How to Connect Google Analytics 4 to Claude with MCP
Connect Google Analytics 4 to Claude with the Data Bloo MCP and pull sessions, users, key events and revenue straight from the chat, no exports, no report builder. It's read-only, so it's safe, and here's the 2-minute setup.
Your Data Bloo MCP URL
Paste this into Claude when you add a custom connector. Then expand Advanced settings and enter 43 in the OAuth Client ID field, leave the Client Secret blank. That's the whole secret; everything below is just where to click.
What connecting GA4 to Claude actually means
You can ask Claude about your Google Analytics data and get a real answer, pulled from your actual GA4 property, in seconds. Type "how did our traffic do last month" and Claude reads your numbers and tells you, no CSV export, no exploration to build. The connection runs through the Data Bloo MCP, takes about two minutes to set up, and it's read-only, so Claude can analyze your traffic but can never touch your property settings.
Comfortable installing Python packages and wiring up Google Cloud credentials? Google's official Analytics MCP is free and worth a look. Want the two-minute browser setup, plain-English answers, GA4 blended with Search Console, Google Ads and your store revenue, and the same data feeding your Data Studio dashboards? That's Data Bloo, and that's what the rest of this guide covers.
Stop digging through GA4 reports. Start asking.
The same GA4 property you already trust for your Data Studio reports, now answering questions in a chat. No explorations to build, no comparisons to configure.
Ask, get instant answers
"Which landing pages drove the most engaged sessions last week?" Claude pulls the numbers and explains them.
Reports inside the chat
Ask for a 30-day traffic view and get tables plus a summary ready to share, no exploration setup, no exports.
Client-ready summaries
"Turn the last 30 days of GA4 into an executive summary." Paste it into Slack, email or a doc.
Diagnose dips fast
"Why did organic traffic drop on Tuesday?" Claude looks across channels, landing pages and dates and explains it.
Blend beyond GA4
Line GA4 up against Search Console, Google Ads, Facebook Ads or Shopify revenue in one answer.
Connect once, use everywhere
The same GA4 property you linked for Data Studio is instantly available to Claude, zero duplicate setup.
Google Analytics 4, plus every other connector
Already connected in Data Bloo? It's instantly available to Claude, connect once, use everywhere.
Three ways to connect Google Analytics to Claude
There are three real options, built for different people. Here's the honest version, including the case where the answer isn't us, so you pick the right one the first time.
A connector like Data Bloo
Paste one URL, sign in once, ask in plain English. No Python, no Cloud credentials, nothing to install or host. Read-only, blends GA4 with your other channels, and the same connection feeds your Data Studio reports.
Google's official Analytics MCP
Google publishes an official Google Analytics MCP server at github.com/googleanalytics/google-analytics-mcp. Free, open source and read-only like ours. The trade-off: it's a developer tool, you install it locally with Python and pipx, authenticate through the gcloud CLI, and it only sees GA4.
A manual CSV export
Export a report from GA4 or an Exploration, save the file, upload it to Claude. Free, and it works once. Then you do it again next week, separately for every property.
Data Bloo vs Google's official MCP vs a manual export
The bottom line: if you're a developer who lives in the terminal and wants raw Data API access, Google's own server is free and very good. If you'd rather skip the credentials setup entirely, ask in plain English, and see GA4 next to the rest of your marketing, that's Data Bloo.
When to use Google's official Analytics MCP instead
Data Bloo isn't the right tool for everyone, and Google's own server is genuinely good. The Analytics team publishes it at github.com/googleanalytics/google-analytics-mcp, currently labelled experimental. It talks to the GA4 Admin and Data APIs and exposes tools like run_report, run_realtime_report, run_funnel_report and get_custom_dimensions_and_metrics, all under the read-only Analytics scope, so it can't change your property either. If you want real-time reports, funnel explorations or every custom dimension you've ever defined, first-party coverage is hard to beat.
The trade-offs are real, and they're all about setup. It runs locally on your machine: you need Python 3.10+, pipx to install it, the gcloud CLI, and Application Default Credentials configured with OAuth or a service account, then every teammate repeats that setup on their own laptop. It also only sees Google Analytics, so it can't line your traffic up against Search Console queries, Google Ads spend or Shopify revenue, and it won't feed the Data Studio dashboards you already send clients.
So it's not a safety question, both servers are read-only and neither can touch your property. It's a question of who you are. If you're an engineer comfortable managing Google Cloud credentials, take Google's. If you're a marketer who wants an answer before the standup, take the one that's already hosted. Plenty of teams run both.
