Connect Google Analytics MCP: prerequisites and authentication troubleshooting

First identify the Google account and GA4 property whose data you will read. The official server uses the Analytics Admin API for accounts and properties and the Data API for reports. A website measurement ID beginning with G- is different from the numeric property ID used in reports.

Prepare a Python environment for the server and a Google Cloud project, and check that both APIs are enabled. For a company account, ask the owner about the existing project and authentication method first; creating a new project is not always necessary.

Use Application Default Credentials (ADC). The scope needed for Analytics read access is https://www.googleapis.com/auth/analytics.readonly. Keep the credentials file on your computer and put its path in the configuration.

Google’s official instructions give authentication examples using an OAuth client file or service-account impersonation. Default gcloud login can require additional Cloud permissions, so check your organization’s allowed client and scopes first. Analytics access and Cloud project permissions are separate.

If pipx is available, register it in a terminal with codex mcp add google-analytics --env GOOGLE_APPLICATION_CREDENTIALS=/path/to/credentials.json -- pipx run analytics-mcp. Replace /path/to/credentials.json with the path to your ADC file.

Codex MCP settings can be stored in your user config.toml or a trusted project’s .codex/config.toml. After registration, check the MCP list and restart the server in the client. A command not found message may mean an executable such as pipx is missing, rather than a server problem.

When the tool list responds, use get_account_summaries to query accounts and properties. Confirm that the property belongs to your website, then request a short-period report. This checks server startup and data access separately.

Keep the first request small: “Query yesterday’s sessions for the target property and show the property ID, date range and time zone.” This is easy to compare with the interface. If properties from several companies appear, do not automatically choose one because its name is similar; check the website and property details.

If “Reauthentication is needed” appears, refresh the saved credentials. For permission errors, check the OAuth scope, API activation and GA4 property access according to the message. Identify the failed step before reinstalling the same server.

If the request succeeds but returns no rows, narrow the dates, property and filters first. If the site-wide report has rows but the /guide filter does not, check the URL condition. Do not assume an empty response means zero visitors; record query conditions separately from an actual zero.

Try this first SEO request: “Compare the latest completed 28 days with the preceding 28 days. Show Organic Search sessions for landing pages beginning with /guide, and separate numerically observed changes from guesses about their causes.”

Check query impressions, clicks and position in Search Console. Connecting Analytics MCP alone does not automatically provide Search Console data. Use GA4 for post-visit behavior and compare a separate source for the queries in search results.