Google Analytics MCP Server

Google's free, open-source connector that lets Claude Code or another agent read a client's GA4 reports directly, with read-only access.

Verdict
Worth testing
The job
Reading client analytics
Last checked

It removes the export step from GA4 audits and monthly reports, and it is free. Setup still needs Python, gcloud credentials and two Google APIs, and Google's guide and the GitHub README disagree about write tools, so test it on one client property first.

What it does for a consultant

Google publishes an MCP server for Google Analytics. Once connected, your agent can list the accounts and properties you can see, run standard, realtime, funnel and conversions reports, and read custom dimensions and metrics. Google’s developer guide says it is for read requests only and cannot change a property’s configuration.

How members would use it

  • Ask the client for Viewer access on their GA4 property, the minimum the guide requires.
  • Connect the server to Claude Code and have the agent pull the numbers for an audit or a monthly report, then write the findings in your format.
  • Pair it with a Search Console connector if you need search data, since this server covers GA4 only.

How to start

  • Install Python 3.10 or later and pipx or uv.
  • Enable the Google Analytics Admin and Data APIs in a Google Cloud project.
  • Sign in with Application Default Credentials using the analytics.readonly scope, then add the server to your agent’s settings.

The catch

Setup is not friendly to non-developers. The GitHub README lists three write tools that stay off unless you start the server with --allow-write, so never use that flag on a client property. If you host it for a team, configure OAuth, or you may expose the service account’s credentials.