Google Analytics
MCP server.

Your data. Better questions.

Connect in three steps

1. Add a remote MCP server. Use an AI client with remote HTTP MCP and OAuth support. Client availability and plan requirements vary.
2. Paste your endpoint. Use https://mcp.liftaven.com/ga. Start the connection from your AI client, rather than opening the endpoint as a normal web page.
3. Authorise with Google. Review the requested access, finish consent and return to your client. Ask it to list your accessible properties or sites first.

Your first conversation

After consent, check that your client lists the server’s tools. Start by listing sites or properties, choose the correct one, then ask a question with an explicit date range. Ask the agent to include the source, dates and any missing data in its answer.

If the connection does not work

A 401 response when opening the endpoint in a browser is expected: connect from your MCP client so it can start OAuth. If your property is missing, check that you authorised the Google account with access to it. If consent expires, reconnect through your client. If the client does not support remote MCP with OAuth, use a compatible client rather than pasting access tokens.

You do not need to create your own OAuth app or paste a provider token. The client receives a scoped MCP connection. You can revoke access when you no longer need it.

Set up the GA4 MCP server in your client

Every client below supports remote HTTP MCP with OAuth. Add the server, start the connection, finish Google consent in the browser and return to the client. The server name is yours to choose; liftaven-ga is used in the examples.

Claude (web and desktop)

Open Settings, choose Connectors, then Add custom connector. Paste https://mcp.liftaven.com/ga as the remote MCP server URL and save. Click Connect on the new connector to start Google sign-in. Availability depends on your Claude plan and organisation settings.

Claude Code

Run the command below in your terminal, then type /mcp inside Claude Code to authenticate.

claude mcp add --transport http liftaven-ga https://mcp.liftaven.com/ga

Cursor

Add the entry below to .cursor/mcp.json in your project, or to the global MCP settings. Cursor opens the browser for OAuth the first time the server is used.

{
  "mcpServers": {
    "liftaven-ga": {
      "url": "https://mcp.liftaven.com/ga"
    }
  }
}

VS Code (GitHub Copilot agent mode)

Create .vscode/mcp.json with the entry below, then start the server from the MCP view.

{
  "servers": {
    "liftaven-ga": {
      "type": "http",
      "url": "https://mcp.liftaven.com/ga"
    }
  }
}

ChatGPT and other clients

Use the client’s custom connector or remote MCP option where the plan allows it, and paste the same URL. If a client only supports local stdio servers, a bridge such as mcp-remote can forward to the hosted endpoint; native remote support is preferable when available.

What the GA4 MCP tools return

Each tool returns a bounded, labelled report rather than a raw API dump, so an agent can cite the property, window and definitions it used.

Start with a useful question

Example prompts. Your agent’s answers depend on the data and permissions available.

Find the GA4 property and web stream for my website. Show its timezone and currency.

Compare organic landing pages over the last 28 days. Which have traffic but weak engagement?

Which acquisition channels bring engaged sessions and key events? Explain the date range and metric definitions.

List the key events on this property and show how many each landing page recorded this month.

Hosted GA4 MCP server vs running your own

Open-source Google Analytics MCP servers on GitHub work well for developers who are comfortable creating API credentials and running a local process. The hosted endpoint trades that control for a two-minute setup and nothing to maintain.

Both approaches read the same provider data. Choose self-hosting if you need custom tools or an air-gapped setup; choose the hosted endpoint if you want your team using GA4 data in their AI client this afternoon.

Access and limits

The standalone GA4 MCP has no Liftaven subscription gate. Google API quotas, your property permissions and per-tool request limits still apply. Analytics access is read-only. Missing tracking, thresholding and reporting delays can affect the results.

Your AI client receives the reports it requests. Review its data policies before connecting. Read Liftaven’s privacy and permissions.

Google Analytics MCP server: questions, answered

Is this for Universal Analytics or GA4?

This server uses Google Analytics 4 reporting and administration APIs. It is designed for GA4 properties. Universal Analytics stopped processing data in 2023.

How is this different from running Google’s own analytics MCP server?

Google’s open-source server runs locally and needs a Cloud project, enabled APIs and application-default credentials on your machine. The hosted endpoint uses browser OAuth from your client, so there is nothing to install or rotate.

Can it change tracking or Analytics settings?

No. It reads reports and configuration. It does not change your tracking, create events, edit audiences or write Analytics settings.

Why does GA4 show fewer sessions than Search Console shows clicks?

They measure different things. A click is counted by Google at the search result; a session needs the page to load and the tag to fire. Consent, blockers and bounce timing all create a gap. Compare them side by side rather than expecting equality.

What does “(other)” or thresholded data mean in an answer?

GA4 withholds rows that could identify individuals and groups rare rows under (other). The server preserves those warnings instead of hiding them.

Does it cost anything?

The standalone GA4 endpoint has no Liftaven subscription gate. Your AI client may have its own plan requirements, and Google’s API quotas apply.

Which clients are supported?

Any client with remote HTTP MCP and OAuth support: Claude, Claude Code, Cursor, VS Code and ChatGPT plans that allow custom connectors.

Need all your SEO data together?

The unified Liftaven MCP uses https://mcp.liftaven.com/mcp. Sign into Liftaven and approve your client to reuse the connections in your dashboard. Standalone provider endpoints have their own consent.