Klaviyo Pointed Its MCP at Claude the Week Anthropic Crossed OpenAI on Ramp
Anthropic now has 34.4% of paying business AI customers on Ramp's AI Index, ahead of OpenAI at 32.3%. Twelve months ago Anthropic sat at 9% while OpenAI lost one point over the same period, closing the gap 26 points across a sample of more than 50,000 companies. Klaviyo announced its expanded Claude MCP integration the same week.
The Ramp number got the TechCrunch headline. The Klaviyo timing is the more useful signal for anyone running a marketing stack right now.
What Ramp is actually counting
Ramp's AI Index doesn't measure raw usage. It counts expense receipts inside Ramp's customer base, which skews tech-forward, biased toward early-stage and growth-stage companies that picked Ramp for spend management in the first place. Ramp economist Ara Kharazian told TechCrunch that Anthropic "has already been in the lead amongst the high adoption groups like finance, tech, professional services" before this month's flip put Claude ahead overall.
So the headline isn't "Anthropic has more enterprise customers than OpenAI." It's something narrower. Anthropic now has more paying logos than OpenAI inside a sample that's biased toward the buyers who pulled MCP, Cowork, and Claude Code into their workflows over the last year. That sample matters because it's roughly the same buyer profile that picks new marketing tools fast.
The cleaner number is the velocity. Anthropic went from 9% to 34.4% of paying Ramp accounts in twelve months. OpenAI held flat. From what I've seen, that kind of velocity curve always reaches martech roadmaps before it reaches anyone's quarterly board deck. Klaviyo confirmed that this week, and they probably weren't the first to run the math.
The Klaviyo move is the part you should actually read
On May 8, Klaviyo shipped an expanded MCP integration that lets Claude (and Claude Cowork) hit raw Klaviyo metrics, build campaign briefs, audit flows, and propose re-engagement work as agentic tasks. Andrew Bialecki, Klaviyo's co-CEO, framed it bluntly: "Marketing teams are drowning in reporting and repetitive production work. We're turning Claude into an agentic surface for Klaviyo."
Translation. Klaviyo decided their best surface for agent work is Claude, not ChatGPT. Not "Claude and ChatGPT." Not "an LLM-neutral layer for any agent that wants to call us." Claude.
A year ago that decision would have looked like risky platform concentration. This year, with Anthropic at 34.4% on a buyer sample that overlaps Klaviyo's ICP almost cleanly, it looks like product strategy. Personally, I think most martech CPOs ran some version of this math in Q1.
Canva, Claude Design, and the rest of the pull
Klaviyo isn't alone, and the timing isn't accidental. Canva extended its Claude integration so AI-generated drafts inside Claude can move into editable Canva assets without a copy-paste step. Anthropic launched Claude Design to produce decks, landing-page mockups, and marketing UIs inside Claude itself. And Claude Marketplace now lets enterprises burn down committed Anthropic spend on third-party tools from Replit, GitLab, Harvey, Rogo, and Snowflake.
That last part is the one CFOs notice. A committed Anthropic spend pool that pays for adjacent SaaS reduces the procurement friction of swapping an incumbent for a Claude-native alternative. If your CRO automation vendor is Claude-native and the old one isn't, the math at renewal gets uncomfortable fast.
This is the same shape that made AWS sticky in 2014. The compute wasn't where the lock-in lived. The lock-in lived in the marketplace that paid for everything adjacent to it. On paper that sounds like an upgrade for buyers, and sometimes it is. In practice it's also a procurement gravity well that pulls roadmaps toward whichever model layer the marketplace runs on.
What this changes in your stack this quarter
A few practical reads.
If you run marketing automation: check whether your incumbent (Klaviyo, HubSpot, Marketo, Salesforce Marketing Cloud, Iterable) has shipped an MCP connector and where it points. Today, Klaviyo is the most aggressive on Claude. HubSpot's MCP is live but positioned LLM-neutral. If your platform has shipped nothing, you have a roadmap question for your CSM before the next renewal cycle, not after.
If you run a content or lifecycle stack: the gap between "Claude can read our data via MCP" and "Claude can run a campaign brief end-to-end via Cowork" is the gap between an AI experiment and an actual ops cost reduction. Klaviyo just removed the first half of that gap for itself. Most teams I've seen still treat MCP as an integration ticket, not a workflow shift. That seems to be where the next quarter's mistakes live.
If you run paid social or paid search: the TikTok MCP server move and Klaviyo's Claude pivot are the same trend pointed at different surfaces. Vendors are aiming their agentic surfaces at whichever LLM their buyers will renew on. Google Ads still being read-only on MCP looks more isolating every week.
If you're an in-house ops lead: stop asking "should we standardize on Claude or ChatGPT" and start asking "which of our martech vendors has the better Claude integration today." That second question has a different answer than the first, and the second one is the one your finance team will actually act on.
A small caveat worth saying out loud
The Ramp number is real, but it isn't proof Anthropic has won the enterprise. SaaStr's breakdown of an earlier Ramp cut pointed out that velocity and penetration can diverge for a while, and that share among newcomers reads very differently from share among entrenched accounts. OpenAI still has a much larger consumer footprint, and the API revenue picture (which Ramp doesn't see) reportedly still favors them in absolute dollars.
What the data does prove is narrower and more useful. For the segment of the market that buys SaaS through Ramp, paying for Claude went from edge case to plurality inside twelve months. That segment overlaps your buyer if you sell marketing software, and it overlaps your stack if you buy it.
Worth flagging one more thing. Ramp's index also showed an industry-wide bump: overall business AI adoption rose nine points in the same twelve months. So the Anthropic story isn't only a share grab. It's a share grab inside a growing pie, which is the harder version to argue against if you're an incumbent vendor still deciding which model to point at.
The bet most martech vendors quietly already made
Klaviyo's move isn't a brave bet on Anthropic. It's a bet that the customers Klaviyo already has are mostly Claude users now. They probably ran the same number internally that Ramp just published, and they probably ran it three months ago.
I think the more interesting question for the rest of 2026 isn't whether Anthropic stays ahead on this particular index. It's how many martech roadmaps got quietly re-pointed at Claude in Q1 that we'll only find out about in Q3 press releases. From what I can tell, that count is closer to "most of the serious ones" than "a handful." And honestly, the vendors who picked Claude this spring look less exposed right now than the ones still calling themselves model-neutral. That's a sentence nobody would have written at this time last year.
Notice Me Senpai Editorial
