Amami MCP docs

AI growth analysis

Amami turns website analytics into an agent workflow. Instead of opening dashboards, choosing date ranges, and copying numbers into chat, your AI assistant can call Amami tools directly and explain what changed.

What the assistant can analyze

Amami MCP gives your assistant access to analytics signals such as:

  • Websites available to your account.
  • Visitors, pageviews, bounces, session time, and conversion events.
  • Pageview trends by minute, hour, day, or month.
  • Top pages and entry paths.
  • Referrers, UTM campaigns, countries, browsers, operating systems, and devices.
  • Live active visitors.
  • Event and session data when available.
  • Reports such as funnel, journey, attribution, retention, revenue, UTM, and web-vitals when the enabled API surface supports them.

These map to the same categories teams already use in Umami-style analytics docs and dashboards. Amami's extra value is that the assistant can combine them into a diagnosis without asking you to manually click through each report.

Growth questions to ask

Start with questions that force the assistant to inspect data and recommend an action:

What drove visitors this week, and which channel should I double down on?
Which pages performed best this month, and what should I copy from them?
Where are users dropping off before signup, and what event should I track next?
Compare traffic from GitHub, docs, and X. Which source brings the highest quality visitors?
Look at the install docs and landing page traffic. Recommend three concrete changes to improve activation.

For the full query guide, read Natural language queries. For proactive review patterns, read Auto insights and AI reports.

Use this workflow after tracking is connected:

  • Discover available websites.
  • Choose the site and date range.
  • Pull stats, pageviews, referrers, top pages, events, and active visitors.
  • Compare the latest period with the previous period.
  • Identify the pages or sources creating the change.
  • Recommend specific edits, tracking events, or distribution actions.
  • Verify the next change after deployment.

AI should not skip instrumentation

When data is missing, the assistant should say what is missing instead of guessing. Useful missing-data recommendations include:

  • Add signup CTA click tracking.
  • Add install command copy tracking.
  • Add MCP config copy tracking.
  • Add website connected tracking.
  • Add first event received tracking.
  • Add pricing CTA and checkout step tracking.

Example growth playbooks

Use these when you want structured analysis:

  • Landing page: compare hero traffic, CTA clicks, signup events, and top referrers.
  • Documentation: compare install pages, copy-button events, and docs-to-dashboard paths.
  • Launch: monitor realtime visitors, top sources, UTM campaigns, and activation events.
  • Funnel: inspect each step from first pageview to signup or install.
  • Retention: compare returning visitors and repeat product usage signals.
  • Performance: connect web-vitals or page-speed signals to conversion changes when available.

Weekly growth review

Use this prompt at the same time every week:

Run a weekly growth review for this website. Compare the last 7 days with the previous 7 days. Explain traffic, sources, pages, events, anomalies, and missing instrumentation. End with three actions and how to measure them.

Example prompt

Analyze my website growth for the last 14 days. Use Amami data to find the biggest traffic change, the top acquisition source, the best page, and one missing event I should track. Then give me three prioritized actions.

Good output should include:

  • The metric that moved.
  • The source or page responsible.
  • The evidence from Amami tool results.
  • A specific next action.
  • A follow-up measurement plan.

How Amami differs from a dashboard

The dashboard remains useful for visual exploration. Amami MCP adds an AI-native layer that can operate where you already build: your editor, terminal, or agent workspace.

That layer is especially useful when you want to know what to do next, not just what number changed.