usage guide

From script tag to first triaged bug

Five steps, about five minutes. This assumes you already have an account (sign in here if not).

1. Create a project and copy your keys

In the dashboard, click New project, give it a name, and optionally list the origins the widget is allowed to run on. Two keys are shown once: a publishable key (safe to embed in your site's HTML) and a secret key (server-side use only, for the direct-report API below). Copy both now, they aren't shown again.

2. Drop the widget on your site

Paste this before the closing </body> tag on any page you want covered:

<script src="https://bugmanager.tech/widget/index.js"
        data-key="bm_pub_..."
        async></script>

That's it, no build step, no dependencies. Your users now see a "Report a bug" button. When they use it, the description plus URL, console errors, and viewport are captured automatically and redacted before anything is stored.

3. Or file a test report without a browser

Useful before you have real traffic. Uses your secret key:

curl -X POST https://bugmanager.tech/v1/reports/direct \
  -H "X-API-Key: bm_sec_..." -H "Content-Type: application/json" \
  -d '{"description":"Save button does nothing on the settings page"}'

4. Watch it get triaged

Open the Reports tab and refresh, then the Agent Activity tab to see the decision trail: a fingerprint check, an embedding comparison against existing categories, and if the bug is genuinely new, an AI worker (and, sometimes, an AI manager) deciding whether it belongs in an existing category or needs a new one. Most repeat reports resolve at the fingerprint step, at zero cost.

5. Work the queue

  1. Open a report and click Copy for AI to get a clean markdown block (description, console errors, suggested fix) ready to paste into Cursor or Claude Code.
  2. Once it's fixed, click Mark resolved on the report, it drops out of your active queue. Reopen if it turns out it wasn't actually fixed.
  3. In Settings → Notification rules, add a rule so you hear about spikes or critical-severity bugs by email instead of having to keep the dashboard open.

Every report is scoped to the project it was filed against, and your dashboard only ever shows projects you own. Nothing here requires you to manage an LLM API key or think about prompts.