The Merchant-of-Record War Is Now an AI-Agent War — and Cross-Border Sellers Are the Real Prize
Cross-border sellers have spent the last three years duct-taping together Stripe, a tax engine, a subscription biller, an affiliate tracker, and a spreadsheet to reconcile it all. Every one of those layers is a place where money leaks: VAT remittance errors, failed affiliate payouts, refund clawbacks nobody modeled, and a checkout that bounces buyers off-domain right at the moment of intent. So when a merchant of record starts pitching itself as a “full commerce surface” that an AI agent can configure end to end, that’s not a SaaS curiosity — it’s a direct shot at the operational mess most DTC and marketplace-adjacent operators live inside. Creem just shipped Creem 2.0 with exactly that thesis, plus a €5M raise. Here’s what I think it means for people who actually move goods and subscriptions across borders.
What Creem 2.0 actually solves — and who it’s really built for
Let me be blunt about the framing. The founder, Gabriel Ferraz, opened the launch with a three-line thesis that reads like a manifesto for the AI-native builder: building a product collapsed from months to a weekend, but getting paid for it didn’t, and therefore the winners are the ones who monetize fastest, not the ones who code fastest. That’s a developer-tools argument, not a cross-border-retail argument. But the mechanics underneath it are genuinely relevant to sellers, because the merchant-of-record model is the same thing that makes Shopify Markets, Paddle, and Lemon Squeezy attractive to anyone selling digital or hybrid products into the EU, UK, and LATAM without standing up local entities.
Creem’s pitch, as stated on the launch page, is that it acts as the seller of record and handles payments, taxes, and payouts across 190+ countries, charging only when you earn — 3.9% + 40¢, no monthly fees. That “only pay when you earn” line is the wedge. Incumbents like Paddle and Lemon Squeezy have historically layered monthly minimums, setup friction, or approval gates that punish pre-revenue sellers; Stripe charges nothing monthly but leaves the tax and remittance burden squarely on you. For a cross-border operator testing a new SKU or a subscription tier in a new market, the difference between “pay 3.9% when a sale clears” and “pay $49/month while you wait for approval” is real cash flow.
The 2.0 feature set is where it gets more interesting for operators who run hybrid models — physical goods plus digital add-ons, or a subscription with usage-based overage. The launch highlights credit wallets, usage-based pricing, custom meters, flexible affiliates, short links, and a CLI that an agent can drive to set up, run, and monitor a store end to end. In the comments, the team confirmed that a monthly subscription can auto-grant credits on purchase and every renewal, claw them back on refund, and auto-refill when a customer runs low. That’s the exact billing primitive that kills most hybrid DTC models — the “you bought 500 credits, you used 480, now what” problem that usually ends up as a manual support ticket.
Why Amazon sellers should care more than Shopify ones
Here’s my contrarian take: the operators who should be paying closest attention to this launch aren’t the Shopify DTC crowd — they’re Amazon FBA brand owners running a side SaaS, a warranty extension, a replenishment subscription, or a digital companion product off-platform. Amazon’s Seller Central gives you almost no native recurring-billing or usage-based infrastructure, and the moment you try to move a customer from a one-time Amazon purchase into a direct subscription, you’re stitching together Stripe Billing, a tax service like TaxJar or Avalara, and an affiliate tool. Creem collapses those into one MoR with a single fee. The catch — and it’s a big one — is that Creem is fundamentally a software-commerce MoR. There’s no evidence on the launch page that it handles physical fulfillment, shipping tax nexus, or marketplace-specific compliance. So for the Amazon seller, this is a tool for the digital and subscription layer of your business, not a replacement for your FBA stack.
Where the affiliate math breaks
The affiliate feature is the one I’d stress-test hardest. The team says affiliate payouts run through Creem’s payout infrastructure automatically, with splits to any number of recipients, per product or store-wide, each with their own dashboard, and refunds clawing back each party’s share. That’s a genuine improvement over the spreadsheet-and-PayPal nightmare one commenter described. But here’s the operational trap: cross-border affiliate programs often involve creators in jurisdictions where you’d normally issue a 1099 or local equivalent, and the MoR model shifts who is legally paying that affiliate. If Creem is the payer, you may lose visibility into your own affiliate tax reporting and relationship ownership. I don’t have data from the launch page on how Creem handles affiliate tax docs or W-8/W-9 collection — that’s a question I’d put to their team before migrating a program with more than a handful of creators.
The AI-agent angle is the real story, and it’s double-edged
The part of this launch that I think most cross-border operators are underestimating is the CLI and MCP story. One commenter asked how hard setup is for a first-time developer; a maker replied that with the Creem CLI, ChatGPT set up all products and pricing in minutes and integrated them into the app, with a team review taking up to two days — so realistically signup to launch in one to three days. Another commenter said they had Grok walk them through every step. A third said they were set up in under 24 hours.
That’s the actual product shift. It’s not that Creem added AI features; it’s that the configuration surface of a merchant-of-record is now something an LLM can drive. For a lean cross-border team — say, two founders running a DTC brand and a digital product — that means you can stand up a new market’s checkout, pricing tiers, and affiliate structure in an afternoon instead of a sprint. The founder’s line that “your agent sets it up in a prompt and keeps running and monitoring the store after” is marketing, but the direction is credible.
