Free for B2B and SaaS teams
Google Analytics MCP for Claude and ChatGPT
Ask GA4 where traffic comes from, where your funnel leaks and which channels drive demos and signups. Answers in minutes.
A free, hosted GA4 MCP server built for B2B and B2B SaaS teams. Connect in about 5 minutes and ask in Claude, ChatGPT, Cursor or our own web chat. No explorations, no Google Cloud project, no code.
Want to look first? See the setup options or browse the prompts.
Where do visitors drop off before signing up?
I traced your signup funnel over the last 30 days.
74% leave between the homepage and pricing
- Homepage visit47,832100% of users
- Pricing page12,45674% drop-off
- Start trial3,24174% drop-off
- Complete signup1,84743% drop-off
Fix this first: add a "View pricing" CTA to the homepage hero, then test social proof on the pricing page.
Where did our traffic come from, and what converts?
47,832 sessions, up 12% on the previous period.
LinkedIn converts best at 5.2%
| Source / medium | Sessions | Conv. rate |
|---|---|---|
| google / organic | 18,421 | 3.2% |
| google / cpc | 12,845 | 4.8% |
| direct / (none) | 8,234 | 2.1% |
| linkedin / paid | 4,521 | 5.2% |
Worth testing: LinkedIn has the lowest volume but the best conversion rate. Consider shifting budget toward it.
What drove our traffic spike last Tuesday?
Tuesday had 8,432 sessions against a 2,430 daily average.
+247% traffic, mostly from one blog post
Watch out: conversion rate fell from 3.2% to 1.1%. The spike brought low-intent visitors, so capture emails for remarketing.
Example answers from a B2B SaaS GA4 property.
Quick answer
What is Google Analytics MCP?
Google Analytics MCP (also called GA4 MCP) is a connector that lets an AI assistant like Claude read your Google Analytics 4 property. It uses the Model Context Protocol (MCP), the open standard Anthropic introduced in 2024, so instead of building explorations you ask in plain English and Claude analyses your traffic, funnels, engagement and conversions, explains what changed and recommends what to fix.
GrowthSpree's Google Analytics MCP is the managed, free version built for B2B and B2B SaaS teams. You connect with Google sign-in, then ask in Claude, ChatGPT, Cursor or our web chat, and start with 20 ready-made GA4 prompts.
Last updated September 21, 2026
- Price
- Free to start
- Setup time
- About 5 minutes
- Works with
- Claude Desktop (free and paid plans), Claude Cowork, Claude Code, ChatGPT, Cursor and any MCP client, plus GrowthSpree web chat and Slack on Premium
- Built for
- B2B and B2B SaaS marketing and revenue teams
- Access
- Google OAuth 2.0, revocable any time
- Google Cloud project
- Not needed
- GA4 access needed
- Viewer
- Changes to GA4
- Claude recommends; you make the change
- Multiple properties
- Yes; set the property per conversation
- Also connects
- Google Ads, LinkedIn Ads, Meta Ads, Search Console, HubSpot, Microsoft Advertising
- Data use
- Covered by our privacy policy
- B2B ad spend managed
- $60M+ B2B ad spend managed From GrowthSpree client audits
- accounts optimised
- 300+ accounts optimised From GrowthSpree client audits
- rating on G2
- 4.9/5 rating on G2 Read the verified reviews
Used by marketing teams at
Built for B2B
Built for B2B and B2B SaaS, not ecommerce
Most AI marketing connectors are built for online stores chasing ROAS. B2B buys differently, whether you sell software, services or industrial products: long cycles, buying committees, demo and quote requests, and pipeline. GrowthSpree's MCP is built around that.
-
Pipeline, not clicks
Connect HubSpot alongside Google Analytics and ask which campaigns, keywords and pages created SQLs and deals, not just leads.
-
The B2B channel mix
Google Ads, LinkedIn Ads, Microsoft Advertising, GA4, Search Console and HubSpot on one account, with Meta Ads on Premium.
-
Long sales cycles
Prompts look at 90-day windows, assisted conversions and demo or quote requests, so a slow month does not look like a failure.
-
B2B and SaaS benchmarks
Prompts written from $60M+ of managed B2B ad spend and our LinkedIn Ads and Google Ads waste reports.
