Bootstrapped by founders Olli and Patrik, ScreenCI is a new tool that treats tutorial videos as end-to-end tests. The concept is deceptively simple: you describe a feature in a prompt, and your coding agent records a narrated walkthrough of your real app. If the application changes and the video can no longer be recorded successfully, the CI run fails just like a broken Playwright test.

The Agent-Driven Workflow

Unlike traditional screen recording tools that require manual editing, ScreenCI integrates directly into the development lifecycle. The source material emphasizes that the script is a test, meaning every take is identical and deterministic. When the product evolves, a single click re-records the video. This ensures that documentation and marketing assets never drift from the actual user experience, solving the common problem of outdated help center content.

Advanced Visual Automation

The platform goes beyond simple capture, offering sophisticated camera controls driven by code. Features like autoZoom allow the camera to track clicks, field entries, and slider drags automatically, eliminating the need for manual zoom points. The system supports true redaction of sensitive data, masking API keys and PII in the browser before the frame is captured, ensuring secrets never enter the uploaded recording. Additionally, it supports programmatic overlays using React, Vue, or plain HTML, allowing developers to embed branded callouts and rings anchored to live UI elements.

Global and Multi-Format Output

ScreenCI supports narration in 79 languages, allowing a single recording to be localized for different markets with synchronized subtitles and AI voices. The tool also handles aspect ratio switching seamlessly; a single setting change can render vertical 9:16 videos for TikTok or Reels without cropping. With features like motion blur for natural-looking cursor movements and speed ramps to skip repetitive steps, the output looks professionally filmed rather than scripted.

Key Takeaways

  • Tutorial videos are treated as CI artifacts that fail if the UI changes.
  • Coding agents generate the walkthroughs from simple text prompts.
  • Supports 79 languages with per-cue narration styles and voice cloning.
  • Includes true data redaction and advanced camera auto-zooming.

The Bottom Line

This is the kind of infrastructure that lets AI agents actually do useful work beyond just writing code. By merging documentation with testing, ScreenCI forces teams to keep their product knowledge up to date, automatically.

Source

Hacker News