Amami MCP docs

Self-hosting

Self-hosting gives you infrastructure control. It also gives you operational responsibility. Choose it when you need private deployment, custom retention, internal network access, compliance control, or deeper customization.

Most teams should start with Cloud unless self-hosting is a real requirement.

What you operate

A production Amami-compatible deployment normally needs:

  • Application runtime or container platform.
  • PostgreSQL.
  • HTTPS reverse proxy.
  • Domain and DNS.
  • Backups and restore tests.
  • Logs and monitoring.
  • Upgrade process.
  • Optional AI provider configuration.

Minimum production checklist

  • HTTPS is enabled.
  • APP_SECRET or equivalent app secret is unique and strong.
  • Database credentials are not checked into git.
  • PostgreSQL backups run automatically.
  • Restore has been tested.
  • Admin access is limited.
  • Tracking script URL is stable.
  • API keys are rotated when access changes.
  • MCP clients use environment variables or local env files.

MCP with self-hosting

Point MCP at your instance:

AMAMI_API_URL=https://analytics.example.com
AMAMI_API_KEY=your_api_key_here
AMAMI_ENABLE_WRITE=1

or in a client config:

{
  "mcpServers": {
    "amami": {
      "command": "npx",
      "args": ["-y", "amami-analytics-mcp@latest"],
      "env": {
        "AMAMI_API_URL": "https://analytics.example.com",
        "AMAMI_API_KEY": "your_api_key_here"
      }
    }
  }
}

AI feature tradeoffs

Cloud can provide managed AI workflows. A self-hosted instance may require you to configure:

  • AI provider keys.
  • Model access.
  • Usage limits.
  • Logging and privacy controls.
  • Network egress policy.

If you do not configure AI providers, the core analytics dashboard and tracking can still be useful, but AI reports, natural language analysis, and anomaly workflows may be limited.

When not to self-host

Avoid self-hosting when:

  • You do not have someone responsible for upgrades.
  • You cannot test backups.
  • You need the fastest path to value.
  • You only need simple website analytics.
  • You want managed AI features without infrastructure work.