Benchmarks
B2B benchmarks your GA4 data should be judged against
B2B analytics fails when it is measured like ecommerce. These benchmarks come from GrowthSpree's B2B SaaS research and $11.3M Google Ads Waste Report, and the prompts below check your property against them.
- 84 to 281
- 84 to 281days: typical B2B SaaS sales cycles, depending on deal size
- 5 to 15%
- 5 to 15%of B2B SaaS revenue captured by a default 7-day click window
- 36.1%
- 36.1%of B2B SaaS Google Ads spend goes to traffic that never reaches SQL
- $50 to $2,000
- $50 to $2,000suggested conversion values from trial to SQL for B2B value-based bidding
Benchmarks from GrowthSpree's published waste reports. Find the benchmark prompts in the prompt library below.
Why GrowthSpree
What makes GrowthSpree's MCP different from a generic connector?
A connector gives an AI tool access to your data. That part is quickly becoming a commodity. What decides the quality of the answer is what the AI knows about B2B, and that is what GrowthSpree adds.
-
Benchmarks no one else has
Your account is compared with 99 audited B2B SaaS accounts from our Google Ads and LinkedIn Ads waste reports, not with generic industry averages.
-
Playbooks from real accounts
Prompts and playbooks encode how our operators run B2B accounts, from $60M+ of managed ad spend across 300+ companies.
-
Pipeline on one key
LinkedIn Ads, Google Ads, Microsoft Ads, GA4, Search Console and HubSpot on one account, so answers are judged on SQLs and revenue.
-
Built into your week
Company context, saved prompts, scheduled audits and a Slack bot turn one good answer into a weekly habit.
- PriceLabs 3.6x ROAS (0.7 to 2.5)
- Trackxi 4x trials, 51% lower cost
- Rocketlane 36% lower cost per demo, 15% MoM inbound growth
Results from GrowthSpree agency engagements, the work these prompts and playbooks come from.
Before and after
What changes when AI can read your GA4 property?
The GA4 work that eats a marketing team's week, and what it looks like when you can just ask.
| The job | Today | With Google Analytics MCP |
|---|---|---|
| Answering a simple question | "Where did traffic come from last month?" turns into 30 minutes of clicking through GA4 reports and exports. | Ask it in plain English and get the channel breakdown, the trend and what to do about it in seconds. |
| Finding the funnel leak | Build a funnel exploration, guess which steps matter, and still not know why people leave. | Claude traces every step to your key event, finds the biggest drop and recommends specific fixes. |
| Explaining a spike or drop | Compare date ranges across channels, pages and countries until something looks different. | Claude breaks the change down by source, page and country and ranks the causes by impact. |
| Trusting the data | Key events, conversions, "(not set)": nobody is sure what is tracked correctly. | Ask for a tracking audit and get every key event checked for drops, duplicates and gaps. |
| Improving pages | Rewrite landing pages on instinct and wait weeks to learn if it worked. | Claude reads how visitors behave on the page, then writes new headlines and CTAs aimed at the problem it found. |
| Reporting to leadership | Executives ask, you scramble to build a custom report, and by the time it is ready they have moved on. | A weekly or board-ready website report with the reasons behind every change, written in seconds. |
Setup
How do you connect Google Analytics 4 to Claude or ChatGPT?
Three steps, about five minutes. No code and no Google Cloud project. Use it in the AI assistant your team already has.
-
Get your login and connect Google Analytics
Enter your work email and your login details arrive by email. Log in, connect Google Analytics with one click and copy the token shown on your dashboard.
-
Add it to your AI tool
Pick your setup below: the Claude Desktop extension on a free plan, a custom connector on paid Claude or Cowork, one command in Claude Code, or a config for any MCP client.
-
Ask your first question
Start with "Where did our traffic come from last month?" and get an answer from your live data.
- Claude free plan
Claude Desktop extension
Install the GrowthSpree .MCPB file in Settings, Extensions, then paste your token when Claude asks. About 2 minutes.
- Claude paid plans
Claude Desktop and Cowork connector
In Customize, Connectors, add a custom connector named GrowthSpree and paste your personal server URL.
- Developers
Claude Code
Run one
claude mcp addcommand with your API key in the header. Done. - Everything else
Any MCP client
Paste the SSE config into Cursor, Windsurf or any client that supports remote MCP servers. ChatGPT connects as a custom connector.
