Google Analytics#

Query GA4 properties using the Google Analytics MCP server via MCPorter. Each workspace supplies its own service account credentials — no shared or global auth.

Prerequisites#

• uvx (from uv) — runs analytics-mcp ephemerally, no install needed
• mcporter — npm i -g mcporter
• Service account JSON with GA read access in the workspace (see Setup)

Per-workspace credentials#

Each workspace stores its own credentials:

text

{workspace}/
  credentials/
    ga-service-account.json   ← Google service account key file
    ga-config.json            ← optional: default project/property

ga-config.json shape (optional):

json

{
  "projectId": "my-gcp-project",
  "defaultProperty": "properties/123456789"
}

If ga-config.json is absent, pass property explicitly in every query.

Workflow#

1. Locate credentials — read {workspace}/credentials/ga-service-account.json (fail clearly if missing).
2. Load config — read {workspace}/credentials/ga-config.json if present; extract projectId and defaultProperty.
3. Run query via the helper script or direct mcporter call:

bash

bash {skill_dir}/scripts/ga.sh <workspace> <tool> [args...]

Or directly:

bash

CREDS="{workspace}/credentials/ga-service-account.json"
PROJECT_ID="$(cat {workspace}/credentials/ga-config.json | python3 -c 'import json,sys; print(json.load(sys.stdin).get("projectId",""))')"

mcporter call \
  --stdio uvx \
  --stdio-arg analytics-mcp \
  --env "GOOGLE_APPLICATION_CREDENTIALS=$CREDS" \
  --env "GOOGLE_CLOUD_PROJECT=$PROJECT_ID" \
  "analytics-mcp.<tool>" \
  [key=value ...]

Available tools#

ToolPurposeget_account_summariesList all GA accounts + properties the SA has access toget_property_detailsDetails for a specific propertyrun_reportStandard GA4 data report (dimensions, metrics, date ranges)run_funnel_reportFunnel analysisrun_realtime_reportRealtime dataget_custom_dimensions_and_metricsCustom dimensions/metrics for a propertylist_google_ads_linksGoogle Ads links for a property

Common queries#

List all accessible properties:

bash

mcporter call --stdio uvx --stdio-arg analytics-mcp \
  --env "GOOGLE_APPLICATION_CREDENTIALS=$CREDS" \
  analytics-mcp.get_account_summaries

Run a report (sessions last 30 days):

bash

mcporter call --stdio uvx --stdio-arg analytics-mcp \
  --env "GOOGLE_APPLICATION_CREDENTIALS=$CREDS" \
  --env "GOOGLE_CLOUD_PROJECT=$PROJECT_ID" \
  analytics-mcp.run_report \
  property=properties/123456789 \
  'dimensions=[{"name":"date"}]' \
  'metrics=[{"name":"sessions"}]' \
  'dateRanges=[{"startDate":"30daysAgo","endDate":"today"}]'

Setup (one-time per workspace)#

See references/setup.md for full instructions. Short version:

1. Create a GCP service account
2. Enable Google Analytics Admin API and Google Analytics Data API
3. Download the JSON key → save to {workspace}/credentials/ga-service-account.json
4. In GA4: Admin → Property Access Management → add the service account email as Viewer
5. Test: bash {skill_dir}/scripts/ga.sh {workspace} get_account_summaries

Error notes#

• PERMISSION_DENIED — SA email not added to GA property access
• credentials not found — check ga-service-account.json path
• API not enabled — enable Admin + Data APIs in GCP console