FAQ

Answers to common questions about Mozark MCP.

Do I need to know tool names?

No. Describe what you want in plain English. Tool names appear in these docs only so you can recognise what the assistant is doing.

Which AI clients are supported?

Any MCP client that supports remote (Streamable HTTP) servers — Claude web/desktop, Claude Code, Cursor, VS Code Copilot, Windsurf, Codex and others.

Can the assistant change things without asking?

Reads happen freely. Creating, saving, updating or deleting always follows an explicit request, and multi-item deletes and plan changes require confirmation.

Does the AI model see my credentials?

No. Sign-in happens in your browser and the token is held by your client. The one exception is the app-upload curl command, which embeds a short-lived token so a coding agent can run it — don't share or log that command.

Why does the assistant ask which organisation to use?

Because saving into the wrong organisation is silent and costly. It only asks when your context is an unconfirmed account default and you belong to more than one organisation.

Why doesn't the assistant just start running my test case?

Capture options can't be added after a session starts, and the app package can't be guessed. Two quick questions up front avoid a wasted device session.

Can I run the same exported script on iOS and Android?

Yes, if it uses text locators — which is the default and the only thing optics_export_yaml produces.

Is there a rate limit or quota?

Device time and AI usage count against your Mozark plan. Ask "Show me our usage report" to check.

Where do screenshots, videos and logs go?

They're stored with the session in Mozark. Ask the assistant to list session assets or outputs to retrieve them.