What you need before you start
- A GA4 property you can view (Viewer access or higher)
- An AI tool you already use: Claude Desktop (the free plan works), Claude Cowork, Claude Code, Cursor or another MCP client. No AI tool? On Premium, the GrowthSpree web chat works on its own.
- A work email to receive your login details and setup steps
Calculator
What is fixing your funnel worth?
Most funnel fixes Claude finds are small: a missing CTA, a slow page, a confusing step. Here is what a small lift in conversion rate adds up to on your traffic.
A relative lift, so 10% takes a 2.0% conversion rate to 2.2%. Pick a number you believe; Claude will show you where the real leak is.
- Extra demos or signups each month
- 80
- Over a year
- 960
- Time to find the leak
- One question
Prompt library
The best GA4 prompts for Claude and ChatGPT, ready to run
Written by our growth team for B2B and SaaS properties. Each one tells Claude to use only real data from your property, state the date range and flag gaps. Save any of them as a custom prompt in GrowthSpree and rerun it in one click.
-
"Where did our traffic come from last month, and which sources convert?"
Channel and source/medium breakdown with conversion rate and trend.
-
"What caused the traffic spike (or drop) on a specific day?"
Root cause of any spike or dip, ranked by impact.
-
"Which countries and devices bring engaged traffic?"
Geo and device quality, not just volume.
-
"Where do visitors drop off between landing and signup?"
Step-by-step funnel with the biggest leak called out.
-
"Which pages have high traffic but low conversion?"
Pages that waste attention, with fixes.
-
"Are my key events tracked correctly?"
A tracking audit before you trust any report.
-
"Which blog posts actually lead to demos or signups?"
Content that drives pipeline, not just pageviews.
-
"Rewrite the headline and CTA on my weakest landing page."
Copy grounded in how visitors actually behave.
-
"Write this week's website report for leadership."
Week-over-week summary with reasons and next steps.
-
"How do my channels compare under data-driven attribution?"
Which channels assist, which close.
-
"Is our attribution window too short for B2B?"
Checks your lookback windows against B2B sales cycles.
-
"How long from first visit to demo request, by channel?"
Time to convert by source.
-
"Are our key events valued like a B2B funnel?"
Conversion value tiers check.
-
"How much of our paid traffic never engages?"
Paid traffic quality against the 36.1% waste benchmark.
-
"Which paid landing pages have the worst engagement?"
Where ad spend lands badly.
-
"Where does mobile break our funnel?"
Device-level drop-off.
-
"Which blog posts lead readers to pricing or demo pages?"
Content that moves buyers forward.
-
"What changed after our site redesign?"
Before and after comparison.
-
"Clean up our (not set) and unassigned traffic."
A tracking hygiene report.
-
"Draft a GA4 tracking plan for a B2B site."
Events, parameters and key events.
Playbooks
Playbooks: how our operators use Google Analytics MCP every week
Prompts answer one question. Playbooks chain them into a routine. Save each step as a custom prompt and rerun it in one click; Google Ads and Meta audits can also run on a schedule.
-
The funnel fix sprint
Find and fix the biggest leak in two weeks.
- Run the funnel prompt from landing to your key event.
- Check mobile versus desktop drop-off.
- Ask Claude to rewrite the headline and CTA on the leakiest page.
- Compare the two weeks before and after the change.
-
The tracking trust check
Make sure every report after this one is right.
- Audit key events for drops and duplicates.
- Clean up (not set) and unassigned traffic.
- Check the attribution window against your sales cycle.
-
The monthly website report
A board-ready view of the site.
- Run the weekly report prompt for the full month.
- Add channel attribution beyond last click.
- Ask for a one-page summary with three actions.
Questions to ask
What can you ask Claude or ChatGPT about Google Analytics?
Anything you would ask a senior analytics lead. These are the questions B2B and SaaS teams ask most in their first week.
- Which channels drive quote and demo requests?
- Where did our traffic come from last month?
- Why did organic traffic drop last week?
- Which landing pages convert best from paid traffic?
- Where do visitors drop off before booking a demo?
- Which blog posts lead to signups?
- Is LinkedIn traffic worth what we pay for it?
- Which countries bring engaged visitors, and which bounce?
- Are all our key events firing correctly?
- How does mobile conversion compare with desktop?
