Live simulator streaming
Stream live iOS simulators and Android emulators — plus Apple TV and Android TV — into any modern browser. No Xcode, no Android Studio, no USB cables. Your team gets the same device on day one that it gets on day one hundred.
- ✓Full touch, keyboard, and gesture support
- ✓Low-latency streaming from the cloud, no plugins or installs
- ✓Share an app build with a teammate via link — they get their own live session
AI-powered test automation
Describe a test in plain English or record it by interacting with the device. The AI agent reads the live UI and executes taps, swipes, text input, and assertions. Broken flows surface in seconds instead of burning a full run.
- ✓Natural language test descriptions
- ✓Adapts when your UI changes, instead of breaking on every layout shift
- ✓Broken flows fail fast with a timeline of what went wrong
24 tools: tap · type · swipe · assert · scroll-until-visible…
Visual regression
Catch UI regressions before they ship. Every suite carries warn and fail thresholds. After each step, VibeView compares the screenshot to the stored baseline and flags anything outside tolerance — mid-run, not only at the end.
- ✓Mid-run warnings on both AI and hybrid modes
- ✓Accept a single step's screenshot as the new baseline with one click
- ✓Tighten baselines iteratively as the UI stabilizes
Apple TV and Android TV
TV sessions run alongside phone and tablet. The AI agent handles d-pad navigation by targeting elements directly instead of chaining arrow presses, so TV tests stay short and readable.
- ✓Apple TV and Android TV in the same browser
- ✓Focus-based interaction — no manual d-pad choreography
- ✓Same dashboards, same suites, same run detail
d-pad navigation · focus-based interaction
CI integration
Run tests and suites from the command line. Results export as JSON and JUnit for any CI/CD pipeline, and any run can be pinned to a specific build for bisecting regressions.
- ✓CLI for local runs and CI/CD pipelines
- ✓Pin a run to a specific build when you need to
- ✓JSON and JUnit output for any CI system
- VibeView / smoke-flows (ios-17)41s
- VibeView / regression (android-14)1m 12s
- VibeView / visual-drift (tvOS)1 warning · +0.8%
- VibeView / e2e (android-tv)In progress…
This branch has no conflicts
VibeView status checks gate the merge button.
Tests as YAML
Define tests in YAML with import and export. Store them alongside your code so test suites are versioned, reviewed, and shared like any other source.
- ✓Store tests alongside your code in version control
- ✓Import and export suites as YAML
- ✓Review test changes in pull requests
Embeddable iframe
Drop a live device into any web page with a single iframe. No SDK to install, no build step — paste the snippet and it streams.
- ✓One iframe, no npm install
- ✓Available on the Starter and Pro plans
- ✓Session isolation — each visitor gets their own device
Every detail,
covered.
From session hygiene to audit logs, the platform handles the details so you can focus on shipping.
Build management and shareable links
Upload .app and .apk files with a short note describing what changed. Every build has a one-click launch into a sandbox pinned to that exact version, and the link you share reopens the same build — the pin survives session reloads.
- ✓Attach build notes ("pre-release hotfix, reverted feature flag")
- ✓Launch any historical build in a fresh session with one click
- ✓Run detail pages show which build was tested, with the note attached
Hybrid test mode
Hybrid mode replays recorded steps first and hands off to the AI agent when the UI changes. Fast on stable flows, resilient on the ones that drift. Fuzzy element matching decides how strictly replay should match — per run.
- ✓Recorded replay with AI fallback when the UI shifts
- ✓Per-run fuzzy matching: strict, text-substring, or default
- ✓Every step logged, so you can see where AI took over
Inline hybrid editor
Fix a recorded step without re-recording the whole flow. Edits validate against the source recording, so a change that breaks the anchor chain surfaces inline instead of at run time.
- ✓Edit text, targeting, or order of recorded actions
- ✓Validation against the source recording catches broken edits early
- ✓Smaller maintenance cost than re-recording
Session hygiene
Sessions start clean. The on-screen keyboard is suppressed at boot, first-launch notification dialogs are dismissed, and common system alerts are auto-dismissed mid-session. Fewer flakes from stray OS dialogs, especially in CI.