What you need
Three things, and you probably already have them.
How to connect Google Analytics 4 to Claude
Pick your Claude client. Desktop & Web are easiest for most marketers; Claude Code is a one-liner for developers.
Launch Claude Desktop or claude.ai. In the chat input, click the + icon on the left to open the tools menu.
Hover over Connectors, then click Manage connectors to open your connector settings.
Click the small + at the top of the connector list, then choose Add custom connector.
Type Data Bloo in the Name field. That's how you'll find it inside any conversation later.
In the Remote MCP server URL field, paste https://mcp.databloo.com/mcp. Then expand Advanced settings and enter 43 in the OAuth Client ID field, leave the Client Secret blank.
Claude opens your browser on Data Bloo's sign-in page, sign in and approve so Claude can read your data. You sign in once.
Start a new chat and try: "Using Data Bloo, pull sessions, total users and key events for my Google Analytics 4 property for the last 30 days." Claude returns a table plus a short analysis.
Open your terminal and run:
claude mcp add --transport http databloo https://mcp.databloo.com/mcp
Next time you start a Claude Code session, Claude opens your browser to sign in to Data Bloo. Sign in once, the connection persists across sessions.
Run /mcp, you should see databloo with status connected.
Confirm it works: in a fresh chat, ask "Using Data Bloo, list all the Google Analytics 4 properties I have connected." If Claude returns your properties, you're fully connected.
Take the MCP for a spin
Walk through connecting Google Analytics 4 to Claude and pulling a report, no signup, right here in the page.
Seven read-only tools, why your property stays untouched
You just ask in plain language, Claude picks the right tool. The typical flow: list_data_sources → get_connector_accounts → list_fields → query_data.
get_userlist_data_sourcesget_connector_accountslist_fieldsquery_datalist_accountsget_datesRead-only, always. Claude can pull and analyze your GA4 data, it can never edit a data stream, delete an audience or change a property setting. Your Analytics configuration stays exactly as you left it.
Read-only safety and your dashboards, in one connection
The developer route makes you choose: spend an afternoon on Python and Cloud credentials, or keep exporting by hand. Data Bloo doesn't ask you to choose. It reads, it reports, and the exact same connection already powering your Data Studio dashboards now answers questions in Claude.
It can't touch your property
All seven tools are read-only. Claude pulls and analyzes your traffic, but it can't edit an event, delete an audience or change a data stream, so there's no wrong instruction that could break the tracking your whole team relies on. No safety checklist to memorize, because there's nothing to switch off.
One connection, two surfaces
The GA4 source you built for Data Studio is the same one Claude reads. No second setup, no second bill, and the numbers Claude quotes always match the dashboards you send clients. It's the one thing a local, GA4-only server structurally can't give you.
What GA4 data you can pull
The same metrics you use in your reports, sliced by channel group, source/medium, landing page, country, device and date.
Sessions, total users, new users and engaged sessions, the core of every traffic report.
Engagement rate and average engagement time, the quality signals GA4 replaced bounce rate with.
Key events (what GA4 now calls conversions) and revenue, so purchase and lead questions come back with real value attached.
Default channel group, source/medium, landing page, page path, country, device and date.
What Claude actually gives you back
Three real prompts and the kind of answer Claude returns once Google Analytics 4 is connected through Data Bloo, numbers, then a plain-language read.
Claude: Last 30 days on Google Analytics 4, 148,392 sessions from 96,418 total users, with 3,847 key events. That's a 61.4% engagement rate, and sessions are up 9% on the prior period.
Claude: Top three, /blog/ga4-reporting-guide with 18,204 engaged sessions, /pricing with 9,412, and the homepage with 8,730. /pricing converts at nearly twice the site average.
Claude: Organic Search, 61,240 sessions and €38,410 revenue. Paid Search, 22,876 sessions and €21,930. Direct, 19,552 sessions and €9,884. Paid Search earns the most revenue per session at €0.96.
Example outputs shown for illustration; your figures come live from your own connected property.
Not just charts. Finished, branded deliverables
Plug your live GA4 data into Claude, ChatGPT or Gemini through our MCP, and generate branded presentations, audits and executive reports straight from a chat prompt.
Client presentations
Slide-ready decks with your logo, colors and commentary, built from live data, ready to send.
Audits
Automated property audits that flag issues, wins and opportunities, with clear next steps.
