Environments/vex/verify/composal/environments
Each environment can be added, renamed and archived on its own; clearing search restores the saved list.
- · Add environment · attempted → New environment
- · Archive · attempted
Testing is slow to set up and goes stale as your product changes. Give Discovery a URL: agents learn your features and the flows inside them, write the Scenarios for you, and keep learning with every run.
No test steps to write first. Give it a URL and the areas that matter; agents work out the features and flows.
You never have to define a Scenario. Observed journeys become saved Scenarios with steps and expected outcomes.
What agents learn becomes knowledge every later run starts from, and finished runs keep it current as your product changes.
Add your URL and, if the app needs a login, a test account. Say which areas matter and what actions agents may take, then select Start discovery.
Agents sign in, click through each area, and work every feature end to end: searching, filling forms, saving, and checking what changed. Along the way they note the roles, data, and setup each flow needs.
Discovery returns recordings, observations, potential issues, and gaps where agents could not continue.
A separate authoring pass turns observed journeys into saved Scenarios with starting conditions, steps, and expected outcomes. Review them, then run them again.
Useful observations become documents in Knowledge: what features do, how to reach them, and what to set up first. Nightly Dreaming at 02:00 refines them from the previous day’s runs.
What you get back
Every discovery hands back what the agents saw, what they learned, and the Scenarios written from it.
Check each drafted Scenario’s steps and expected outcomes, and change what you want before you rely on it.
Run the saved Scenarios from Composal, from CI, or with com verify scenarios run, and get a result for each.
Open a document in Knowledge and edit it, or use com knowledge edit. Nightly Dreaming keeps your edits.
On a finished discovery, select Deep Scan and pick the flows it found. Each gets its own team of testers looking for bugs.