- What changed on the pricing page after the redesign?
- Which channels assist conversions without getting credit?
- Rewrite the headline on our lowest-converting landing page.
- What should we track that we are not tracking today?
- Compare our marketing site with our product app property.
- Write this month's website report for the board.
- What content should we create next quarter based on engagement?
Example answers
What does a Google Analytics MCP answer look like?
Two more example answers from a B2B SaaS GA4 property.
Asked: "Which channels are we undervaluing?"
Attribution, beyond last click
Channel Conversions Assist rate Organic search 342 32% Paid search 287 18% Organic social 156 45% Email 134 52% Insight: email assists 52% of conversions but rarely gets last-click credit. Don't cut it based on last-click reports.
Asked: "Which pages keep people engaged, and which lose them?"
Engagement and exits
Page Avg. time Conv. /case-studies 3m 21s 15% /blog/complete-guide 4m 12s 12% /features 3m 45s 8% Insight: case studies convert best at 15%, while /pricing has a 68% exit rate. Feature case studies on pricing.
Who uses it
Who on a B2B revenue team uses it, and what they ask first
- CMO or VP Marketing
"Which channels bring visitors who become pipeline?"
- Growth and web teams
"Where do visitors drop off before a demo?"
- RevOps
"Do GA4 key events match what sales counts?"
- Content teams
"Which posts lead readers to pricing?"
- Agencies
"Which client site lost the most conversions this month?"
- Founders
"Is the website working?"
Compare
Which Google Analytics MCP should you use? Official vs GrowthSpree vs the rest
Google publishes its own open-source Google Analytics MCP server. It is a strong choice for developers. Here is how the options compare for a marketing team.
| Best for B2B and SaaS GrowthSpree Google Analytics MCP | Google's official server | DIY from GitHub | Other hosted GA4 MCPs (e.g. Data Bloo, Stape, Windsor) | |
|---|---|---|---|---|
| Price | Free to start | Free | Free, plus your time | Free tiers to paid plans |
| Setup time | About 5 minutes | 30+ minutes | Hours | Minutes |
| Google Cloud project and API setup | Not needed | Required | Required | Usually not needed |
| Coding | None | Python and terminal | Python or Node and config files | None |
| Changes to GA4 | Analysis; you make the changes | Read-only | Depends on the repo | Usually read-only |
| Ready-made analysis prompts | 20, built for B2B and SaaS | None | None | Generic |
| Ad platforms in the same chat | Google Ads, LinkedIn Ads, Meta Ads | No | Rare | Some vendors |
| CRM and pipeline context | Yes, with the HubSpot MCP | No | No | Rare |
| Where you can use it | Claude Desktop (free and paid), Cowork, Claude Code, ChatGPT, Cursor, any MCP client, plus web chat and Slack on Premium | MCP clients you configure yourself | Depends on your build | Usually Claude and ChatGPT |
| Built for B2B and B2B SaaS | Yes: LinkedIn Ads, Microsoft Ads, HubSpot pipeline and B2B prompts | No, general purpose | No | General purpose |
| Scheduled reports to email or Slack | Slack bot answers on Premium; scheduled audits for Google Ads and Meta | No | If you build it | Some vendors |
Official server details from Google's Analytics developer documentation and the googleanalytics/google-analytics-mcp repository. Competitor details come from each vendor's public pages in September 2026 and may change.
Capabilities
What can Claude or ChatGPT do with Google Analytics MCP?
Almost anything a senior analytics lead, CRO specialist and report writer would do with full access to your GA4 data. Here is what that covers, and the few things it deliberately will not do.
Your AI analyst can
- Break down traffic by channel, source, campaign, page, country and device
- Find where your funnel leaks, step by step, and why
- Explain any spike or drop, with causes ranked by impact
- Audit key events and flag tracking gaps before they skew reports
- Show which content and landing pages lead to demos and signups
- Rewrite landing page headlines and CTAs from how visitors behave
- Draft a tracking plan and suggest new key events
- Write weekly, monthly and board-level website reports
- Combine GA4 with Google Ads, LinkedIn Ads and HubSpot for full-funnel answers
- Answer questions in Slack through the GrowthSpree bot (Premium)
- Use your company context (ICP, goals, brand voice) so every answer fits your business
- Save any prompt as a custom prompt and rerun it in one click
- Share a conversation with a colleague through a private link
It won't
- Make changes to GA4 on its own. It recommends; you decide what to apply.
