MCP connectorsGoogle Analytics

Google Analytics MCP: connect Google Analytics to Claude and ChatGPT

Ask Claude or ChatGPT which pages drove the most sessions last week, where this month’s traffic came from or how mobile compares with desktop. The answer comes straight from your GA4 properties. No explorations to build, no screenshots of dashboards.

With Deployed OS everyone connects with their own Google account. The AI sees the properties that person can see in Google Analytics, and it only runs reports: nothing in Analytics is created or changed.

3 read actions · read-only · Works in Claude, ChatGPT and Copilot

In practice

What your team can do with Google Analytics in Claude and ChatGPT

Marketing

Traffic before the Monday meeting

Sessions, users and the biggest sources last week against the week before, summarised in the chat in two minutes. No report to build first.
Content

Which pages actually get read

Top pages by views and engagement rate over the last 30 days, so you know what to update and what to write next.
Management

One number, a straight answer

How many active users did the site have last quarter, and from which countries? Ask in plain words and get the figure with the property and dates it covers.
Growth

Where the visitors come from

Sessions by source and medium, day by day, to see whether the campaign moved traffic or not. Custom dimensions too, looked up by name.
Examples

Ask it like this

  • “How many active users and sessions did the webshop have last week compared with the week before?”

  • “Top 10 pages by views in the last 30 days, with engagement rate.”

  • “Which traffic sources brought the most sessions in September?”

  • “Split last month’s users by country and device category.”

  • “Show sessions per day since the autumn campaign started on 1 September.”

  • “Which properties do I have access to in Google Analytics?”

The connector

What the Google Analytics MCP connector does today

  • Readsgoogle_analytics_list_properties

    List the Analytics accounts and GA4 properties you can see, with the property id every report needs.

  • Readsgoogle_analytics_run_report

    Run a GA4 report on one property: up to 10 metrics by up to 9 dimensions over a date range, at most 100 rows, the last 28 days up to yesterday unless you say otherwise.

  • Readsgoogle_analytics_search_metadata

    Look up the exact metric and dimension names a property offers, custom ones included, so a report never rests on a guessed name.

Need it to do more? We extend connectors for the companies that use them.

Security

Who sees what

Reports only
The connector reads GA4 reports: properties, metrics and dimensions over a date range, up to 100 rows per report. It cannot create, change or delete anything in Google Analytics.
Each person’s own rights
Everyone connects their own Google account, so the AI sees the properties that account sees. Someone without access to the webshop property gets nothing from it through the AI either.
You choose who gets it
Admins decide which groups at the company can use the Google Analytics connector at all. Everyone else never sees it.
Every call logged
Who asked, which connector, when. Deployed OS shows reads per connector and per person, and the log can be exported.
Setup

Set up the Google Analytics connector in three steps

01
An admin creates one Google Cloud project for the company, turns on the Analytics Data and Admin APIs and sets up an Internal consent screen. The same Google app serves Drive, Gmail, Calendar and Google Ads, and your company owns it.
02
The admin creates a Web application OAuth client, adds its Client ID and secret to Deployed OS and chooses which groups get Google Analytics.
03
Each person connects their own Google account. From then on they just ask, about any property they can see.
Compared

Google Analytics’s own MCP server or Deployed OS?

Google publishes an experimental Google Analytics MCP server. It runs locally with Python on one computer, uses that person’s own Google credentials and reads reports, real-time and funnel reports included. For one analyst or a developer it is a good fit.

Deployed OS is built for the whole company. One Google app that serves Analytics, Ads, Drive, Gmail and Calendar, each person’s own login, admins choosing which groups get what, and a log of every call. The same connection works in Claude, ChatGPT and Copilot, next to your company’s skills and guardrails.

Their documentation: Google’s Analytics MCP server

Questions

Google Analytics MCP: common questions

Yes. Google publishes an experimental Google Analytics MCP server that you run locally with Python and your own credentials. It suits one analyst or developer. Deployed OS gives the whole company one connection, with each person’s own Google login, access per group and a log of every call.

No. It runs reports and looks up metric and dimension names, nothing more. No property, audience, event or setting is created or changed. If you want it to do more, we can extend it for you.

Yes. Deployed OS is your company’s own MCP, so you connect Google Analytics to Claude, ChatGPT and Copilot in one go. People keep the assistant they prefer.

One for the company, and Google Analytics shares it with Drive, Gmail, Calendar and Google Ads. If you already have the Google app for another connector, the admin turns on the Analytics APIs and adds one redirect URI. The saved keys stay.

Reports cover the last 28 days up to yesterday unless you ask for something else, since today’s figures are still coming in. Ask for today or any range you like. The answer says which property and dates it covers.

It depends on the size of your team. Get in touch and we will walk you through it.

More connectorsAll connectors
Let’s talk

See it on your own tools.

Tell us which systems your company runs on. We’ll show you what your team can ask, and what IT gets to control.

Poyan Karimi
You’ll hear from Poyan Karimi
Co-founder · Deployed AI