AI Test Fix: Grounded Repair of Failing Tests
What It Does
AI Test Fix turns a wall of red into a short review queue without turning red into green by force. Before it proposes anything, it decides whether the test is wrong or the API is: a test whose expected field was renamed is a test defect, while a 500 on a valid request is a possible product bug and is left for a person to report. Each fix is grounded in evidence (the specification, the requirement the test covers, the run, and the behavior observed on passing runs), and a fix that cannot cite its evidence is rejected. You see the change as a per-field before/after diff and choose Apply to form to edit it first, Apply & Save, or Discard. For a failed pack, failure analysis groups the failures into causes and makes one AI call per cause rather than per test, and the bulk review lets you drop any test before applying. Projects can opt in to automatic suggestions when a test fails twice in a row; suggestions stay pending until someone applies them unless auto-apply is turned on. When a test failed because its record is missing, the fix points to the positive test in your project that lists or creates such records. Creating one asks first, never runs against production, and uses only a value from the run you just made, so an old id is never copied.
Overview
When a test starts failing, Fix with AI reads the test and its intent, the specification, the requirement, the last run and what the platform has learned about the endpoint, then classifies the failure: test defect, data issue, possible product bug, environment issue or flaky. For a test defect it proposes a fix (updated assertions, parameters, headers, auth or request body) with a confidence score, and shows it as a before/after diff you apply or discard. Suspected product bugs and environment problems are never patched, a negative test is never weakened into one that cannot fail, and every change must cite its evidence. When a whole run fails, the same repair works per cause, so one review covers every test with that cause. If a test failed because the record it asks for does not exist, the fix names the test in your project that lists or creates such records, and can fetch a real one for you.
Key Capabilities
How It Works
- 1
A test fails during a run and is flagged with the failure reason
- 2
Open the failing test and choose "Fix with AI", or open the run’s causes
- 3
Review the classification, the confidence and the before/after diff
- 4
Choose Apply to form to edit first, Apply & Save, or Discard
- 5
Re-run to confirm the test passes
Available on
All Plans
Included in the free trial — no credit card required.
Documentation
Step-by-step guides for setting up and using AI Test Fix.
- Fix with AI: Grounded RepairHow Fix with AI grounds each proposed repair in your API contract, your requirements and the last run, and what the workbench Diagnostics show about what was actually sent.
- Fix Tests with AIWhen a test fails, use Fix with AI to get a proposed patch you can review and apply—plus a fix-suggestion lifecycle that classifies whether the failure is a bad test or a product bug.
Go deeper
Guides and playbooks that show this capability in practice.
Related Features
Failure Analysis
A handful of causes with verdicts, instead of a long list of red rows.
Learn moreQuality Agent
A scheduled review that finds what needs attention and, on Trial and Enterprise, prepares fixes for you to approve.
Learn moreChange Detection
Detect breaking API changes before they reach production.
Learn moreTry AI Test Fix Today
Start your 15-day free trial — no credit card required.