- Guess numbers. Prompts tell Claude to use real property data and flag gaps.
Testimonials
What B2B and SaaS marketing leaders say
-
Hours saved every week, and leaks found that nobody knew about
"This tool has saved us hours each week. The real-time recommendations are spot-on, and we've plugged budget leaks we didn't even know existed."
Arnav MahajanChief of Staff, Fleetpanda -
Reporting went from a chore to a question
"GrowthSpree's AI platform has made reporting so easy. These custom reports are intuitive, and they've made decision-making so much faster."
PranayHead of Marketing, Hubilo -
Like a dedicated analyst, without the overhead
"From budget leaks to performance issues, it identified and fixed problems we didn't even know existed. It's like having a dedicated analyst, but without the overhead."
DebpriyaDirector of Marketing, Sogolytics
Security
Is Google Analytics MCP safe for your GA4 property?
-
Official APIs only
Data comes from Google's official APIs through the access you approve on Google's own consent screen. No scraping and no shared logins.
-
Tokens stay server-side
OAuth tokens are stored server-side and the AI never sees your password.
-
Your own API key
Every teammate gets a personal API key. Treat it like a password and ask us to reset it if it is ever exposed.
-
Disconnect any time
Disconnect any integration from the Integrations page, or revoke access from your Google account. How data is handled is set out in our privacy policy.
Integrations
Go beyond sessions: connect the rest of your funnel
GA4 tells you what happens on your site. Add your ad platforms and CRM and Claude can tell you which spend and which pages created pipeline.
-
Google Ads MCP
Match paid clicks to on-site behaviour and find wasted spend.
Learn More -
HubSpot CRM MCP
See which sources and pages turned into SQLs and deals.
Learn More -
Search Console MCP
Join search queries to GA4 engagement by page.
Learn More -
LinkedIn Ads MCP
Check whether LinkedIn visitors engage and convert.
Learn More -
Meta Ads MCP
Follow Facebook and Instagram traffic through your site.
Learn More -
All marketing MCPs
Plus Microsoft Advertising, Shopify and Slack, all on one GrowthSpree account.
Learn More
FAQ
Google Analytics MCP questions, answered
What is Google Analytics MCP (GA4 MCP)?
Google Analytics MCP is a connector built on the Model Context Protocol (MCP), the open standard Anthropic introduced in 2024, that lets an AI assistant like Claude read your GA4 property. Instead of building explorations or exporting reports, you ask questions in plain English and Claude answers from your live traffic, engagement, funnel and conversion data.
Is there an official Google Analytics MCP server?
Yes. Google publishes an open-source Google Analytics MCP server on GitHub (googleanalytics/google-analytics-mcp). It is aimed at developers: you install it with Python, create a Google Cloud project, enable the Analytics APIs and set up credentials. GrowthSpree's Google Analytics MCP is a managed alternative for marketers that connects with Google sign-in and works alongside Google Ads, LinkedIn Ads and HubSpot.
Is GrowthSpree's Google Analytics MCP free?
You can start free. The Basic plan covers Google Ads, Google Analytics, Search Console, LinkedIn Ads, Microsoft Advertising and HubSpot, used inside the AI tool you already have. Premium adds Meta Ads, the GrowthSpree web chat and Slack.
Do I need a Google Cloud project or any code?
No. You connect Google Analytics with Google sign-in and GrowthSpree generates your personal API key. There is no Google Cloud project to create, no APIs to enable and no Python to install.
Which setup should I use?
On Claude's free plan, use the Claude Desktop extension. On a paid Claude plan or Cowork, add GrowthSpree as a custom connector. Developers can connect Claude Code with one command. For Cursor, Windsurf, ChatGPT and other MCP clients, paste the server config with your API key. Every option uses the same token from your dashboard.
How long does setup take?
About 5 minutes: enter your work email to get your login, connect Google Analytics and copy your token, add GrowthSpree to your AI tool (Claude Desktop extension, Claude connector, one Claude Code command or an MCP config), then ask your first question.
Can Claude change my GA4 settings?
The connection is built for reading reports and property details so Claude can analyse them. When Claude recommends a tracking or configuration change, you make it yourself in GA4 or Google Tag Manager.