The compliance question nobody on the launch page answered
Here’s where my judgment says the enthusiasm needs a cold shower. A merchant of record is a regulated financial intermediary. Letting an AI agent configure pricing, credits, and affiliate splits is fine right up until the agent misconfigures a tax rate or a refund clawback and you’ve now got a VAT remittance error in three jurisdictions. The launch page shows no mention of audit logs, approval workflows, or human-in-the-loop controls for agent-driven changes. For a solo SaaS founder, that’s acceptable risk. For a cross-border seller doing seven figures with real tax exposure, “my agent reconfigured my store” is a compliance incident waiting to happen. I’d want to see change history, role-based permissions, and a staging-to-production promotion flow before I’d let an agent touch a live store with meaningful volume.
What cross-border sellers can borrow from this playbook
Even if you never sign up for Creem, there are three transferable moves here.
First, unbundle your payment stack from your tax stack from your affiliate stack — then decide what to re-bundle. The reason Creem’s pitch lands is that most sellers have never actually mapped the true cost of their payment infrastructure. Add up Stripe fees, your tax service subscription, your affiliate tool, your subscription-billing add-on, and the hours of reconciliation, and the effective rate is often well north of 5%. Creem’s 3.9% + 40¢ is only compelling relative to that fully-loaded number, not relative to Stripe’s headline rate. Do the math on your own stack this week.
Second, treat “agent-configurable” as a buying criterion going forward. Not because you’ll use it today, but because the tools that let an LLM drive setup and monitoring will compound faster than tools that require a human in a dashboard. The comment thread shows real users doing exactly this with Creem’s docs and third-party models. If your current billing provider’s API requires a human to click through five screens for every new pricing tier, you’re accumulating operational debt.
Third, watch the embedded-checkout feature closely. A commenter specifically asked whether checkout can be embedded in-app rather than sending buyers off-domain; a maker confirmed an iframe embed with React, Vue, and Svelte wrappers, with Apple Pay and affiliate attribution working inside the embed. For cross-border DTC, on-domain checkout is a conversion and trust issue — buyers in Germany and Japan are measurably more sensitive to being redirected at payment. If Creem’s embed works as described, it’s a genuine differentiator against older MoRs that force a redirect.
The €5M raise: signal or noise?
The founder announced a €5M raise to build “the next generation financial OS for the lean, modern teams building with AI.” I’d treat the number as directionally meaningful but not decisive. €5M is enough to build product and hire, not enough to win a payments war on capital alone. What matters more is the stated mission: letting lean teams of humans and AI agents do business globally. If that’s real, the competitive threat isn’t Paddle or Lemon Squeezy — it’s Stripe moving further into tax and MoR territory, or Shopify expanding its own merchant-of-record offering for Markets Pro. Creem’s window is the next 18–24 months before the incumbents close the agent-configurability gap.
Where my judgment says it falls short
Three honest concerns.
One: the review history shows real friction. A tester flagged that during product creation, the same product gets created twice, and that in the API layer, setting a custom price becomes mandatory for pay-as-you-go. The team hasn’t publicly confirmed a fix on the launch page. Duplicate product creation in a billing system is the kind of bug that produces double charges and chargebacks — the single most expensive failure mode in cross-border commerce. If you’re evaluating Creem, test product creation in sandbox mode aggressively before going live.
Two: the MoR model concentrates risk. When Creem is the seller of record, they own the customer relationship for tax and payment purposes. If they change pricing, get acquired, or hit a compliance issue in a key market, your ability to collect revenue in that market is suddenly their problem, not yours — but you feel it. This is the same trade-off every Paddle and Lemon Squeezy customer accepts, and it’s worth being explicit about rather than pretending MoR is free.
Three: no physical-goods evidence. The entire launch is framed around SaaS, AI products, and digital commerce. There’s nothing on the page about shipping, returns, physical SKU tax treatment, or marketplace integration. If you’re a pure-play Amazon FBA or TikTok Shop seller, Creem is not your checkout — it’s a potential layer for your subscription or digital add-on business. Don’t let the “190+ countries” headline fool you into thinking it replaces your marketplace infrastructure.
What I’d watch / test next
This week, if you run any recurring or usage-based revenue off-platform, do three things. First, pull your last 90 days of payment-related costs — processor fees, tax service, affiliate tooling, subscription add-on, and reconciliation hours — and compute your true effective rate. Second, sign up for Creem in test mode (the team says sign up at creem.io, build products in test mode, then submit for verification) and deliberately try to break product creation and the pay-as-you-go pricing flow to see if the duplicate-product bug still exists. Third, if you have an affiliate program, ask Creem directly how they handle affiliate tax documentation and whether you retain visibility into creator relationships — that’s the unanswered question that would block my own migration.
I’d also watch two signals over the next two quarters: whether Creem ships audit logs and approval workflows for agent-driven changes (that’s the enterprise-readiness tell), and whether an incumbent like Paddle or Stripe announces comparable agent-configurable billing. If neither happens, Creem has a real window. If both happen, the MoR market commoditizes fast — and the winner will be whoever owns the seller relationship, not whoever has the slickest CLI.
