Your real users are your QA team. Retrace finds the bugs they hit, writes the tests, and opens the fix PR.
- Capture live user sessions from the browser SDK or PostHog replay.
- Detect UX failures: console errors, broken network calls, rage clicks, dead clicks, blank renders, error toasts, and session abandonment.
- Cluster repeated failures into replay-backed issues with severity and affected users.
- Auto-generate deterministic UI regression specs so real bugs become repeatable tests.
- Fix by linking each issue to a repo, ranking likely source files, and opening a draft PR with a fix prompt.
- Track issues as new, ongoing, regressed, or resolved, and re-run specs after fixes.
curl -fsSL https://raw.githubusercontent.com/txmed82/retrace/master/install.sh | bashOr with Docker Compose:
curl -fsSL https://raw.githubusercontent.com/txmed82/retrace/master/install.sh | bash -s -- --dockerThen:
retrace quickstart # mints an SDK key and prints a ready-to-paste <script> tag
retrace api serve # start the ingest API
retrace ui # open the local dashboardretrace github connect --repo your-org/your-app --local-path /path/to/repo
retrace qa auto --repo your-org/your-appThis picks the highest-priority open incident, reproduces it with a generated test, scores the repo, and opens a draft fix PR.
| Problem | How Retrace helps |
|---|---|
| Users find bugs first | Session replay + automatic detection |
| Reproducing bugs is slow | Replay-derived regression specs |
| Debugging takes hours | Fix prompts with replay evidence + likely source files |
| Too many tools | One incident queue for replay / UI test / API test / error monitor / PR review |
| Budget constraints | Self-hostable, BYOK for LLM and third-party services |
Browser / PostHog → Ingest API → Replay Processing → Detectors → Clusterer →
Unified Incidents → Auto-Repro → Repo Scoring → Fix Prompt → Draft PR