Pip, Otto, Mabel, FerrisDex, Moss, Nell, Ada, and the rest of the crew. Add your own.

A QA crew for your site

A QA crew that tests your site while you watch.

A crew of agents tests your flows, forms, accessibility, design, and security in real browsers. A referee re-runs every finding three times before it reaches you.

Runs on your own machine with your own model key. Local runs are free.

See how it works

Free plan for one person. Sign in after you install.

How it works

Three steps from install to a verified finding.

How it works

  1. Plan a run

    Pick a project and the flows to cover. Seat the crew: Pip for checkout, Otto for sign-in. The agents run in browsers on your machine, each in its own clean session.

  2. Watch the crew

    Every agent is a cursor with a pill that says what it is doing. Steer one, take over, or hand over an OTP when it asks.

  3. Get verified findings

    The ref re-runs each finding in three fresh browsers. A finding reaches your inbox when it happens in all three, with the video, the steps, and the network log to check it yourself.

  1. Point it at a site

    Type a URL or import a project from your host. Staging is a good first target.

  2. The crew runs in real browsers

    Each agent takes a section: checkout, sign-in, forms, small screens, the console. You watch the grid or take over.

  3. The ref checks every finding

    Three fresh browsers, three replays. Only findings that reproduce reach your list, with steps and video.

The crew

Meet the crew. Then add your own.

Every agent has a face, a color, and a section it owns. Click a face to read its card. Make your own and it lines up with the rest.

Day shiftWears the color
Night shiftWears it inside out
Your agentAny shift, any color
quibbit crewNo. 01

Pip

the checkout stickler

Shift
Day
Section
Flows
Color
Aqua

Counts every cent twice. Won't let a coupon apply two times.

Found on Acme

A coupon applies twice if you click Apply twice.

staging.acme.com · Checkout · step 18

The referee

Every finding is replayed 3 times before you see it.

The ref isn't on the crew. When an agent finds something, the ref opens three fresh browsers and tries to reproduce it. A finding reaches you only when it happens in all three, with the video to check it yourself.

Verified 3/3

Verified 3/3 or it doesn't count.

Replays make no model calls. Every verified finding is already a script you can keep as a Playwright test.

Sections

Twelve sections. Every one runs on your machine.

  • Flows

    Walks your real user journeys end to end, with test accounts from the vault.

    Pip, Otto, Ferris, Wren, ZuzuDex, Moss, Ada, Rook, Zuzu
  • Pen test

    Probes auth, injection, and access control on domains you have verified. Each run needs approval from an Admin or a member they allow.

    Pen test agents go faceless and numbered.
  • Design review

    Compares every screen to your design system and flags the drift.

    Nell, the pixel matcher
  • Accessibility

    Keyboard, contrast, labels, and reading order on every page it visits.

    Ada, the accessibility ally
  • Performance

    LCP, CLS, and payload on each step, on a fast and a slow network.

    Tally, the load-time timer
  • Responsive

    Re-runs the flow at phone, tablet, and desktop widths.

    Bea, the small-screen tapperRemy takes the night shift
  • Cross-browser

    Chrome, Firefox, WebKit, and Edge, side by side on the same steps.

    Remy, the browser hopper
  • Content

    Typos, placeholder text, broken copy, and prices that don't add up.

    Clem, the fine-print readerZuzu takes the night shift
  • Forms

    Validation, error states, and what happens on bad input.

    Mabel, the form fusserNell takes the night shift
  • Console and network

    Errors, failed requests, and slow calls, tied to the step that caused them.

    Dex, the console listener
  • Regression

    Replays the last release's findings to confirm each fix held.

    Moss, the regression keeper
  • Exploratory

    Leaves the happy path, opens every menu, and tries zero, then a million.

    Juno and RookTally and Rook

Pricing

Free to run locally. Pay for what the team shares.

Compare plans
  • Free

    $0

    Unlimited local runs for one person. 1 synced project, 3 days of history.

  • Solo

    $20a month

    One person. 5 synced projects, 30 days of history, 300 hosted minutes a month.

  • Team

    $39per member a month

    Shared runs, findings, and crew. 600 hosted browser minutes per paid member. Viewers are free.

  • Enterprise

    Talk to us

    SSO, SCIM, audit log streaming, your own bucket and runners.

Your keys, your machine.

Bring your own model key or gateway. Prompts go straight to your provider and never touch our servers. Passing runs keep no video. Failures stay on your machine until you share them.

Read the Security page

Free to run locally. Pay for what the team shares.

  • Free$0, unlimited local runs
  • Solo$20 a month, $17 yearly
  • Team$39 per member, $32 yearly
  • EnterpriseFrom $24,000 a year
Compare plans

Invite only for now.

Windows for now. Leave your email and we'll send a download link when your spot opens.

Put the crew on your site.

Download, add a model key, and watch your first flow run in about ten minutes.

Free plan for one person. Sign in after you install.