For asking questions of your GA4 data
Google Analytics MCP
By Google. Official, experimental server in Google’s googleanalytics GitHub organization (Apache-2.0).
The authoritative way to connect GA4 to an assistant, because it is Google’s own and uses the Analytics Data and Admin APIs with a read-only scope. The trade-off is setup: it runs locally with Python and needs a Google Cloud project and credentials, so it won’t work in browser-only assistants.
Disclosure: the publisher of this site has commercial relationships with some of the companies whose products we rank. Read our full disclosure.
- Reads data
- Yes
- Makes changes
- No
- Runs
- Runs on your computer
- Cost
- Free
What it does
- get_account_summaries, get_property_details and list_google_ads_links describe your accounts.
- run_report and run_funnel_report run Data API reports; get_custom_dimensions_and_metrics lists your custom fields.
- run_realtime_report reads realtime data.
Reads
Account and property details, standard reports, funnel reports, realtime reports and custom dimensions.
Makes changes: No
Read-only: credentials use the analytics.readonly scope and the tools only retrieve data.
Marketing tasks it helps with
Answer “where did traffic come from last month?”
Without building a report in the GA4 interface.
Try: “Which channels drove the most conversions last month, compared with the month before?”
Check a funnel
Uses the Data API funnel report.
Who it’s for
A good fit if
- Analysts and marketers comfortable with a terminal.
- Teams that want Google’s own implementation for data governance reasons.
Look elsewhere if
- You want search queries and indexing, not site analytics. See Search Console MCP (mcp-gsc)
- You need a hosted connector that works in claude.ai or ChatGPT on the web.
Requirements and setup
- Python with pipx.
- A Google Cloud project with the Analytics Admin and Data APIs enabled.
- Credentials for a user with access to your GA4 properties.
- Install pipx.
- Enable the Google Analytics Admin API and Data API in a Google Cloud project.
- Create Application Default Credentials with the analytics.readonly scope.
- Add the server to your MCP client configuration as shown in the README.
Step-by-step for each assistant: How to connect an MCP.
Limitations
- Labelled experimental by Google.
- Runs only on your computer; there is no hosted endpoint.
- Setup requires Google Cloud configuration.
Sources and evidence
Documentation reviewed. We read the maker’s current documentation and linked sources. We have not installed it or run a workflow, so this page says nothing about speed, reliability or output quality.
Facts checked against these sources on .