All work
Case study · 01

Bug reports that replay themselves

Screen, voice, and every click captured in one shareable link - bugs get fixed instead of bounced back as "can't reproduce."

AIDeveloper toolShipped
01

Most bug reports fail before they reach a developer - not because engineers ignore them, but because they lose enough context in transit that nobody can reproduce the problem. This tool closes that gap. A human does two things: hit record, then hit stop. Everything between is captured automatically - screen, clicks, console, and network. An OpenAI model synthesizes the whole session into a detailed, reproducible bug report, then files the ticket. Claude Code picks it up, applies the fix, validates it, and closes the ticket out. The developer never writes a report. The product manager never has to ask for more detail.

HUMANCAPTUREPROCESSOUTPUT● RECORD■ STOP→ sends to AIScreen captureClick eventsConsole outputNetwork callsSynthesisOpenAIDetailed bug reportTicket createdClaude CodeFix appliedValidated✓ Closed
CAPTURE LAYER - Chrome Extension · MV3getDisplayMedia()DOM event streamDevTools ProtocolService worker bufferANALYSIS LAYER - Serverless · GPT-4o VisionFrame samplerGPT-4o VisionEvent log parserReport composerPM INTEGRATIONLinear · Jira · ticket auto-filedCLAUDE CODE PIPELINEticket → fix → PR → close
QA Recordersystem
Capture
  • Screen recording
  • Click events
  • Console output
  • Network calls
  • Voice narration
Browser layer
  • getDisplayMedia()
  • DOM event stream
  • DevTools Protocol
  • Service-worker buffer
Pipeline
  • Frame sampler
  • GPT-4o Vision
  • Event-log parser
  • Report composer
Delivery
  • Shareable replay link
  • Detailed bug report
  • Hand-off to Claude Code

Want one built for your business?

These are my own builds. Yours would be shaped to your exact workflow - and you'd own it outright.

Email copied - valters@valters.solutions