- ✓On-screen keyboard suppressed on boot, iOS and Android
- ✓First-launch notifications and autofill prompts dismissed
- ✓System alerts handled automatically during the run
Sub-flows
Reusable step sequences for login, checkout, or navigation. Compose them into larger suites to cut duplication and keep flows tight.
- ✓Named, versioned sequences you can drop into any suite
- ✓One fix updates every suite that uses it
- ✓Good fit for onboarding, auth, and repeated setup
GitHub status checks
Post test results as commit statuses on GitHub. See pass/fail directly on pull requests and block merges until tests pass.
- ✓Commit statuses on every pull request
- ✓Block merges when tests fail
- ✓Works with the VibeView GitHub Action
Webhooks and Slack
Get notified when test runs complete. Configure webhook URLs or connect a Slack channel for real-time alerts on regressions.
- ✓Webhook URLs for downstream pipeline steps
- ✓Slack channel notifications for run completion
- ✓Routed by status, so green runs stay quiet
Bring your own key
Connect your own API keys for Anthropic, OpenAI, Google, or OpenRouter. Keys are encrypted at rest. No vendor lock-in on the AI provider.
- ✓Anthropic, OpenAI, Google, OpenRouter supported
- ✓Keys encrypted at rest
- ✓Swap providers without changing tests
Organization and access control
Manage teams with role-based access and organization-scoped resources. Issue API tokens for CI/CD and programmatic access.
- ✓Roles: Admin, Developer, Viewer
- ✓Organization-scoped sessions and resources
- ✓API tokens for CI/CD pipelines
Session history and audit logging
Every session, mutation, and administrative action is logged with timestamps, user attribution, and device metadata. Trace a test run, review a support session, or keep a full audit trail for compliance.
- ✓Every session logged with user and device context
- ✓Full audit trail for administrative actions
- ✓Filter by user, date, or action type
Test analytics
Track reliability over time with pass rates, duration trends, and flakiness scores. Identify flaky tests before they slow a release and measure quality improvements over weeks, not days.
- ✓Pass rates and duration trends
- ✓Flakiness score per test
- ✓Failure pattern analysis and grouping
The device lab,
finally in a tab.
No credit card required. Start testing in minutes.
Already have an account? Sign in
Common questions
What does the AI testing agent do?
The AI testing agent (beta) reads the live UI of your app and drives it — tapping, typing, scrolling and asserting — from steps you describe in plain English. Because it works from what is on screen rather than from stored selectors, a moved or renamed element does not automatically break the test. Recorded steps replay first for speed, and the agent takes over only where it is needed.
How does this run in CI?
A command-line tool uploads a build and runs a test suite against a cloud device from any pipeline, returning JUnit output your CI already knows how to read. With GitHub Actions the result also appears as a status check on the pull request, so a regression blocks the merge instead of being noticed later. Any CI system that can run a shell command can run it.
Which platforms can I test?
iOS and Android are fully supported, as are the Apple TV and Android TV simulators, which are driven by d-pad navigation rather than touch. Roku is in beta: you can drive the remote and inspect the UI on real Roku hardware, because Roku publishes no emulator, but app upload and the test runner are not available yet.
How is this different from a real-device cloud?
A device cloud gives you physical phones; VibeView gives you simulators and emulators. Simulators start in seconds, cost less to run and can be handed to anyone with a link — but they cannot reproduce bugs that depend on physical hardware, such as camera capture, Bluetooth peripherals, NFC, cellular conditions or GPU performance. Most teams use simulators for day-to-day work and physical devices for a pass before shipping.
Can I put a live device on my own site?
Yes. The JavaScript SDK embeds a running device in an iframe, so a prospect, reader or reviewer can use the real app inside your landing page, documentation or pitch deck with nothing to install. Sessions can also be shared as a plain link if you do not want to embed anything. With the per-key Agent Control setting turned on (it is off by default), your own team can additionally drive an embedded session from the command line, an MCP client or the API while the visitor watches — the setting changes nothing about what a visitor can do.
What kind of build do I upload?
For iOS, a simulator build — a .app bundle, uploaded as-is or compressed to .zip, .tar.gz or .tgz. An .ipa is a device build and will not install on a simulator, so it is rejected with a message explaining why. For Android, upload the .apk your existing build already produces. The upload limit is 500 MB.