dbgr
← dbgr

About dbgr

Share a session. Keep your secrets.

dbgr lets you share a Claude Code session as a debuggable artifact — without leaking what was on your screen. Every transcript is sanitized in your browser before it is uploaded, so the raw file never leaves your machine.

What it does

A Claude Code transcript is a rich record of your work — and it can carry things you would never want to publish: credentials, personal details, the layout of your machine. dbgr takes that transcript, strips out the sensitive material, and gives you a clean, shareable view of where the context actually went.

Sanitized before anything is sent

Redaction happens on your device, the moment you drop a file in. Your original transcript is never transmitted — only the sanitized version is, and only after you choose to publish. If you close the tab beforehand, nothing has left your browser.

Defense in depth

We do not trust a single pass to be perfect. After your browser sanitizes a transcript, the server independently re-scans it on arrival and refuses anything that still looks like a live secret. The two layers are built to catch each other’s misses — a credential has to slip past both, on purpose, to ever be stored.

What gets redacted

Sanitization spans the categories most likely to do harm if shared:

  • Secrets & credentials. API keys, tokens, and private keys — the things that grant access if they leak.
  • Personal information. Identifying details such as emails, phone numbers, and other personal data.
  • Machine-specific paths. Home-directory and local paths that reveal who and where you are.
  • Whatever else you choose. Optional toggles let you redact more — file contents and other context — before you publish.

The detection is broad on purpose, erring toward redacting too much rather than too little. We keep the exact rules out of public view so they cannot be reverse-engineered and tested against.

You see it before you share it

Sanitizing is not a black box. Before you publish, dbgr shows you a preview of exactly what was redacted, lets you toggle categories, and only uploads once you are satisfied. Structural details that carry no risk — token counts, timestamps, model names — are kept intact, so the shared session stays genuinely useful.

An honest note on limits

No automated redaction can promise to catch everything, and dbgr does not claim to. It is built to be thorough and conservative, with overlapping checks — but we still recommend a quick look at the preview before you publish anything sensitive. You stay in control of what goes out.

For the specifics of what we collect, how long we keep it, and your rights, see the privacy policy.