Run GA4 from the chat you already use

What your assistant can do in GA4

Connected, your assistant works in Google Analytics 4 under your own Google sign-in. AdCopilot's Analytics tools number 29 once Analytics is connected, 15 that read and 14 that write, as of v2.45.2. It covers both halves of the product: the numbers, and the setup the numbers depend on.

Reports. It runs five kinds, each over a date range you name:

• Any report. Dimensions by metrics: sessions by channel, sign-ups by landing page, revenue by country.
• Realtime. The last half hour.
• Funnels. An ordered set of steps, and where people drop out of it.
• Retention. People grouped by when they first visited, and how many came back in each later period. A period still running is marked incomplete, not read as a drop.
• Cross-tabs. One set of dimensions down the side and another across, with totals.

It can also list the dimensions and metrics your property has before it asks for one, so a report is built from names that exist.

Setup. It reads your accounts and properties, data streams, key events, custom dimensions and metrics, the links to Google Ads, and the attribution settings. That is the part a report cannot show you: why a number is missing is often a key event nobody marked.

Set up measurement end to end

For a new site, or one whose tracking was never finished, this is the order the work goes in. Your assistant does each step you allow and says plainly which one only you can do.

1. Start an account, if your login has none. Your assistant gives you a link to open once and accept Google's Analytics terms. Only you can accept them, so it cannot do that step for you.
2. Create the property. It takes the time zone and currency from your Google Ads account when one is connected, and asks you otherwise, so the two products agree from the first day.
3. Add a web data stream. That gives the measurement ID your site's tag needs. With Tag Manager connected, your assistant can stage the tag, and you publish it.
4. Mark the events that matter as key events. Key events are what Analytics counts as results, and what Google Ads can later import. Your assistant also lists the ones already marked, and says when one is an event Analytics records on every visit.
5. Add custom dimensions and metrics. One for each parameter you send that a report needs to split by or add up.
6. Link the property to Google Ads. Your assistant checks the existing links before it adds one, so a property already linked to the account is not linked twice.
7. Read the attribution settings. The reporting attribution model and the lookback windows. It changes them only when you ask.

Prompts to try

These only read. Paste them more or less as they are once your assistant is connected.

• “Which landing pages turned paid clicks into sign-ups last month?”
• “Who is on my site right now, and which pages are they on?”
• “Build a funnel from first visit to purchase for the last 28 days and tell me where most people drop out.”
• “Show retention for people who first visited last month, week by week.”
• “Which key events does this property have, and are any of them events Analytics records on every visit?”
• “What attribution model and lookback windows is this property using?”

These change something, so AdCopilot proposes each one first, and your AI client asks before it runs if it is set to ask before AdCopilot's write tools.

• “Mark sign_up as a key event on my main property.”
• “Create a custom dimension for the plan_type parameter.”
• “Add a web data stream for my website.”
• “Link my Analytics property to my Google Ads account.”

What a hosted GA4 MCP server can and cannot change

A hosted GA4 MCP server is a service your assistant calls, so there is nothing to install or keep running. This one changes the definitions Analytics needs in order to measure: key events, custom dimensions and metrics, data streams, property settings, attribution settings, a new property, and the link to Google Ads. It does not touch your site, your tags or your data.

What it cannot do, and will say so when you ask:

• Delete or archive anything. No tool in Analytics removes a property, a stream, a key event or a definition.
• Unlink Google Ads. There is no tool for it, which is why your assistant checks the existing links before it adds one.
• Accept Google's terms or create an account for you. That step is yours, behind a link.
• Reach a property your login cannot open. It runs as you. A property you cannot open in Analytics is out of reach, and where your role cannot edit, it cannot either.
• Make a model change forward-only. Google applies a new reporting attribution model to past and future reports for everyone on the property. A change to a lookback window starts the day it is made. Your assistant says which it is about to do.

How approval works

Nothing changes until you allow it in your AI client — approval is your client's own tool-permission prompt. If you tell your client to always allow a tool, it will run that tool without asking again — which is why the limits that hold either way are server-side: it cannot delete, it cannot publish a container, it will not switch on the campaign it built, and it asks for your budget before it builds.

So for a hard stop, set your AI client to ask before AdCopilot's write tools: Needs approval in Claude; in Cursor, set the run mode to Allowlist and keep them off the allowlist (Cursor's cloud agents never ask). Reads, including every report, run without asking; they change nothing.

Connect it

Add the connector once, in Claude, ChatGPT or any MCP client, and sign in with Google. Analytics is its own consent beside Google Ads, granted in Google products, and it works without the others. You can connect it with read-only access, and your assistant then reads but cannot change anything until you press Enable editing.

The short overview is on the Google Analytics page, and the tool reference is in the docs.

Google's own documentation

What the calls above are built on, and what each setting means, in Google's words:

• Google Analytics Data API overview. The reporting calls behind every report above.
• Google Analytics Admin API overview. The calls behind properties, data streams, key events and custom definitions.
• About key events. What a key event is and how Analytics counts it.
• About custom dimensions and metrics. When to add one, and what it can hold.
• Connect Google Ads to Google Analytics. What the link shares, and how to use it from Google Ads.
• Get started with attribution. The models and windows the attribution settings choose between.