Amami MCP docs
Auto insights
Auto insights turn analytics from a passive dashboard into a review loop. Instead of manually scanning every chart, ask Amami to look for changes worth attention.
Insight types
Trends:
- Traffic growth or decline.
- Search, social, direct, or referral source changes.
- Page categories gaining or losing attention.
- Mobile or desktop behavior shifts.
- Country or region changes.
- Conversion trend changes.
Anomalies:
- Sudden traffic drop.
- Sudden traffic spike.
- Conversion event stops firing.
- One source or campaign becomes abnormal.
- A key page receives traffic but no downstream events.
- A country, browser, or device segment changes sharply.
Opportunities:
- A page with high traffic and weak CTA performance.
- A source with lower traffic but better conversion.
- A doc page that should link to signup or pricing.
- A campaign that deserves a follow-up post.
- A missing event that blocks better analysis.
Prompt for an insight scan
Scan the last 30 days for trends, anomalies, and opportunities. Use Amami data, explain the evidence, and give me a prioritized action list.
Prompt for anomaly detection
Detect unusual changes in the last 7 days. For each anomaly, tell me the metric, expected baseline, actual value, likely cause, and next action.
Prompt for opportunity discovery
Find growth opportunities in my analytics. Look for pages with strong traffic, sources with high conversion, and missing events that make future analysis harder.
How to evaluate an insight
A useful insight should include:
- What changed.
- Where it changed.
- When it changed.
- How large the change was.
- Which segment or source explains it.
- What action to take.
- What to measure after the action.
If the assistant cannot produce evidence, ask it to call narrower tools or recommend instrumentation.
Avoid false confidence
AI analysis is only as good as the data available. Treat an insight as directional when:
- The date range is too short.
- Event tracking is incomplete.
- Campaign URLs are missing UTM parameters.
- A deployment changed the page structure.
- The site has very low traffic.