Amami MCP docs

Realtime monitoring

Realtime analytics helps during launches, campaigns, incidents, and content spikes. The dashboard shows what is happening now; MCP lets your assistant explain the live signal and suggest what to do.

What to watch

  • Active visitors.
  • Current top pages.
  • Live referrers and campaign sources.
  • Countries, devices, browsers, and operating systems.
  • Conversion events and goal events.
  • Recent event stream.
  • Sudden traffic increases or drops.

Launch workflow

Before launch:

  • Confirm tracking is installed.
  • Confirm the signup or install event fires.
  • Add UTM parameters to launch links.
  • Open the realtime dashboard.
  • Ask your assistant to watch source, page, and event changes.

During launch:

Monitor realtime traffic for this site. Tell me the top sources, active pages, and whether signup or install events are keeping up with the traffic spike.

After launch:

Summarize the first two hours of launch traffic. Which source produced the highest-quality visitors, and what should I change before the next post?

Incident workflow

Realtime data can expose tracking or site problems quickly:

  • Traffic drops to zero after a deployment.
  • One region disappears from traffic.
  • A conversion event stops firing.
  • A launch source sends traffic but no users reach signup.
  • Mobile visitors spike and bounce immediately.

Ask:

Check realtime data for signs of a tracking or conversion issue after the latest deployment.

Campaign workflow

Use realtime monitoring when paid, social, newsletter, or community campaigns go live:

  • Compare campaign source against normal baseline.
  • Check whether visitors land on the intended page.
  • Verify important events fire.
  • Compare mobile and desktop behavior.
  • Watch for irrelevant traffic from unexpected sources.

AI anomaly checks

Amami should describe anomalies with evidence:

  • The metric that moved.
  • The expected baseline.
  • The actual value.
  • The source, page, country, device, or event related to the change.
  • The recommended action.

Good prompt:

Look for realtime anomalies in the last 60 minutes. Separate good anomalies from risky anomalies and give me the next action for each.