Skip to main content
    My stack

    Plug Google Analytics into your assistant

    If you're comfortable with a terminal, or someone can set it up for you once, it's worth the effort: you ask your traffic questions in plain words instead of building reports. Otherwise, skip it: it takes Python, a Google Cloud project and credentials to configure.

    In Google Analytics, your assistant can read your data without changing anything. The server runs on your computer.

    A server to install on your computer and connect with Google Cloud credentials: a technical setup.

    Official, published by Google · checked on 16 November 2026

    What changes for you

    Everyday situations, and what to write to your assistant.

    1. Monthly review

      The month's review without opening a report.

      How many visitors did my site get last month, where did they come from, and what changed from the month before?

    2. Live traffic

      Your site's pulse on launch day.

      How many people are on my site right now, and on which pages?

    3. Converting pages

      The pages that bring you clients.

      Which pages bring in the most quote requests this quarter, and through which channel do those visitors arrive?

    From one freelancer to another

    Google Analytics is the gold mine everyone has and nobody has time to dig. This server hands you a pickaxe that talks: “where do my visitors come from?”, “which pages bring in quote requests?”. It only reads your numbers, so nothing to break there. The price is paid at the door: setup happens on your computer, by hand. One user puts it well: too hard to install, but “magical” once it runs.

    What works

    • Viewer role is enough

      The account used only needs the Viewer role on the property. For a client's site, ask them for that role, nothing more.

      Source: developers.google.com
    • Full paths

      In the config file, write paths out in full: many assistants don't understand “~” or terminal shortcuts.

      Source: developers.google.com
    • The right package

      The official server is called “analytics-mcp”. A project with an almost identical name exists, and the Google team has seen users install the wrong one.

      Source: github.com

    Where it falls short

    • A setup for insiders

      Python, a Google Cloud project, APIs to turn on, credentials: users find setup out of reach for most non-developers. Workaround: get help once from someone comfortable with it, following Google's guide step by step.

      From user reports · github.com
    • Error 403 on first try

      With your personal credentials, Google requires a Cloud project with the Analytics APIs turned on; without it, access is denied. Workaround: turn on the Admin and Data APIs in your project and put its ID in the config.

      Source: developers.google.com
    • On Windows, it hangs

      Claude users on Windows saw every question spin until it gave up. Workaround: update the server: one user confirms the latest version fixed it.

      From user reports · github.com
    • Credentials expire

      The access granted to your computer doesn't always last, and missing read rights cause a “scopes” error. Workaround: run the Google sign-in again from the terminal, asking for Analytics read access.

      Source: developers.google.com

    Plug in Google Analytics

    Pick your assistant: the steps follow.

    1. It runs on your computer: keep Google Analytics open while you use it.

    2. Google Analytics does not detail this setup for Claude: follow its documentation, linked in the sources.

    3. Check it by asking:

      How many visitors did my site get last month, and where did they come from?

    Stay in control

    The assistant never has more rights than your account.

    Your safeguards

    Read only

    It only reads: the access requested is read-only on your Analytics data.

    Your properties, no more

    It only sees the properties where your account has at least the Viewer role.

    What it can do

    Google's documentation says the server is read-only; the official repository also describes three optional write tools, off by default. Only Google's list is shown here.

    ActionIdentifier
    Read only9No effect on your data
    List accounts and properties
    Read a property's settings
    List Google Ads links
    List property annotations
    Read custom dimensions and metrics
    Run a report

    Actions as the vendor documents them.

    Our view

    Useful if

    You track your site in Google Analytics and want answers without building a report.

    Less useful if

    You want a one-click setup: this server is installed on your computer, with Python and Google Cloud credentials.

    Going further

    Compare the 19 official MCP servers

    The guide to get started