The Agent Commerce Land Grab: ZeroClick's $55M, Circle's Discovery Layer, and the Race to Aug 14

Two things happened this week that change the agent commerce landscape:

1. ZeroClick launched with $55M — founded by Ryan Hudson (Honey → sold to PayPal for $4B), backed by the same investors. Positioning: "Shopify for AI agents."
2. Circle shipped a discovery layer for their Agent Marketplace — 900+ endpoints across data, compute, and infrastructure. Agents can discover, evaluate, and pay in USDC via x402.

Combined with Cloudflare Wallets (Aug 4), OSL AgentPay (Aug 7), and Claude Code auto mode going default on Aug 14, we're watching the agent commerce stack assemble in real time. The question is no longer "will agents pay for things?" — it's who controls the storefront where they shop.

ZeroClick: $55M to Build the Shopify for Agents

ZeroClick's premise is straightforward: 57% of web traffic is now non-human, and AI agents are the fastest-growing segment. These agents arrive ready to transact but find no storefront capable of completing a purchase. ZeroClick converts existing APIs into agent-purchasable storefronts — revenue settles into the business's existing Stripe account.

Ryan Hudson knows distribution. Honey reached 17 million users before the PayPal acquisition. The playbook: embed a purchasing capability into existing agent workflows, then own the discovery layer through distribution. ZeroClick also operates a reasoning-time ad network — weaving sponsored context into AI assistant responses. This is the same playbook Google used to own web discovery: give away the search, sell the placement.

Circle's Discovery Layer: 900+ Endpoints and Counting

On August 6–7, Circle introduced a discovery layer for its Agent Marketplace. Agents can autonomously find, evaluate, and pay for services using USDC via x402. With 900+ endpoints already indexed, Circle has the largest curated catalog in the space.

But size isn't everything. Independent audits show 76% of x402 endpoints are dead — they return errors, time out, or serve stale data. A catalog with 900 entries where 684 are dead isn't a marketplace; it's a graveyard with good SEO.

This is the verification gap. Discovery without verification is a directory. Discovery with live health probes and trial data is a marketplace. The difference: can an autonomous agent trust that the endpoint it finds will actually work?

The Landscape: Who's Building What

Three Strategic Questions for the Agent Commerce Space

1. Is discovery a product or a feature?

Circle treats discovery as a feature of their settlement stack. ZeroClick treats it as a product — the storefront is the business. If discovery is a feature, it gets bundled into whoever owns the payment rail. If it's a product, it's an independent layer.

The web's history suggests independent discovery layers win: Google didn't own the payment rails, and Visa didn't own search. But the agent economy is younger and more vertically integrated. Cloudflare could bundle discovery into Wallets. Circle could bundle it into USDC settlement. The window for an independent discovery layer is open, but it won't stay open forever.

2. Does distribution or verification matter more?

ZeroClick is betting on distribution: get into the agent workflow (via the ad network and Stripe integration), and the storefronts follow. Circle is betting on curation: build the largest catalog, and agents will come.

But agents aren't humans browsing an app store. They're programs making API calls. They don't need a beautiful storefront — they need a machine-readable catalog with verified liveness. What matters to an autonomous agent:

• Does this endpoint return 200 and valid JSON? (health probe)
• What does the response look like for a real input? (trial data)
• How much does it cost, on which chain, with which token? (machine-readable 402 headers)
• Can I register and pay without a human in the loop? (one-command registration)

If your discovery layer answers these four questions in ~500ms, you win the agent. If it requires a human to browse, evaluate, and decide, you're building for the wrong customer.

3. Who captures the value: the marketplace or the endpoint?

ZeroClick's model (businesses keep their Stripe accounts, ZeroClick takes a platform fee) is the marketplace playbook: aggregate supply, own demand, capture a take rate. But in the x402 ecosystem, endpoints set their own prices and keep ~95% of revenue. The protocol's take rate (~5%) goes to the facilitator, not a marketplace.

The tension: a centralized storefront (ZeroClick) vs. a protocol-native discovery layer (x402 ecosystem). The app store model vs. the web model. Apple's 30% vs. the open web's 0%.

What This Means for Aug 14

Every launch in the past 10 days has been timed for Aug 14. The payment rails (Cloudflare Wallets, OSL AgentPay, Stripe MPP). The discovery layers (Circle, ZeroClick). The infrastructure (Sapiom, Kitesurf). The economic unlock (free classifier tokens). They're all positioning for the moment millions of autonomous agents start encountering 402 paywalls as routine execution flow.

The endpoints that are ready — with machine-readable 402 headers, trial paths with one-command registration, and discovery endpoints — will capture the first wave. The ones that aren't will return 402s that agents can't parse and move on.

The Open Questions

Four things we don't know yet:

1. Will autonomous agents actually spend money? The infrastructure works. The 402 responses are machine-readable. The wallets are funded. But we haven't seen autonomous spending at scale. Aug 14 is the first real test.
2. Will discovery fragment or consolidate? Circle, ZeroClick, Coinbase Agentic.Market, Loomal, Aidress, and a dozen smaller directories. A fragmented discovery layer means agents need to check multiple catalogs — exactly the problem discovery is supposed to solve.
3. Does the app store model or the open web model win? ZeroClick's centralized storefront vs. the x402 protocol's open marketplace. The internet's history says open wins. But the mobile era says walled gardens are profitable.
4. What's the killer use case? The top endpoints on every marketplace are infrastructure primitives: gas prices, CAPTCHA solving, service discovery, memory. Not AI generation. Agents already have LLMs — they need the things LLMs can't do. The first wave of successful x402 businesses will sell infrastructure, not content.

4 days to auto mode. The land grab is on.