Bug Reports
One-hotkey, voice-first bug filing — captures transcript, screenshot, console logs, and session replay, then hands the bundle to an agent that creates issues in your team's chosen tracker.
The Bug Report feature lets any user file a high-fidelity report without interrupting their flow. Hold a hotkey, describe the problem aloud, release — the platform transcribes the recording, collects screenshots and console logs, and runs the bundle through an agent your admin configured. The agent creates issues, posts threads, or pages on-call using whatever integrations your org has wired up. No vendor is hardcoded; the agent decides.
Bug reports are off by default for customer organisations. An admin must enable them under Settings → Organisation → Bug Reports and configure the agent before any reporter can file a report.
Enabling bug reports
Only organisation admins and owners can access this page.
Pick the tools the agent can call (for example, a GitHub integration or a Jira integration). Add agent instructions and optionally override the default LLM. You must complete this step before the enable toggle becomes active.
Flip the Enable bug reports toggle. All signed-in org members can now file reports using the keyboard shortcut.
Filing a report
The keyboard shortcut is Ctrl + Shift + B (Windows / Linux) or Cmd + Shift + B (macOS).
| What you do | What happens |
|---|---|
| Press and hold the shortcut | A floating panel appears. The screenshot and session replay buffer are captured immediately. The microphone starts recording. |
| Speak your description | Describe the problem while the panel shows a live waveform. |
| Release the shortcut | Recording stops. The panel shows Sending… while the report uploads. |
| Hold < 0.5 s | Treated as a misfire — nothing is sent. |
| Press Esc while recording | Discards the recording. Nothing is uploaded. |
Once the upload is complete, the panel switches to a Processing view that polls for status in real time. When the agent finishes you see the created issue link (or links) directly in the panel, with a Done button to dismiss.
If the agent fails, the panel shows the failure reason instead.
What gets captured
Every report bundles:
| Artifact | Details |
|---|---|
| Audio transcript | Your voice recording, transcribed by the STT provider your org has configured. |
| Viewport screenshot | A static image of the screen at the moment you pressed the shortcut. |
| Session replay | Up to 30 seconds of DOM activity before the shortcut (rrweb). |
| Console logs | The last 50 browser console lines, collected automatically. Secrets and tokens are redacted on a best-effort basis before the bundle is sent. |
| Recent server errors | Up to 10 server-function errors observed in the current session. |
| Page context | App name, route shape (IDs replaced with placeholders), viewport size, and browser/OS. |
Console log redaction is best-effort. Avoid logging sensitive data to the browser console in production — the filter catches common patterns (bearer tokens, API keys, passwords) but cannot guarantee complete coverage.
Redaction at capture time
Before anything leaves the browser:
- All form inputs are masked (content not captured).
- DOM elements marked
data-privateor.sensitiveare blurred in the screenshot and suppressed in the replay. - Route paths have tenant IDs and UUIDs replaced with shape tokens, preserving the URL structure without leaking the actual identifiers.
What the agent receives
The agent prompt is seeded with your org name and instructs the agent to:
- Include the transcript and any notes verbatim.
- Embed the screenshot inline in the created issue.
- Add a Diagnostics section with console logs, recent server errors, page context, and session replay link.
Admins can customise the instructions under Settings → Organisation → Bug Reports → Agent configuration. The instructions field is a free-form text area. The default prompt is vendor-neutral and works with any issue tracker.
Server logs
If your organisation has an integration that exposes a server-log query tool (for example, a custom Grafana/Loki integration built with the Integrations Authoring SDK) and it is added to the bug-report agent's tool integrations, the agent can fetch correlated server logs as part of triage.
When a report is submitted the agent receives the reporter's user ID, the trace IDs from any recent server-function errors captured in the session, and the capture time window. The default agent instructions direct it to call the log-query tool with that context to retrieve matching log lines, then include the relevant lines in the Diagnostics section of the created issue — alongside the console logs and session replay that are always present.
Server-log triage is agent-driven, not deterministic. The agent decides which log lines are relevant and whether to include them. Correlation quality depends on whether your application logs consistently emit user IDs and trace IDs in the text of each log line (the tool matches them as raw substrings — LogQL line filters — rather than requiring structured label fields).
To enable:
- Install and configure an integration that exposes a server-log query tool.
- Go to Settings → Organisation → Bug Reports → Agent configuration and add that integration to the Tool integrations list.
What the reporter sees after submitting
The panel stays open and shows a live status line while the report is processed:
- Sending — artifacts uploading.
- Processing — agent running.
- Done — the created issue link(s) appear as clickable chips. Press Done to dismiss the panel.
- Failed — the failure reason is shown. The report is saved to history regardless.
Resolving integrations before submitting
If the admin has configured a tool that requires a personal OAuth account (for example, a tracker that identifies issues under the reporter's own user), the panel shows an Integrations section after you stop recording and before you send. For each tool:
- A green check means you have a connected account — nothing to do.
- A picker appears if you have more than one connected account — select which one to use.
- A Connect button appears if no account is connected — clicking it opens Settings → Account → Integrations in a new tab. Return to the panel after connecting; it re-checks automatically.
The Send report button stays disabled until every tool has a resolvable credential.
Configuring the agent
Under Settings → Organisation → Bug Reports → Agent configuration:
| Field | What it does |
|---|---|
| Tools | The integrations the agent may call. Select one or more from your installed integrations. |
| Instructions | Free-form instructions for the agent. Seeded with a default prompt that uses your org name and works with any issue tracker. |
| Model override | Optional. Defaults to the org's default LLM resolved via the integration registry. |
Default instructions
The default prompt is generated from your org name and instructs the agent to triage the report, include the transcript and notes verbatim, embed screenshots inline, and add a Diagnostics section. You can edit it freely.
Rate limits
Admins can configure rate-limit values under Settings → Organisation → Bug Reports → Advanced:
| Limit | Configurable range |
|---|---|
| Per user, per hour | Configurable |
| Per user, per day | Configurable |
| Per org, per day | Configurable |
Rate-limit values are configurable in settings. Runtime enforcement (returning a rejection when the limit is exceeded) is planned and not yet active — reports are not currently blocked when these thresholds are reached.
Artifact retention
The bug-report workflow record (transcript, agent log, issue links, status) is retained for the admin-configured retention window (minimum 7 days, maximum 90 days) and purged by a scheduled job when it expires.
Screenshot images embedded in issues are served via permanent serve URLs — the same mechanism that renders organization avatars — so the inline image keeps loading for the life of the issue and never breaks. The screenshot object is stored under an unguessable, report-scoped key in your organization's storage; read access is public-by-unguessable-key (so the issue tracker's image proxy can fetch it), while the issue or repository itself remains the access-controlled surface. Coupling automatic deletion of the stored screenshot objects to your configured retention window is a planned follow-up.
Screenshots render inline in private-repository issues because the issue tracker's image proxy caches the bytes at render time. The report history page (accessible to org admins and the original reporter) retains the full report record for the configured retention window.
Report history
Settings → Organisation → Bug Reports → History shows all reports filed in your org. Each entry includes the status, transcript, agent log, external issue links, and links to download the artifacts. Only org admins and owners can view the full org history; reporters can view their own reports.
Related
- Integrations — how to install and connect integrations.
- Account integrations — connecting personal OAuth accounts.
- Security → Redaction — the platform-wide redaction model.