Think inside your AI world.
How do I connect GA4 to Claude?
GA4 connects to Claude through an official read-only MCP server, with community alternatives alongside — and one documented caveat every honest page about it should carry.
Direct answer: add the official GA4 MCP server (read-only) in Claude via Settings → Connectors and authorise your property; Claude Code users run 'claude mcp add'. The read covers sessions, users, events, conversions, engagement, traffic sources, landing pages, device and geography, and e-commerce revenue. One caveat belongs up here with the answer: independent testing by Coupler.io caught the official server returning materially wrong figures in places — wrong periods, silent sampling, in one case a +35% report where reality was −60%. Verify important numbers against the GA4 interface before acting on them.
Useful, with eyes open
For orientation questions — what moved this week, which sources drive signups, how the landing pages rank — the conversational read is a real convenience. For decisions with money attached, treat the connector as a draft of the answer and confirm in the interface. That is not a knock; it is the current documented state of the server, and the frame judges the data it is given without verifying the source's numbers.
The spend rule the analytics never held
Worked case: you ratified a channel rule — paid spend follows a blended-CAC ceiling, and any channel over it for two consecutive months gets cut to maintenance budget; set after a channel ran hot for a quarter on vanity clicks. GA4 attributes and counts; your ceiling, the two-month test and the vanity-quarter lesson are commercial policy, resident with you.
Traffic, read as policy
Through Unl the channel read arrives against the rule: one channel over your CAC line for the second month running — cut to maintenance, by the policy you set after the vanity quarter. The monthly marketing review starts where it should: at your own thresholds, with their history in the room.
GA4 to Claude is an official read-only connection with a documented accuracy caveat worth respecting; GA4 through Unl reads the same traffic against the spend policies you ratified — verdicts from your rules, applied to figures you still sanity-check.
Reads through Unl arrive with measured context — in the presence of the decisions you’ve already settled. The reach lane is live: one box, paste anything. If it speaks MCP, Unl can reach it. Readings arrive unprompted, the data beside the criterion; Unl is a courier, not a warehouse, and keeps only your keys and the frame.
Read further
Questions people ask
What does the official GA4 MCP server read?
Sessions, users, events, conversions, engagement, sources, landing pages, device and geo, and e-commerce revenue — read-only, from the property you authorise. Community servers exist alongside the official one.
How serious is the accuracy caveat?
Serious enough to state plainly: independent testing (Coupler.io) documented materially wrong figures from the official server — wrong periods and silent sampling included. Use the conversational read for orientation; confirm decision-grade numbers in the GA4 interface.
Does the caveat undermine the measured read too?
The measured read inherits whatever the source serves — the frame judges the data it is given; it does not verify the source's accuracy. Your ratified rules apply cleanly to the figures; the figures deserve the same scepticism either way.
What this is
Think inside your AI world — you stay in command
Unlimitless (Unl to friends) holds what you've settled, reads what your tools are showing, and catches what's changed out in the world — and hands your AI whatever bears on the work, the moment it's needed, without you asking. The right thing, in front of the model, unprompted, with you in command of the call. So you keep moving toward what you set out to build, on top of everything you've already decided.
It plugs into Claude, Claude Code, ChatGPT and Cursor as an MCP connector. Quick to connect, in a couple of steps.
Unlimitless is open now to invited Alpha. Apply for the Beta waitlist to come in ahead of the full launch:
Alpha is invite-only · free at launch.