Amami MCP docs

Amami docs

Amami is an AI-native analytics product built by modifying and extending Umami-style website analytics. It keeps the familiar analytics foundation: websites, pageviews, visitors, events, referrers, devices, countries, sessions, reports, API keys, teams, sharing, and privacy-minded tracking. The difference is the AI workflow around it: your assistant can help connect a site, inspect the data, explain what changed, and recommend the next growth action.

With Amami MCP, Codex, Claude, Cursor, Windsurf, Cline, Continue, Zed, VS Code, OpenCode, and custom agents can connect to your analytics account, create or inspect tracked websites, read traffic data, and turn that data into growth recommendations without making you hunt through a dashboard.

Product layers

Amami has three layers:

  • Analytics foundation: the Umami-style dashboard, tracking script, website management, events, referrers, UTM campaigns, reports, teams, and sharing.
  • MCP access layer: local server setup, browser authorization, API-key handling, and client configuration for agent tools.
  • AI growth layer: agent workflows that explain changes, compare segments, recommend page or event improvements, and help users decide what to do next.

What this documentation covers

Use these docs to move from first setup to repeatable growth analysis:

  • Start: learn the product model, run quick setup, and choose Cloud or self-hosted deployment.
  • Setup: connect Codex, Cursor, Claude, Windsurf, Cline, Continue, Zed, VS Code, OpenCode, or another MCP client.
  • Product: install tracking, define events, use goals, watch realtime traffic, read reports, and collaborate with teams.
  • Workflows: ask natural-language questions, generate AI insights, produce reports, and run recurring growth playbooks.
  • Reference: configure MCP, understand the API surface, migrate from other analytics tools, and review the security model.

What Amami inherits from Umami-style analytics

  • Lightweight website tracking.
  • Pageviews, visitors, sessions, bounce rate, and visit duration.
  • Referrers, countries, browsers, operating systems, and devices.
  • Custom events and conversion actions.
  • UTM campaign analysis.
  • Realtime active visitor counts.
  • Reports such as funnel, journey, attribution, retention, revenue, and web-vitals when the data and API surface are enabled.
  • Team and sharing workflows for collaboration.

What Amami adds

  • Browser-based login or registration during MCP setup.
  • Local API key creation through an explicit authorization screen.
  • MCP tools that let agents discover sites, create tracking, read stats, inspect trends, analyze traffic sources, and verify events.
  • AI analysis workflows for explaining why traffic changed and what to improve next.
  • A security model where login, registration, and authorization stay in the user's browser.
  • Read access by default, with write tools enabled only when setup uses --write or AMAMI_ENABLE_WRITE=1.

For a new user:

For a developer:

For a growth or marketing team:

What you can build

  • Ask your AI assistant to connect a website to analytics.
  • Let an agent create a tracked website and return the script snippet for your app.
  • Ask what changed in traffic, referrers, pages, events, countries, devices, or campaigns.
  • Get recommendations such as which CTA to move, which page to rewrite, which source to double down on, and which signup event to track.
  • Keep the dashboard available for deep inspection while using AI for daily diagnosis and next actions.

Short setup prompt

Paste this into an AI assistant when you want guided setup:

Install the Amami MCP server, then guide me through browser login and authorization.
https://dashboard.amami.dev/install/mcp-install.md