poolabsGet started

Multi-user AI testing

Put AI users
inside your app

Poolabs launches multiple autonomous AI users in separate browser sessions and makes them interact with your application — and with each other — like real customers.

See it in action

Why multi-user testing

Most bugs only show up when two people use your app at once

01

Real browsers, side by side

Each AI user gets its own isolated browser session with its own cookies and login, so a host and a guest really are two different people to your app.

02

Goals, not scripts

You describe what each user wants to achieve. The AI decides what to click and type, so changes to your layout do not break the test.

03

Proof for every step

Every click, page and error is recorded with screenshots from the actual browser. If nothing ran, the run says so instead of guessing.

How it works

  1. 1

    Add your app URL

  2. 2

    Define AI users & roles

  3. 3

    Give each user a goal

  4. 4

    Start the run

  5. 5

    Review every action

What every run records

Nothing here is estimated. Each item is written only from what the browsers actually did.

Action
timeline

Every step each AI user took, in the exact order it happened.

Step types recorded

clicktypenavigatewaitdonefail
02

Live browser state

Where each user is right now while the run is going.

03

Screenshots

Captured from the real browser after actions and on finish.

04

Issues with evidence

Problems logged with severity and the steps that led there.

05

Run outcome & report

Pass, fail, partial or error with duration and summary. Copy, export or share it.

Runs on your worker

Browsers run in a separate Node + Playwright service you deploy. It authenticates to Poolabs with a shared secret.

Passwords stay with you

Only test account emails are saved here. Passwords never leave your worker. Read about security.

Questions

What do I need to run a test?+

A Poolabs account, a publicly reachable URL for your app, and the browser worker deployed with its address and secret added to the app. Without the worker, runs are marked as not started.

Do AI users need real accounts?+

If your flow needs sign-in, give each AI user a test account. Only the email is stored here; passwords stay on your worker.

How many users per test?+

As many as your scenario needs. Each gets its own isolated browser context.

Is anything simulated?+

No. Every status and action you see was reported by a real browser session.