Google Analytics MCP: Google's server, or a hosted one?

Google publishes an official Model Context Protocol server for Google Analytics 4. It is free, open source, and it works. That fact is the first thing anyone researching a google analytics mcp server runs into, and it deserves a straight answer rather than a sales page: for a meaningful number of people, the official server is the right choice and you should stop reading here and go install it.

This guide is about the part that is less well covered — what installing it actually involves, where it stops being convenient, and what a hosted connector does differently. We make one of the two options, so read accordingly; every factual claim below is about steps you can verify yourself in ten minutes.

What running the official server involves

The setup is not hard, but it has more steps than the phrase "just install it" suggests:

1. Create a Google Cloud project, or pick an existing one.
2. Enable the Google Analytics Data API and the Admin API on that project.
3. Create credentials — either OAuth client credentials you then complete a consent flow against, or a service account whose email you add as a reader on each GA4 property.
4. Install the server, usually with pipx, which means having Python and pipx on the machine.
5. Register it in your MCP client's config file, pointing at the local command.
6. Keep that process available whenever you want to ask a question.

None of those steps is unreasonable for someone who works in a terminal. Taken together they are the reason a lot of people bookmark the repository and never finish, and the reason "no Google Cloud project, no JSON key, no terminal" appears in the marketing of every hosted alternative — it is the genuine friction, not an invented one.

Where local stops being convenient

Two consequences follow from the server running on your machine rather than somewhere else, and neither is a criticism of the software.

It runs where you installed it. The connector works in the assistant on that laptop. It does not work in the same assistant on your phone, on a second machine, or in a web client, because there is no process there to talk to. If you only ever work in one place this costs nothing.

It is one platform. The GA4 server reads GA4. Search Console, Meta Ads and anything else are separate servers with separate setups and separate credentials, and an assistant holding three of them still cannot answer "did the ad spend increase move organic sessions" without you having wired all three. For a question that spans platforms, the integration work moves from the vendor to you.

What a hosted connector changes

A hosted MCP server — ours or anyone else's — moves the OAuth registration to the vendor. You approve a consent screen, the connector is a URL, and there is nothing running locally. The trade you are making is explicit and worth stating plainly: you are handing a third party read access to your analytics in exchange for not doing the setup yourself. Whether that trade is good depends entirely on how much you trust the third party and how much your time is worth, and anyone who tells you it is obviously correct is selling something.

What you should check before making it, of any vendor:

• Which scopes do they request? Read-only is a much smaller surface than read-write, and it is visible on the Google consent screen before you accept.
• Are tokens encrypted at rest, and can you revoke access? Revocation should be immediate and in the product, not an email to support.
• What is stored? A connector that answers questions live needs a short cache. A pipeline that copies your data somewhere is a different risk profile with a different answer.

Which one to pick

Use Google's official server if you are comfortable in a terminal, you work from one machine, GA4 is the only source you need, and you would rather trust software you can read than a company you cannot. It is free and it is first-party, which are two very good arguments.

Use a hosted connector if the setup steps above are the reason you have not started, if you want the same connector to work everywhere you use your assistant, or if your real questions span Search Console, analytics and ad spend at once. Mevnix is one of these: one URL, read-only, three sources in the same connector, free for one property. The Search Console guide and the GA4 guide cover what each one answers.

The honest summary is that this is a build-or-buy decision with a low price on both sides, and the deciding question is not features. It is whether the twenty minutes of setup is a pleasant afternoon or the thing that stops the project.