A renamed button or a regenerated id shouldn’t cost anyone an afternoon. AutoTestX finds the element the step meant, keeps the run going where that is safe, and proposes the fix with its evidence. Your team decides; rules can take the safe decisions for you.
The current and the proposed target, every signal that was compared with its value then and now, a confidence score, and the run that found it. Approve it, edit and approve it, or reject it. Only an approval changes the test, and the new target is stamped so you can always see where it came from.
The test never changes during a run. A healed run tells you the test would pass with the fix.
#btn_8f3arole button, name “Save changes”Illustration of a suggestion card
Healing runs only when a step’s element cannot be found.
Another way of finding the element already worked, so the step was warned. Suggest making it the main one.
Free, deterministicCompare the page with the element as it was on the last passing run: its name, label, test id, attributes, text, heading, row and position.
Free, deterministicTier 1 found look-alikes but couldn’t prove one. An AI chooses among them, or says none. Off until a project turns it on.
One AI generation per new questionAn AI pick is never acted on during a run and rules never apply one: it always waits for a person. AI in AutoTestX →
The run continues with the found element, the step is marked healed, and a fix waits for approval. One run finds every broken step, not just the first.
The step still fails, with the proposed fix attached, because acting on it during the run could do harm or hide a defect.
These fail clearly, with the reason, so a real regression is never painted green.
A rule decides some suggestions without a review: apply low-risk click fixes seen in two runs or more, always require review for payment tests, or reject fixes you never want. Rules act only after a run has finished, and a fix they apply goes through the same checks as yours and can be reverted.
Try a rule safely first: in dry-run mode it changes nothing and records what it would have done.
A rule says
Apply automaticallyAlways require reviewReject automatically
The self-healing audit trail is append-only and hash-chained, and Verify integrity reports any break. Each entry names the actor (a person, or a rule and its version), the test and step, the locator before and after, and the evidence it was decided on. Any applied fix can be reverted in one click.
Each application can use the profile that matches how its screens are built, or your organization’s own. A preview shows what a change would have done before you save it.
Most applications: stable ids and test hooks count most, then the name and label.
Applications that generate ids and classes, such as Maximo, Salesforce or ExtJS: trust the label, heading and table row over the id.
Applications with data-testid on every control: a matching hook decides, a renamed one is a real change.
Screens with many look-alike controls in rows: the row and section count more.
In 30 minutes: a live demo of AI drafting, healing and explaining tests on a representative flow, and what your team would need for a pilot.