Stop describing.
Start capturing.
Record what went wrong: clicks, console errors, failed requests, screenshots, and your voice explaining what you expected. Export a session package and paste its markdown handoff into any coding agent. One person, one browser, nothing to configure.
Start solo, nothing to set up
- 1 Record. Hit Alt/Option + Shift + R and reproduce the bug. Clicks, console errors, failed requests and environment are captured as they happen.
- 2 Talk. Press Alt/Option + Shift + M where it breaks and say what you expected instead. Your voice becomes the bug description.
- 3 Hand it over. Export, then paste the handoff markdown into Claude Code, Cursor, Codex, or any other agent. Repro steps and a Playwright script ship in the same package.
No relay, no account, no second person. See what the export actually looks like →
Then, when you want it: the live loop
Everything above works on your own. When a non-developer is the one finding the bugs, your agent can watch their session as it happens and ask them questions directly.
- Tester Shares a session code, then hits record. No terminal, no API key, no account.
- Agent Joins from the developer’s machine and watches interactions, console errors and failed requests stream in live.
- Agent Asks a question, which appears as a toast in the tester’s browser, mid-session.
- Tester Answers by voice and highlights the spot on screen. The agent keeps working with the answer.
Tester and developer can sit on different machines anywhere in the world, end-to-end encrypted through a relay that only ever sees ciphertext. Or keep the whole loop on localhost.
Developers
Set up the MCP bridge once, then hand your agent a session, or join a tester’s live and let it ask its own questions.
Developer guide →Testers
Record, talk, done. Nothing to configure and nothing to write up: your voice and your clicks are the bug report.
Tester guide →Coding agents
Watch sessions live over MCP, interview the tester, and reproduce with generated Playwright scripts.
Agent integration →Built to be allowed in
Local by default
Recordings, logs and markers stay in your browser. No account, no upload, no cloud project to provision.
End-to-end encrypted
Remote sessions are AES-256-GCM encrypted on your device. The relay only ever sees ciphertext, so we cannot read your session.
Secrets scrubbed at the source
API keys, tokens, passwords, card numbers and SSNs are stripped from logs, network data and form fields before they leave the machine, on exports and live streams alike. It's pattern-based, so treat it as a safety net rather than a guarantee, and note that screenshots, video and voice can't be scrubbed at all.
Yours to host, yours to swap
The hosted relay runs on Cloudflare’s edge network. It forwards ciphertext and stores no session content, but it is global, so we do not claim a data region today. Need to control where it runs? npx tracegist-relay on your own infrastructure and point the extension at it. Agent-neutral by design: any MCP client works, not one vendor’s.
Live streaming and AI analysis are both opt-in and off until you turn them on. Full detail in the privacy policy.
What’s included
- One-file markdown handoff: paste into any coding agent, no integration required
- Voice markers with triage prompts: press Alt/Option + Shift + M to annotate
- Auto-capture of JS errors, failed network requests, and console logs
- Playwright repro script generated from interaction log
- DOM interaction timeline with
data-testid-priority selectors - Environment snapshots: viewport, storage, scroll position, user agent
- MCP bridge for direct agent access to session packages
- Live shadowing: your agent watches the session and asks clarifying questions as it runs
- Remote Live Sessions: tester and developer on different machines, end-to-end encrypted
Rolling this out to a team?
We are working with a small number of teams, especially in Germany, Austria and the Netherlands, where the people finding bugs are employees and screen recording needs sign-off from a works council or a data protection officer. If that review is ahead of you, we would like to hear how it goes, and we will help you through it.
Get in touch →