranger, one motivation, one report.
One stranger explores one goal for 10 to 14 minutes. It is one user for one session, not an audit: it finds where a person gives up, not every bug. Two runs find different things.
Web apps in a browser. A run can create a real account and real data, so staging is the default target. It does not test native mobile or desktop apps. A run that never gets past your front door does not use a session.
Yes, on a Mac with Chrome installed. You install once, sign in once, and start each run from your terminal. No terminal? Join the waitlist and tick the box, and you hear first if a version without one ships.
No. The model calls run on our own gateway. Nothing of yours is metered.
Staging is the default. To run against production you confirm once, per project, that you own the product or have authority to test it, and you see exactly what the run will do before it starts.