Tests in AutoTestX are steps, not scripts. AI drafts them from your test cases, analysts read and change them without a developer, and every run leaves evidence your change board can read.
Each step says what it does: Select Vendor from lookup, Expect message. Values come from your data sheet or your environment's secrets, never typed into the test.
The test editor in the docs →
The failed step, what was expected, what appeared, and the screen at that moment. No digging through logs to explain a red result in a meeting.
Runs and reports in the docs →
Paste a written test case and the optional AI step writer drafts steps from it, reusing fields your saved tests already find. Each draft step shows the sentence it came from, and nothing is saved until someone reviews it.
No run acts on an AI answer: every step executes exactly as written, so the same test on the same data gives the same result. If you turn on AI healing or AI diagnosis, the AI only files suggestions and explanations for a person to review, and an organization admin can turn AI off altogether.
AI in AutoTestX →
A renamed button or a regenerated id shouldn't cost an analyst an afternoon. When a step can't find its field, AutoTestX compares the page with how that field looked on the last passing run and proposes the match, with the evidence and a confidence score.
Nothing changes in the test until a Test Developer or Test Manager approves it, from the Self-healing page, the run report or the editor, or, on Business and Enterprise, a rule they set applies a safe fix, with an audit trail and one-click revert. A run never acts on a healed field in a risky step like Delete or Approve, and when the field is really gone, the run says so instead of hiding a defect.
How self-healing works →
#btn_8f3arole button, name “Save changes”The run continues with the found field, the step is marked healed, and a fix waits for approval.
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.
| Task | Manual UAT spreadsheet | With AutoTestX |
|---|---|---|
| Re-running a regression cycle | Analysts click through the spreadsheet, case by case | One suite run, from the app, a schedule or your CI pipeline |
| Testing with many data rows | Each row repeated by hand, or skipped | One test, one run per spreadsheet row |
| Evidence | Screenshots pasted into Word | A report per run: steps, data used, a screenshot of every step |
| Small screen changes | Noticed when someone spots them | Flagged as drift, or healed with a fix to approve |
| Who maintains the tests | Whoever wrote the spreadsheet | Analysts, in a plain-language step editor |
| When the application is upgraded | Start the spreadsheet again, at a bigger scale | Run the same tests as the baseline |
Every test is a JSON document in a documented format. Copy it out from the DSL tab, keep it in version control, and run it with the command-line runner, which returns a non-zero exit code and a JSON result your pipeline can read.
The capability list for the October 2026 release of the hosted service. The website, the documentation, our demos and our proposals all describe this list. We'd rather you hear about a gap from us before a pilot than discover it during one.
| Capability | Status | Details |
|---|---|---|
| Recording and building tests | ||
| Recording on Windows | Works, with a caveat | Through the AutoTestX Agent, installed per user with no admin rights. The installer is not code-signed yet, so SmartScreen or endpoint protection may warn about it. More → |
| Recording on macOS | Works, with a caveat | Through the Mac agent, installed with one Terminal command because it is not yet signed and notarized by Apple. More → |
| Recording on Linux | Not planned | Use a Windows or Mac computer, or a Windows VDI. |
| Visual step editor with a plain-language view | Works today | Build and change tests from steps anyone can read. |
| Lookups, tabs, frames and table rows | Works today | Table rows found by business key. In the Maximo edition, Maximo’s own messages become checks. |
| Data-driven tests | Works today | Excel or CSV sheets, an API request, or a pool of single-use records. |
| API steps and custom code steps | Works today | Create or check data through an API; custom code for the rest. |
| AI step writer (optional) | Works today | Drafts steps from a written test case. A person reviews them before anything is saved. |
| Native mobile or desktop (non-browser) applications | Not planned | Browser applications only. |
| Running tests | ||
| Runs from the app, suites and schedules | Works, with a caveat | They run on the AutoTestX server, which must be able to reach your test environment. More → |
| Scheduled runs | Works today | Daily, once per suite, on Free and Team; any timing (cron) on Business and Enterprise. |
| CI pipelines: GitHub, GitLab, Jenkins, Azure DevOps | Works today | Command-line runner with exit codes and a JSON result. |
| Watching a run on your own screen | Works today | Runs in the tester’s browser through the agent, inside your network. |
| Self-healing and AI | ||
| Self-healing: suggested fixes you approve | Works today | When a field moved or was renamed, the run finds it, carries on and proposes the fix with its evidence. More → |
| Self-healing: automatic fixes by rule | Works today | Rules a Test Manager sets apply safe fixes, with an audit trail and revert. Business and Enterprise. |
| AI-assisted healing and AI failure diagnosis | Works today | Optional; an organization admin can turn AI off. Runs on the AutoTestX AI key or your organization’s own Anthropic key. |
| Healing timeouts, pop-ups and changed flows | Planned | Fix these in the editor today. |
| Amazon Bedrock or Google Vertex as the AI provider | Planned | Anthropic’s API only today. |
| Evidence and reports | ||
| Run report with a screenshot of every step | Works today | Steps, data used, expected and actual values; video and trace on failure. |
| Release report | Works today | Regressions, new and known failures across one release’s runs. |
| PDF evidence report for audits | Works today | Run, suite run and release reports download as PDF, with the screenshots inside. |
| Retention settings for runs and screenshots | Planned | Kept until deleted; deleted on request today. |
| Security and access | ||
| Five roles per project, organization admins | Works today | Admin, Test Manager, Test Developer, Tester, Viewer. |
| Secrets redacted from logs and reports | Works, with a caveat | Text is redacted. Screenshots are not masked: they show what was on the screen. More → |
| Single sign-on (SAML / OIDC) and MFA | Planned | Email and password accounts today. |
| SOC 2 or ISO 27001 | Planned | Not started. The IT pre-check and questionnaire answers today. |
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.