The agent directory is the new affiliate channel — and most sellers are going to miss it
Cross-border operators have spent a decade optimizing for one thing: being findable at the moment of intent. That meant Amazon search rank, then TikTok Shop’s For You feed, then Temu’s algorithmic shelf. The next discovery surface looks nothing like a search bar. It looks like a contact list. Agent Store, launched by Stanley Virgint under the Soar banner, is a marketplace of 137 texting agents — travel companions, matchmakers, betting copilots, recruiters, health coaches — that live as phone numbers rather than apps. For a seller, the interesting question isn’t whether you’d text a recruiter bot. It’s who controls the directory when your customer’s first question about a product gets asked inside iMessage instead of on your PDP.
What Agent Store actually is, stripped of the launch-day gloss
The pitch is deliberately simple: browse by category or trending, text any agent that speaks to you. The entire marketplace can be navigated over text — you can message the store itself at +1 (424) 278-0248 to explore listings and get weekly updates on new agents. There’s also a native iMessage app that keeps the directory in your keyboard, which the maker argues makes agents “so much easier to remember to use.”
The numbers are modest but the trajectory claim is aggressive. The ecosystem grew to 137 agents in three months at launch, and the maker told one commenter it had already hit 139, with an expectation of “millions of agents over the next decade.” Monetization mirrors the App Store: most agents are free to text, some carry in-agent purchases or subscriptions, and the team is building paid/unpaid filters. Quality control is human-in-the-loop — submissions go through a submit link and most go live within 24 hours if they pass “standard safety and usability checks.”
That’s the whole product. No API docs, no revenue share disclosed, no pricing page. Which is exactly why it’s worth a seller’s attention rather than dismissal.
Why this is a directory play, not a chatbot play
The comment thread gives away what’s actually being built. One user compared it to the yellow pages; another said having agents “live natively in my phonebook” is what makes them memorable. Both are describing distribution infrastructure, not AI capability. The hard part of the agent economy was never building a bot — it’s the same problem every DTC brand has: getting a stranger to remember you exist at the moment they have a need.
Whoever owns the directory owns the default. That’s the lesson from Shopify building an app store, from Amazon Seller Central owning the buy box, and from Klaxoon-style marketplaces that aggregate supply. Agent Store is trying to be the App Store for a category that doesn’t have one yet.
The comparison that matters: it’s not competing with ChatGPT
The lazy read is “another AI wrapper.” Wrong frame. Compare it to three real incumbents instead.
Against ChatGPT and its GPT Store: OpenAI’s directory is web-first, requires account context, and has struggled with discoverability and spam. Agent Store’s bet is that SMS/iMessage is a lower-friction surface — no app install, no login, just a saved contact. Whether that bet pays off is unproven, but the friction argument is real for older demographics and low-tech markets.
Against Intercom and Zendesk: those are B2B support stacks priced per seat, sold to brands. Agent Store is consumer-facing and, for now, free to browse. Different buyer entirely.
Against Gorgias and the Shopify app ecosystem: those tools live inside a store’s own funnel. Agent Store lives outside it, in the customer’s messages app. That’s the strategic threat — and the opportunity.
Why Amazon sellers should care more than Shopify ones
Shopify operators already own their customer relationship: email list, SMS list, Klaviyo flows, Postscript for SMS. If a texting agent marketplace emerges, they can plug in and treat it as one more channel.
Amazon FBA sellers can’t. They have no customer email, no phone number, no direct line. Their entire discovery model is rented from Amazon’s search algorithm. If a meaningful slice of product discovery migrates to conversational agents — “find me a good travel stroller under $200” texted to a bot — the seller who isn’t listed in that agent’s inventory is invisible, and there’s no Helium 10 keyword tool that fixes it. That’s the asymmetry. Marketplace sellers should be watching this category harder than DTC brands, because they have fewer fallbacks.
What a cross-border seller can actually borrow from this
Three transferable plays, in order of how fast you could test them.
First, treat conversational listing as a new SKU surface. If agents are going to recommend products, they need product data they can parse. That means clean, structured feeds — title, price, shipping window, return policy, country of origin — exposed wherever an agent might query. Most sellers’ product data lives in a Shopify admin or an Amazon flat file and nowhere else. The operators who win the next two years will be the ones who publish structured catalog endpoints agents can actually read.
Second, build your own agent before you rent someone else’s shelf. A brand-owned texting agent that handles order status, sizing questions, and restock alerts is a defensible asset. Shopify Inbox and Tidio already do fragments of this. The version that matters is one your customer saves as a contact — same mechanic Agent Store is betting on.
Third, watch the paid/unpaid filter as a signal. The moment Agent Store introduces paid placement or sponsored listings, the economics change from curiosity to channel. That’s the trigger to actually budget against it.
Where the math breaks
The honest problem: 137 agents is not a marketplace, it’s a beta. The maker’s own comment projects “several million” agents over the coming years — a claim with no supporting evidence in the thread. Discovery at 137 listings is trivial; discovery at 137,000 is a search problem the team hasn’t demonstrated they can solve. And “millions” of agents implies a spam problem that human-in-the-loop vetting cannot scale to.
There’s also the platform risk nobody in the thread raised. Apple owns iMessage. The entire product runs as a guest on infrastructure Apple could restrict, reprice, or absorb into its own Siri/Apple Intelligence roadmap. Anyone building a business on top of Agent Store needs to price that risk in.
What’s genuinely unclear (and I’d want disclosed)
- Revenue model: not disclosed. No take rate, no listing fee, no subscription tier mentioned.
- Data handling: what does Agent Store see when a user texts an agent? Not disclosed.
- Cross-border readiness: no mention of multi-currency, localized agents, or non-US phone number support. The store number is US-only.
- Agent portability: can a brand migrate its agent off the platform? Not disclosed.
What I’d watch / test next
Three concrete moves this week.
- Text the store number yourself. Save +1 (424) 278-0248, browse a category relevant to your vertical, and note how the agent handles a product-intent query. Ten minutes tells you more than any launch post.
- Audit your product feed for agent-readability. Pull your top 20 SKUs and check whether title, price, shipping window, and return terms are exposed in a structured, machine-readable way outside your storefront. If they aren’t, fix the top five.
- Submit a test agent. The submission process is free and reportedly live within 24 hours. Even if you never commercialize it, you’ll learn what the vetting bar looks like — and whether free distribution here is worth the setup cost.
The category is early enough that being wrong costs you a weekend. Being late costs you a channel. I know which trade I’d take.
