Skip to main content

Feature

Bug Reports With Full Reproduction Context

Every report includes the session replay, browser metadata, console errors, network logs, and AI-generated reproduction steps. Stop the "can you reproduce?" loop.

Why teams pick this

Full context attached

Replay link, console errors, network calls, browser + OS, viewport size, Git SHA, and feature-flag state — all attached to every ticket automatically.

AI-generated repro steps

Our model reads the session and writes the reproduction recipe in plain English: "Log in as a free-tier user, click upgrade, observe spinner that never resolves."

One-click ticket push

Linear, Jira, GitHub, Notion — pick your tracker, map the project, and ship. The full evidence packet lands in the ticket body.

Auto-tagging by impact

Severity, page, component, and customer tier are inferred from the session. Triage stops being a daily 90-minute meeting.

How it works

  1. 1

    User hits a bug

    They click the in-app feedback widget, or you spot a frustration spike in the dashboard. Either way, the session is already recorded.

  2. 2

    Click "Open issue"

    Choose the tracker (Linear / Jira / GitHub / Notion). Title, severity, and labels are pre-filled from the session metadata.

  3. 3

    AI drafts the body

    Repro steps, observed behaviour, expected behaviour, and full context links — drafted in seconds. Edit if you want; ship if you don't.

  4. 4

    Ticket lands

    The evidence packet is in your tracker before the user has finished their coffee. Two-way sync keeps Relyv in step when the ticket closes.

Under the hood

What gets attached to a ticket

A direct replay link (Relyv-hosted, password-gateable), console logs filtered for errors and warnings, the network HAR file, browser + OS + viewport, the Git SHA at the time of capture (if you wire up release tagging), feature-flag state for that user, and the AI-generated reproduction steps.

How the AI gets repro right

The model reads the DOM mutation stream and the user-input timeline, not the video. That's why it can write "click the second Save button (the one inside the modal)" instead of pixel-coordinate gibberish. Misses are flagged and never fail silently — every report has a confidence score you can sort on.

Triage automation

Pro tier and above can run rules against incoming reports — "if a Stripe API error is in the console, ping #payments and tag @on-call". Rules can fire webhooks, route to specific teams, or auto-merge duplicates by stack-trace fingerprint.

Frequently asked

Which trackers are supported?

Fourteen trackers natively: Linear, Jira (Cloud + Data Center), GitHub Issues, GitLab, Asana, ClickUp, Trello, Notion, Slack, Teams, Discord, HubSpot, Salesforce, and email — plus a generic webhook for everything else. Two-way sync is supported on Linear, Jira, GitHub, and GitLab; HubSpot creates a ticket pre-filled with the AI bug report; Salesforce creates a Case attached to the same data.

How accurate are the AI reproduction steps?

Every generated repro carries a confidence score. High-confidence steps are reproducible by an engineer who has never seen the bug; anything below 0.7 is surfaced for human review instead of shipped as a "ready" ticket. Published accuracy numbers will land once our launch-cohort eval is peer-reviewed.

Can we redact specific fields from bug reports?

Yes. PII is masked by default. You can also block specific selectors (e.g. ".user-avatar img", "#api-key-field") from ever being captured, or strip specific console namespaces.

What happens if the user is anonymous?

Reports work without identification — the session ID is the link. If you call identify() later (e.g. after sign-up), the session retroactively stitches to the user, and any pending tickets get the user attached.

How do CRM integrations (HubSpot / Salesforce) work?

HubSpot creates a Ticket in the pipeline you select, pre-filled with the AI bug report, network failures, environment metadata, and a deep link back to the replay. Salesforce creates a Case (with RecordType picker for orgs that use them), tied to the same payload. Both are first-class providers in gt-relyv-integrations alongside the bug trackers — no third-party Zapier hop.

Can my AI agent file tickets directly?

Yes — Relyv is MCP-native. Claude, Cursor, and Continue can call get_session_intelligence to read the session, then chain into the integrations service to file the ticket. The same flow a human triggers from the dock works for agents.

Ready to try it?

Every report includes the session replay, browser metadata, console errors, network logs, and AI-generated reproduction steps. Stop the "can you reproduce?" loop.