FAQ

Which computers are supported?

Local automation currently uses the directly distributed macOS app. Check supported systems and installer availability on the download page. The Mac App Store distribution does not provide local automation.

Can I use reopt QA without AI?

You can write tests, run them, inspect results, and compare runs without AI. Generation, investigation, and other AI features require a separate model connection. See Getting Started.

Does it work offline?

Stored reports and tests against accessible local services can work locally. External sites and model providers require network access. A local service may still depend on external authentication or APIs.

Where is data stored, and what can leave my computer?

The directly distributed app stores user data in agent-data under its local application data directory. This is separate from the replaceable runtime.

AI generation, investigation, fixes, and failure classification can send required code, page content, and failure records to external providers. Conditional fix approval evaluation can involve a separate transmission. Secret masking does not remove every kind of sensitive information, and screenshots need separate review.

Manual fix approval controls file application; it does not block all AI data transmission. Do not use AI features for data that must not leave the computer. See Architecture and data flow.

Are AI costs included in the app fee?

Usage charged to your model provider account is separate from app and setup support costs. Conditional fix approval evaluation may also require separate authentication and incur costs. App pricing is listed under Pricing.

Project cost and token limits are per AI session. An empty limit means unlimited, and monetary tracking depends on the provider reporting cost. AI failure classification is not covered by that session budget. These settings are not a total billing cap; check provider-side limits too.

Does a page on the service map mean it is verified?

No. The map shows discovered pages and connections, not proof of every user state or successful feature. Find relevant pages in the Service map, then test actual user goals.

How do I prepare scenarios and personas?

Create a definition and executable code through Verification → Test list → New test, or use Generate with AI. Descriptive steps alone are not a complete executable check.

Manage personas in Settings → Personas. Templates may still need environments and credentials. A role label does not grant permissions in the target service. See Scenarios and personas.

Can I compare mobile and desktop behavior?

Run the same test with personas using different viewports. Check account, environment, and locale when comparing. A mobile browser viewport does not reproduce every behavior of a physical device.

What if execution stops or fails?

Check the run state and stop it if needed. If the engine itself is unavailable, inspect it in Agent Manager and restart it. Partial records or missing artifacts do not represent completed verification.

Read the failing step and error, then inspect available logs, network records, screenshots, and traces. Check AI hypotheses against evidence. See Reports and analytics for comparisons and reruns.

Will schedules run with the app closed?

The local engine must be running. Missed intervals are not all replayed later. See Scheduled checks.

Can external AI read results and fix tests?

An MCP connection can read the latest reports and session states. Starting tests, exploration, investigations, or fixes requires project permission. Fixes started by external agents always require human approval in the app. The bridge currently comes from the repository CLI, not the app installer.

Can I share results?

Individual reports export as CSV or Markdown; the list supports batch CSV and ZIP. Check screenshots, logs, and URLs for sensitive information before sharing. Sharing test definitions does not prepare another person's environments or accounts.

Are the website's charts and numbers my project data?

Website animations and sample metrics illustrate workflows. They are not actual results or guarantees of product quality. Inspect your project's execution evidence in the app.