Fume

Product Introduction

1. Fume is an AI-powered QA testing platform that automatically generates and maintains Playwright browser tests from user-recorded product videos. Users describe their testing requirements through screencasts (using tools like Loom), and Fume converts these into executable test suites that run twice daily on its cloud infrastructure.
2. The core value lies in replacing manual test design, implementation, and maintenance with autonomous AI agents, enabling engineering teams to focus on development while ensuring critical user flows remain validated. It eliminates the need for dedicated QA teams to write or update tests for dynamic web applications.

Main Features

1. Video-to-Test Conversion: Fume analyzes screencasts with voice descriptions to generate Playwright tests covering 80% of critical user flows in under 10 minutes. It interprets spoken assertions (e.g., "Check if the checkout button redirects to payment") and translates them into automated browser interactions.
2. Self-Healing Tests with AI Fallbacks: When Playwright scripts fail due to UI changes, Fume’s AI agent resumes testing by simulating human-like browser navigation. The system automatically repairs broken scripts in the background, reducing flakiness without manual intervention.
3. Cloud Execution with Free Runners: Tests run twice daily on Fume’s infrastructure at no cost, with failure alerts sent to shared channels. Users retain full ownership of Playwright code and can integrate tests into CI/CD pipelines via API for additional execution control.

Problems Solved

1. Manual Test Maintenance Overhead: Engineering teams spend 20-40% of their time fixing broken tests due to UI/UX updates. Fume autonomously adapts tests to application changes, eliminating this maintenance burden.
2. Target User Group: SaaS development teams (particularly startups and mid-market companies) lacking dedicated QA resources, and DevOps engineers seeking to automate regression testing for continuous deployment pipelines.
3. Typical Use Cases: Validating checkout flows after e-commerce platform updates, ensuring authentication workflows remain functional post-UI redesigns, and testing data-heavy dashboards with dynamic content without writing custom selectors.

Unique Advantages

1. Autonomous QA vs. No-Code Tools: Unlike no-code test builders requiring manual test design, Fume acts as a full AI QA team—designing, implementing, and maintaining tests without human input. Competitors like Cypress or Selenium Studio still demand manual selector updates.
2. Native AI Recovery System: Fume uniquely combines deterministic Playwright scripts with LLM-driven fallback agents that resume failed tests mid-flow, a feature absent in traditional testing frameworks. This dual-layer approach reduces failure rates by 63% compared to static scripts.
3. Code Ownership & Portability: Users receive 100% owned Playwright code compatible with local execution, avoiding vendor lock-in. Competitors often encapsulate tests in proprietary formats, while Fume provides exportable TypeScript files that run natively in any Playwright environment.

Frequently Asked Questions (FAQ)

1. How does Fume handle broken tests from UI changes? Fume’s AI agent takes over when Playwright scripts fail, completing the test via direct browser interaction while logging the discrepancy. The system then updates the Playwright codebase to reflect UI changes, ensuring future runs use the repaired script.
2. Can I run Fume-generated tests in my CI/CD pipeline? Yes, all tests are provided as standard Playwright scripts that integrate with GitHub Actions, CircleCI, or other systems. Fume also offers a single API endpoint to trigger test runs within existing pipelines without migrating code.
3. Do I retain access to Playwright code if I cancel my subscription? Users maintain perpetual ownership of all generated Playwright code, which remains executable locally or in self-hosted environments. Subscription cancellation only disables cloud execution and AI maintenance features.