TraceBug — Bug Reports for AI bởi Prashant Singh

One capture. Everything a developer — or an AI coding agent — needs to fix the bug.

TraceBug records what actually happened in your browser: a DOM replay, console errors, network requests, screenshots, and your repro steps — and bundles it into a single self-contained .html file that lives on YOUR machine. No account. No upload. No telemetry.

How it works
1. Click the TraceBug icon → Capture Bug (or press Ctrl+Shift+B)
2. Trigger the bug — TraceBug captures the evidence
3. Review the ticket → Export replay (.html)
4. Hand the file to a developer, attach it to an issue — or let an AI agent read it directly via TraceBug's local MCP server (npx tracebug mcp) and fix the bug

What makes it different
- Local-first — nothing ever leaves your machine. The report is a file you control.
- AI-native — every report is machine-readable by Claude Code, Cursor, and any MCP-capable agent; TraceBug even generates a failing Playwright test that goes green when the bug is fixed.
- Complete evidence — screen recording (with mic narration), interactive DOM replay, console, network, environment, annotated screenshots.
- One-click hand-off — GitHub issue, Jira, Linear, Slack, PDF, HAR, zip.

Firefox notes (honest ones)
- Screen recording asks for one extra click ("Share screen") — Firefox requires the gesture.
- Firefox can't capture tab audio; turn on the microphone toggle to narrate instead.
- Site access is a one-time prompt on first capture.

Free. Open source (MIT): https://github.com/prashantsinghmangat/TraceBug-ai
Try it on a page with intentional bugs: https://tracebug.dev/try.html