Google Analytics MCP server
Ask Claude or ChatGPT about your Google Analytics in plain English - traffic, channels, conversions, funnels. Free, and set up in about two minutes.
Continue with Google · You choose which accounts to connect · Disconnect any time · No card needed
- Free to connect
- Sign in with Google - no Cloud project
- Listed in Claude's connector directory
MCP (Model Context Protocol) is the open standard AI assistants use to call outside tools. A GA4 MCP server exposes the Google Analytics Data and Admin APIs as tools, so when you ask “which channels drove the most conversions last month?” the AI runs the real report against your property instead of guessing or asking you for a CSV.
Set it up in 2 minutes
Sign in with Google
Use the account that can see your Google Analytics 4 data.
Approve access
Google's own consent screen. You never handle a token or key file.
Add it to your AI
In Claude, open Caybl.ai in the connector directory. Anywhere else, add the MCP server URL https://api.caybl.ai/mcp.
Ask
"Which channels drove the most conversions in the last 28 days, and how does that compare to the 28 days before?"
Full walkthrough with screenshots: Step-by-step: connect GA4 to Claude.
What the Google Analytics 4 MCP server can do
Reading is always on. Tools marked write are on by default and can be switched off for the whole connection or a single account. Claude can dry-run a change first, so the platform checks it without applying anything.
| Tool | What it does |
|---|---|
run_report | Any dimensions × metrics over a date range, with filters, sorting and up to 4 date ranges |
run_realtime_report | Who is on the site right now, and what they're doing |
run_funnel_report | Step-by-step drop-off between events, e.g. pricing view → sign-up |
get_account_summaries | Every account and property the connection can see |
get_property_details | Time zone, currency, industry and other property settings |
get_custom_dimensions_and_metrics | The custom fields set up on a property, so reports use the right names |
list_google_ads_links | Which Google Ads accounts are linked to the property |
mark_key_eventwrite | Turn an event into a key event (conversion) |
create_google_ads_linkwrite | Link a Google Ads account to the property |
delete_google_ads_linkwrite | Remove a Google Ads link |
What you can ask
- Channels“Which channels drove the most conversions in the last 28 days, and how does that compare to the 28 days before?”
- Pages“Which landing pages have the most sessions but the lowest engagement rate?”
- Funnel“What's the drop-off between viewing pricing and signing up, by device?”
- Realtime“How many people are on the site right now, and which pages are they on?”
- Audit“Audit this GA4 property: missing key events, (not set) traffic, self-referrals and anything else that looks broken.”
- AI traffic“How much traffic came from ChatGPT, Claude, Perplexity and Gemini this month, and did it convert?”
Your options for a Google Analytics 4 MCP server
| Setup | Access | Hosting | Best for | |
|---|---|---|---|---|
| Google's official GA4 MCP server | Google Cloud project, enable APIs, Application Default Credentials, install with pipx/Python | Read-only | Runs on your own machine, next to your MCP client | A developer who already lives in Google Cloud and wants a free, local, read-only server |
| Caybl.ai (hosted) | Sign in with Google, add one URL to Claude or ChatGPT | Read and write, with writes switchable off per account; changes can be dry-run first | Already running (EU region) | Marketers and teams who want it working today, in Claude or ChatGPT on the web, for more than one person |
| Data-pipeline tools with an MCP add-on | Create an account, build a data source or flow, then add their connector | Usually read-only, from a copy of your data | Hosted | Teams already paying for that tool for dashboards or warehouse syncs |
If you're a single developer who already uses Google Cloud and Claude Desktop or Claude Code locally, Google's own server is a good, free choice. Caybl.ai exists for everyone else: no Cloud project, works in claude.ai and ChatGPT in the browser, and each teammate just signs in with their own Google account.
Related
- Caybl.ai vs. self-hosting Google's GA4 MCP server
- Connect GA4 to ChatGPT
- Google Analytics in Claude Code
- Free Claude skills for GA4 audits and reports
Last updated 22 October 2026.
Connect Google Analytics 4 to Claude
Sign in with Google, add one URL, ask in plain English. No exports, no Cloud project.
Get started free →FAQ
A GA4 MCP server is a small service that gives an AI like Claude or ChatGPT tools to read your Google Analytics 4 data. “GA4 MCP” and “Google Analytics MCP” mean the same thing: the AI calls the Google Analytics Data and Admin APIs through it, so answers come from your real property. Caybl.ai hosts one, so there is nothing to install.
Yes. Google publishes an open-source GA4 MCP server (googleanalytics/google-analytics-mcp on GitHub). It runs locally, is read-only, and needs a Google Cloud project with the Analytics APIs enabled and Application Default Credentials set up on your machine.
Yes. Connecting Google Analytics and asking questions about it in Claude, ChatGPT or another AI is free, with no time limit and no card. Pro adds Google Ads, Search Console and Bing in your own AI.
Yes. It's a standard remote MCP server at https://api.caybl.ai/mcp with OAuth sign-in, so any client that supports remote MCP servers can use it: claude.ai, Claude Desktop, Claude Code, ChatGPT (developer mode), Cursor and others.
No. Each question fetches the report it needs from Google at that moment and returns it to your AI. Caybl.ai keeps your connection, not your data.
The MCP server is built on the same official Google Analytics Data API. The difference is who writes the query: with MCP, the AI picks the dimensions, metrics and filters from your question, so you don't need to know the API.