Documentation
Documentation
Step-by-step guides and reference for AI Testing Probe, from a Jira story to a reviewed Playwright test in your repo. Search the sections or ask a question; answers come only from these pages.
No matching sections.
You are chatting with an AI system (Anthropic Claude). Answers are based only on these docs, may be wrong and are not a binding offer; the current status of each capability is on the Capabilities page.
Example questions
No answer this time
You have asked several questions in a short time. Please wait a minute, or use Search meanwhile.
No answer this time
Ask AI is not available right now. Search still works.
No answer this time
We could not verify the request. Reload the page and try again, or use Search.
No answer this time
The answer could not be loaded. Check your connection, or use Search.
No answer this time
The question could not be processed. Shorten it (up to 500 characters) or use Search.
No answer this time
Something went wrong on our side. Please use Search for now.
Not covered in the documentation
The documentation does not answer this question. Try Search, or ask us directly in a call.
Clarified in the pilot conversation
Pricing, timelines, service levels and commitments are not answered here. These topics are clarified in the pilot conversation.
Get started
-
Quick start
What AI Testing Probe does, its six stages, and how one Jira story travels from requirement to a reviewed Playwright test in your own repo.
-
Concepts
Approval Gates, AI, human and system roles, user roles, maturity labels, evidence-backed findings, REPLAY mode and the hash-chained audit log.
Flow
-
Requirement review
How the pre-review and the code and data consistency check work, how analysts decide on findings, write fixes to Jira and get product owner approval.
-
Test cases
How AI drafts test cases from an approved story, how analysts review and edit them, and how approved cases are published to Jira or Xray.
-
Code generation
How approved test cases become template-bound Playwright code, verified step by step in the running app and delivered as a pull request.
-
Break-to-Verify
How Break-to-Verify proves a generated test catches a real fault by running it green, red on a deliberately broken app, then green again.
Platform
-
Architecture
Services, network isolation, the model gateway, REPLAY and live modes, what data reaches the model, the audit log and today's deployment model.
Setup
-
Integrations setup
What Jira Cloud, Xray Cloud, GitHub, GitHub Actions, Sauce Labs and the model provider need, with least-privilege accounts and permissions.
Start with one story.
In a 30-minute call we look at your stack and a few of your stories, and tell you honestly whether a pilot fits.