What GA4 access level do I need?
Viewer access to a GA4 property is enough to connect it and run reports.
Can I connect multiple GA4 properties?
Yes. You choose the GA4 property when you connect, or use a manager-level connection, and then set which property each conversation uses. That keeps multi-brand and agency work separate, for example your marketing site versus your product app.
Can I combine GA4 with Google Ads, LinkedIn Ads and HubSpot?
Yes. Connect GrowthSpree's Google Ads, LinkedIn Ads, Meta Ads, Search Console and HubSpot MCPs as well, and Claude can follow a visitor from ad click to site behaviour to pipeline in one conversation.
Can Claude write copy and recommendations from GA4 data?
Yes. Because Claude can see which pages hold attention, where visitors drop off and which channels convert, it can rewrite landing page headlines and CTAs, suggest content to create, draft a tracking plan and write stakeholder reports grounded in your own data.
Does it replace GA4 explorations or Looker Studio?
It replaces most of the one-off questions you would build an exploration for. Dashboards are still useful for always-on monitoring; Claude is faster for why questions, root causes and ad hoc analysis.
How accurate are Claude's answers?
Claude pulls numbers directly from the Google Analytics Data API, so it follows the same data, thresholds and sampling rules as your GA4 reports. Ask it to show the date range and metrics behind any answer and you can check them in GA4.
How is Google Analytics MCP different from the GA4 API?
The GA4 Data API is the interface developers write code against. An MCP server wraps that API as tools an AI assistant can call on its own, so Claude decides which reports to run, pulls the data and explains the results without you writing code.
Does Google Analytics MCP work with ChatGPT?
Yes. GrowthSpree's MCP server is hosted, so it works with any client that accepts a remote MCP server: add it to ChatGPT as a custom connector, or paste the config into Cursor, Windsurf and other MCP clients. Claude has the most setup options: the Desktop extension on the free plan, a custom connector on paid plans and Cowork, and one command in Claude Code.
Is my Google Analytics data used to train AI models?
GrowthSpree uses your data to answer your questions and run the reports you set up. How it is stored and processed is set out in our privacy policy and your agreement with GrowthSpree. The AI assistant you choose, such as Claude or ChatGPT, also applies its own data policy.
How is GrowthSpree's Google Analytics MCP different from a generic MCP connector?
A generic connector only gives an AI tool access to your data, which is quickly becoming a commodity. GrowthSpree adds what decides the quality of the answer: benchmarks from 99 audited B2B SaaS accounts, prompts and playbooks from $60M+ of managed B2B ad spend, every B2B channel plus HubSpot on one account, and workflow features such as company context, saved prompts, scheduled audits and a Slack bot.
Where do the benchmarks come from?
The benchmarks come from GrowthSpree's published B2B SaaS research, including the $11.3M Google Ads Waste Report across 43 accounts, and from GrowthSpree's work across 300+ B2B companies. Each benchmark on the page links to its source.
Why is GrowthSpree's Google Analytics MCP built for B2B and B2B SaaS?
Because B2B companies, SaaS or not, are measured on pipeline, not online sales. GrowthSpree connects Google Analytics with LinkedIn Ads, Microsoft Advertising and HubSpot on one account, and its prompts look at demo and quote requests, SQLs, assisted conversions and 90-day windows that match B2B sales cycles.
Can the AI learn about my company?
Yes. Set your company context once, including what you sell, your ICP, goals and brand voice, and it is applied to every conversation so recommendations and copy fit your business.
Can my whole team use it?
Yes. Each teammate gets their own login and API key, and you can share any conversation through a private link. On Premium, the Slack bot answers questions for the whole team in your channels.
Do I need a paid Claude plan?
No. On Claude's free plan you install the GrowthSpree extension (.MCPB file) in Claude Desktop and paste your token. On paid Claude plans and Claude Cowork you can add it as a custom connector instead, which also gives you higher usage limits for large properties and long reports.
Who is Google Analytics MCP for?
B2B and B2B SaaS marketing leaders, growth and web teams, agencies reporting across many client properties, and founders who want GA4 answers without learning GA4. That includes software companies and B2B firms in IT services, professional services and manufacturing.
Still have a question? The MCP help center covers setup, integrations, plans and troubleshooting.
Get free access