Skip to main content

The idea behind onboarding flows

Getting started with a new application is usually the slowest part of test automation: someone has to understand the application, decide what's worth testing, and then write the first tests by hand. msg.ZenTestAI's onboarding flow lets an AI agent do the exploring and drafting, while you stay in control of what is kept.

From "human designs everything" to "AI proposes, you approve"​

Onboarding is where the autonomy spectrum is most visible. Instead of starting from a blank test list, you point an agent at your application and it:

  1. Scans the application — navigating it like a user to discover its screens and the flows between them.
  2. Proposes test coverage — a draft list of the tests worth having, based on what it found.
  3. Lets you review — you edit, accept or reject the proposals; nothing is committed without you.
  4. Builds the tests you accepted — turning them into ordinary test cases.

The result is a set of normal, reviewable tests that run on the same deterministic core as everything else. The AI accelerated the setup; it didn't create a black box.

Two things the scan produces​

A scan produces two durable artefacts on the application:

  • An Application Map — a visual graph of the application's screens (surfaces) and how they connect. It is a living picture of the application that you can refresh as the application changes.
  • A seeded Knowledge Base — the agent records what it learned about the application as articles, so later assistant sessions start with that context instead of rediscovering it every time.

Explorative tests are temporary​

While scanning, the agent needs to actually drive a browser, so it creates throwaway explorative tests to navigate the application. These are an implementation detail of the scan — they are hidden from your normal test list and are cleaned up automatically a short time after the scan. You will see them only as executions named like "Explorative Scan – …" or "AppMap Scan – …" in the execution list.

How much you let it do​

You decide how far the flow goes each time you start it:

  • Map only — just (re)discover the application's screens, without proposing tests.
  • Map and propose — discover the screens and draft a coverage proposal for you to review.

And on the review step you remain the gate: every proposed test is yours to accept, edit or drop before it becomes a real test.

Where to go next​

  • Application Map — the concrete walkthrough of running a scan and working with the resulting map.
  • The AI Assistant — the chat surface the onboarding sessions are built on.