Executive reports
Concise, narrative summaries for leadership, the numbers that matter, explained in plain language.
Prompts to paste into Claude
Once connected, drop any of these into a fresh chat. Name Data Bloo in your first message, then drill in.
Built for how you actually work
Same connection, different job to be done. Here's the prompt each type of marketer tends to start with.
Agencies
Pull and compare across every client property without switching logins, then draft the monthly report in the chat.
E-commerce & DTC
Track revenue and key events by channel, spot the sources that bring visitors but no buyers, and line GA4 up against Shopify.
SEO & content teams
Watch landing pages and organic engagement, pair GA4 with Search Console, and catch a decaying post before rankings slide.
Common issues and fixes
You're likely not signed in yet, or Google Analytics 4 isn't connected in Data Bloo. Ask "Using Data Bloo, list the accounts I have connected." If it comes back empty, open Data Bloo, confirm your GA4 source is connected, then try again.
Small gaps are normal. GA4 keeps processing recent data for up to 48 hours, your property's timezone shifts which day a session lands in, and when Google signals thresholding kicks in, the interface can hide rows the API aggregates differently. For trends, comparisons and reporting the Data Bloo pull is solid; for today's traffic, give GA4 time to finish processing.
That's GA4, not the connection. "(not set)" appears when a session has no value for the dimension you asked for, and "Unassigned" is traffic that doesn't match any channel rule. Ask Claude to exclude those rows, or to explain how big a share they are before you worry about them.
That's on purpose. The connection is read-only, so nothing in your GA4 setup, events, audiences, data streams, can be changed from a chat. If you need to edit your property, do it in Google Analytics itself.
Query notes & FAQs
Add a custom connector in Claude, name it Data Bloo, and paste the MCP URL https://mcp.databloo.com/mcp. Sign in once, then ask Claude about your Google Analytics 4 data in plain language. No code, no exports, about two minutes end to end.
It's free to start, no credit card. The free tier includes MCP access with usage limits; paid plans raise the limits and add more connected accounts.
No. All seven MCP tools are strictly read-only. Claude can pull and analyze your data but can never edit events, audiences, data streams or any property setting. Your tracking setup stays exactly as you left it.
Yes. Google publishes an experimental Google Analytics MCP server on GitHub. It's free and read-only, but it's a developer tool: you install it locally with Python and pipx, authenticate with the gcloud CLI and Application Default Credentials, and it only sees GA4. Data Bloo is the hosted, no-code route that also blends GA4 with your other channels.
No. Universal Analytics stopped processing data in 2023 and Google has since removed access to it. The connector works with Google Analytics 4 properties, which is where all your current data lives.
Any MCP-compatible client, Claude Desktop, Web and Code, plus ChatGPT, Gemini and Cursor. The same URL works across all of them.
Yes. As long as they're connected in Data Bloo, ask Claude to compare GA4 against Search Console, Google Ads, Facebook Ads or your Shopify revenue, and it pulls one blended answer. That's the piece a GA4-only server can't do.
Claude queries your data live through Data Bloo's connectors the moment you ask. Keep in mind GA4 itself finalizes recent data over 24 to 48 hours, so the freshest numbers can still shift slightly.
No. A free Claude account is enough to add the Data Bloo connector and start pulling GA4 reports. Your Claude plan is separate from Data Bloo, and the connector works on Claude Desktop, Web and Code.
Yes. You sign in once through Data Bloo's secure OAuth flow, and the connection is strictly read-only. Claude can pull and analyze your GA4 data but can never edit an event, delete an audience or change any property setting.
Yes. Ask Claude to list your connected GA4 properties, and it can pull and compare across them, handy for agencies running multiple clients.
Yes. Pair the MCP with Claude's scheduled tasks or Claude Code for an automatic Monday-morning traffic digest, anomaly alerts or a recurring client update, unattended.
Scale your reporting without limits
Every plan includes all data sources. Start free, upgrade only when your dashboards demand it.
Best for individuals or early-stage teams starting with marketing data.
Create free accountFor individuals who need unlimited refreshes with a single account.
Choose StarterFor freelancers & growing teams that need more accounts and volume.
Choose ProFor agencies managing multi-client reporting at scale.
Choose AgencyNo credit card required to start · Cancel anytime · See plan FAQs →
Chat with your Google Analytics 4 data today
Connect Google Analytics 4 to Claude in under 2 minutes and pull reports straight from the chat. Free to start, no credit card.