Google Analytics MCP Server

Google Analytics

Google Analytics tracks website traffic, user behavior, and conversion metrics. Access account/property info, run reports with 200+ dimensions and metrics, analyze real-time data.

Data Model

• Accounts & properties — organizational hierarchy: an account owns one or more GA4 properties.
• Report pulls — arbitrary dimension/metric slices (200+ built-ins) with paging tokens.
• Custom definitions — user-defined dimensions and metrics registered on a property.

Reporting workflow

Historical reports time out after 60 seconds upstream. Stage large questions instead of one export-shaped call:

1. Totals — metrics only, small row limit (works for YTD and other long ranges).
2. Trend — one time dimension such as yearMonth or date, row limit about one bucket per period.
3. Breakdown — one breakdown dimension at a time, moderate row limit, paginate with offset.
4. Long-range detail — split by month or narrow with dimension_filter instead of stacking multiple session dimensions across YTD.

Use the account/property discovery call first, then custom dimension/metric discovery when non-standard fields are involved.

Gotchas

• property_id: Numeric string like 123456789 or properties/123456789; resolve from the account summary in the linked Google account.
• Dimensions/metrics: Names must exist in the GA4 schema; inspect custom definitions before referencing non-standard fields. users is invalid — use activeUsers, newUsers, or totalUsers. Filtering on source with session-scoped metrics (bounceRate, averageSessionDuration, sessions) is incompatible — use screenPageViews with pagePath or split into two reports.
• Timeouts: Heavy dimension × date × row-limit combinations may exceed 60s — decompose and retry a smaller shape instead of repeating the same call.
• Realtime vs historical: Realtime supports a smaller dimension/metric set and reflects the last few minutes only.