Fume
Fume Technologies, Inc. · Coding
Fume is an AI QA tool that turns screen recordings into Playwright test suites, runs them twice a day in the cloud, and repairs broken tests on its own.
About Fume
What Is Fume
Fume is an AI testing tool for web apps, built by Fume Technologies after the team went through Y Combinator's Winter 2024 batch. You record a short walkthrough of your app while describing what should happen. Fume watches the video, pulls out test cases, and writes the Playwright code that checks those flows.
The pitch is narrow and worth saying plainly: it doesn't sell test authoring software. It sells the work a QA hire would do, which means designing the tests, writing them, and keeping them alive as the interface changes. Release cycles keep shrinking while regression suites keep rotting, and that gap is the problem Fume targets.
The catch is price and scope. As a software testing platform it bills per team, not per seat, and plans start at $500 per month. There is no free tier, and the output is Playwright, so anything outside a browser is out of reach.
Getting Started
- Sign up at app.fumedev.com and start the trial. No payment details are needed up front.
- Record a screencast of your app with Loom, QuickTime, or any screen recorder. Narrate the steps and the expected results while you click through.
- Upload the video. Fume extracts test cases within seconds and shows the list before it starts writing code.
- Connect Slack and let the bot post results to a shared channel, or pull the suite into CI/CD with an API call.
Product Information
A quick look at Fume's pricing, supported platforms, and performance.
Best for
The users, tasks, and scenarios where this tool fits best.
Users
- QA engineers at small teams
- Engineering leads without a QA hire
- Founders shipping weekly
Tasks
- Building a first regression suite
- Keeping flaky tests alive
- Migrating an existing suite
Scenarios
- Pre-release smoke checks
- Teams with no QA budget line
- Fast-moving products with changing UI
Key features
Video to Test Cases
AI test generation starts from a screen recording, not a spec document. You talk through the app while using it, and Fume turns the narration into named test cases with assertions. The company says teams can cover around 80 percent of their important user flows in under ten minutes of recording. Vague narration produces vague assertions, so the quality of the first draft tracks how clearly you describe expected behavior.
Autonomous Test Generation
The AI QA agent explores the app the way a QA engineer would and writes the suite in Playwright. That includes the page objects, the selectors, and the assertions, not just a click log. Tests land as readable files in your repository, which means a developer can open one and see exactly what it checks. Playwright test automation here looks like hand-written code, not a recorded macro.
AI Fallbacks for Fragile Scripts
Playwright breaks when the DOM shifts. Fume handles that by handing the stuck step to a browser-use agent that finishes the flow in a live browser, then repairs the script in the background. You still get a pass or fail result the same day, and the suite gets sturdier without anyone filing a maintenance ticket.
Free Cloud Test Runners
Test runners are a real line item at scale. Fume Cloud runs the suites at no extra cost, and CI/CD hookup takes a single API call from your pipeline. Teams that already pay for a runner service can retire it, or keep both and compare results during a migration.
Slack Reports and Failure Insights
Results post into a shared Slack channel: how many tests ran, which passed, and what broke. The bot attaches context for failures, such as a 404 page on checkout, so triage starts from a description instead of a stack trace. Runs happen twice a day by default, and settings for CI, notifications, and org rules come later.
Full Playwright Code Ownership
Every test is 100 percent Playwright, and the codebase belongs to you. You can clone it, run it locally, and keep running it after a subscription ends. That matters for teams that have been burned by no-code test builders that lock the suite inside a vendor dashboard.
Pros and cons
Pros
- Turns a recorded walkthrough into a working Playwright suite fast, which compresses weeks of test authoring into an overnight job.
- Self-healing tests cut the maintenance tax that usually kills regression suites within a quarter.
- Free cloud runners remove a separate infrastructure bill and a separate vendor.
- Full code ownership means no lock-in: the suite keeps running if you walk away.
- Import support for Playwright, Selenium, and Cypress tests gives existing suites a path in.
Cons
- $500 per month with no cheaper self-serve tier, so a two-person team pays enterprise-adjacent money for capacity it may not use.
- No free plan at all. The 30 day trial is the only way to evaluate, and a full trial needs a real app to record.
- Playwright only means native mobile apps and desktop clients are outside the scope.
- Quality depends on the recordings you feed it. Skip steps in the video and the first draft has holes that someone has to patch by hand.
Frequently asked questions
Fume watches a screen recording of your app and writes Playwright tests for the flows you demonstrated, then runs those tests on a schedule and repairs them when the UI changes. It replaces the design, authoring, and maintenance work of a QA team for web apps.
Related content
Explore related tools, skills, and articles for Fume.
Fume Alternatives
Base Code
Base44 (Wix) · Coding · ProductivityBase Code is a shared cloud development environment from Base44. Teammates describe changes to an existing GitHub codebase in chat, preview them live and send them back as pull requests.
Google EmbeddingGemma 2
Google DeepMind · Writing · Image · CodingGoogle EmbeddingGemma 2 is a 740M-parameter open model that turns text, code, images, audio, and video frames into vectors in one 768-dimension space, running on phones and laptops.
GitGlow
GitGlow · Coding · ProductivityGitGlow is a local-first desktop Git client that lets independent developers review their own code and their AI agents' changes in one place before shipping.