Everything your cast can do
An actor is an AI user driving a real browser. Every feature below points one at your product: an interview, a checklist, a design review, a deploy. Each result arrives with the screenshots behind it.
Actors that use your product
An actor is a persona with instructions, knowledge and its own browser. It observes the page, takes an action, checks the result and repeats until the task is done.
Actors
Give a persona a role, a background and the knowledge it needs, from product facts to test credentials. It works through your product the way that person would.
Real identities
Every actor has its own email inbox and browser session. Sign-up codes, magic links and verification emails arrive and get used, so those flows are tested end to end.
Locations and devices
Pick a city and the actor browses through a residential proxy there, on a phone, tablet or desktop viewport. Geo-gated content and local pricing get checked from where your users are.
Runner
The assistant inside the app. Ask in plain language and it creates actors, checklists, interviews and automations, then takes you to the result.
Hear it before you build it
Put an idea, a prototype or a live flow in front of a panel and read what they did, where they stopped and why.
Interviews
Write a guide of sections and questions, cast a panel, and each actor works through your product while it answers. Participants run side by side, so a five-person round takes an afternoon.
Synthesized reports
When the last session ends, the round becomes themes. Each theme cites the sessions, quotes and screenshots behind it.
Interview video
BetaA round also comes back as a synthesized video edit. Every theme with video evidence has a Watch button that jumps to the moment behind it.
Actor chat
Message an actor, share a screen or float a headline, then ask a follow-up. The actor keeps the context, so the conversation stays coherent.
Prototypes and Figma
Anything with a reachable URL works: production, staging, a preview deployment, a Figma Make site or a Claude artifact. Create a project straight from a Figma file.
Check every change against your requirements
Write the requirement once. An actor checks it on every deploy and shows you the screen it was looking at.
Checklists
An actor runs each plain-language check in a hosted browser and returns a screenshot, a verdict (passed, failed, skipped or inconclusive) and, for weighted checks, a percent or points score. A start page that will not load is reported as Site unreachable or Site error.
Run video
BetaTurn it on in a checklist's Instructions tab and each finished run gets a short narrated video: the outcome, then every check with the steps, the verdict and the screenshot. The words come from that run's results. Paid plans only, off for checklists marked 18+, and each video counts toward compute and AI usage.
Checklist library and replays
Start from templates for sign-up, checkout, pricing pages, forms, accessibility, mobile and search. After its first run a checklist replays the recorded path, so repeat runs are faster and land the same way.
Design reviews
Point actors at a page or a Figma frame and each one reviews it in character. Findings come back tagged by severity and category, in the dashboard or in a Slack thread.
Guidelines
Give a project the rules it is judged by: brand, tone of voice, design system, legal copy. Import them from a web page, a published Storybook or a GitHub repository, and findings cite them.
Topology
PreviewEvery screen in your product on one live map, with the paths between them and broken pages marked. Each release shows what was added, changed or removed.
Automations
Chain checklists, conditions, delays, notifications, pull request reviews and GitHub comments into one run. Start it by hand, on a schedule, from a webhook, or on a GitHub, Vercel, Linear, Slack or Figma event.
Know what to fix next
Findings from actors and comments from people land in one place, each with the evidence attached.
Insights
Observations that fall outside pass or fail: a confusing step, a broken link, a result nobody wrote a check for. A short list instead of a pile of recordings.
Recordings and screenshots
Every run is recorded. Step through each action with the screenshot captured at that moment, or watch the session play end to end.
Shared reports
Turn on a private link and anyone can open a read-only report, with checks, evidence and screenshots, without an account. Switch the link off at any time.
Feedback widget
One script tag on your site or prototype. Visitors pin a comment to the place it applies, with a screenshot, the page URL and device details attached.
Feedback triage
Comments group by page and prototype version, and move through new, reviewed, resolved and dismissed. Summarize the open queue into themes and an implementation brief.
Issues and goals
Failed checks, open design review notes, untriaged comments and insights land on the Issues tab, each linked to its evidence. Link insights and feedback to the goals they support.
Runs where your work already happens
Start a run from a conversation, a pull request or a Slack thread, and send the result back to the same place.
Stunt Double Studio
The free local app in the Claude Code plugin. Your agent runs a checklist in Chrome on your machine, on your existing Claude plan, and each run is kept as files in your project.
MCP server
Connect Claude, the Cursor plugin, or any MCP client to app.stuntdouble.io/api/mcp for cloud runs, interviews and Index reports. Sign-in is OAuth 2.1, with separate scopes to read, write and start runs. Listing tools and the public Index reads do not need a token.
Enterprise Managed Auth
A workspace admin trusts the company identity provider in Settings, then Security. Members then connect Claude with no separate Stunt Double sign-in, and the token still acts only as that person.
Claude
Install the Claude Desktop extension in one click, then run checklists, workflows and interviews from a conversation.
GitHub and Vercel
Run a checklist on every preview deployment and post the outcome on the pull request, so a regression shows up before merge.
Slack and Linear
Ask for a design review in a Slack thread and get run summaries where your team talks. Push findings into Linear as issues, ready to triage.
Devices
On a paid plan the Mac app runs the browser on a machine you choose. Private and internal addresses need Startup or above. Enterprise adds VPN, VPC and self-hosted apps. The app dials out, so there are no inbound firewall changes, and a domain allowlist sets what it can reach.
Mac app
The desktop app for macOS. On a paid plan it pairs as a device, so actors run in a browser on your machine. Solo and Hobby open public URLs. Private addresses start on Startup.
Get started
Your cast is ready to start
Create your first actor in minutes.