Skip to main content

Test apps and prototypes built with AI

Building got cheap. A working prototype exists an hour after the idea, from v0, Figma Make or a Claude artifact, and nobody has watched a user try it. An actor opens the published link in a real browser and works through it as your user, so the first person to find the dead end is not a customer.

Prototype testing workflow
  1. 01An idea
  2. 02v0, Figma Make or a Claude artifact
  3. 03A published link
  4. 04A panel of actors
  5. 05A report

How it works

Captures on this page are from a workspace testing Parlour, a demo app simulating a restaurant booking flow.

  1. 01

    Publish a link

    Anything with a reachable URL works: a v0 link, a Figma Make site, a Claude artifact, or a preview deployment.

    Parlour, a demo app simulating a restaurant booking flow, open on a phone.
    A published app, as the actor opens it on a phone.
  2. 02

    Put it in front of a panel

    Run an interview to watch a group of actors try the idea, or a checklist when you already know what has to work.

    An interview panel in Stunt Double: five participants, each an actor with a different persona.
    An interview panel: five actors, five different participants.
  3. 03

    Change it and run it again

    Regenerate the app, publish the new link, and run the same interview or checklist. The comparison is the point.

    An interview report in Stunt Double: themes, each citing the sessions behind it.
    The report: themes, each citing the sessions behind it.

The requirements you would write

Plain language, one per line. An actor checks each one in a real browser.

  1. 01The main action on the first screen is obvious
  2. 02Every item in the navigation goes somewhere
  3. 03Submitting a form gives feedback
  4. 04The layout holds on a phone viewport
A completed checklist run in Stunt Double: three checks on a booking flow, two passed and one failed, with a weighted score of 72%.
A checklist run: each requirement comes back passed or failed, with the screenshot the actor took.

What comes back

  • 01

    Each check passed or failed, with the screenshot and the reasoning behind it

  • 02

    A recording of the whole session, step by step

  • 03

    Insights for anything a check did not ask about

  • 04

    For an interview, a report grouped into themes, each citing the sessions behind it

Where it stops

A prototype built on placeholder data tests the flow, not the content. Run it again once the real copy and data are in.

Questions people ask

Which tools does it work with?

Any tool that gives you a link. Stunt Double needs a page to open, not an integration, so production, staging, preview deployments, Figma Make sites and Claude artifacts all work.

Can I run it without leaving Claude?

Yes. Connect the MCP server, or install the Claude Desktop extension, and ask Claude to test the artifact it has built.

Keep reading

Go further

Related terms

Other use cases

Try it on your own site

Give an actor your URL. It opens a real browser, works through the site, and shows you every step it took.

Free to try, no sign-up needed.