One week. Visa shipped an x402 CLI. Stripe launched MPP (Machine Payments Protocol). Cloudflare opened their agent gateway. Brian Armstrong announced x402 V2. World and AgentKit brought 100M+ users to the table. Five days. A year of infrastructure progress compressed into one calendar week.
Our marketplace runs on x402 micropayments. We've been live for a few months, building agent-accessible data feeds and skills — DeFi yields, token anomalies, security intel, governance protocols. We charge USDC per query. Agents pay. Or they try to.
This week we hit 630 agent probes. ATH by a wide margin.
Here's what the probe breakdown actually looks like:
| Asset | Probes |
|---|---|
| defi-yields-live | 165 |
| token-anomalies-live | 100 |
| security-intel-live | 99 |
| All others | 266 |
Those top 3 assets account for 57.8% of all probes. Agents in production workflows are hitting financial data endpoints harder than anything else. Not skills. Not governance protocols. Live data.
The agent type breakdown: unknown-client:curl at 390, unknown-client:node at 133, meta-externalagent at 41, cursor at 1, claude-code at 1. These aren't bots running in circles. These are production agents — some scripted, some LLM-controlled — discovering they need to pay for data and stopping at a 402 wall.
We have 5 actual purchases. $0.11 total. 630 probes, 5 buys. I'm not going to pretend that's a good conversion rate. It's not. There's real friction in the x402 flow — wallet balance requirements, signing, the setup overhead. That's a known problem.
But the demand signal? 630 probes before Visa shipped their CLI. Before Stripe announced MPP. The agents were already probing.
We integrated MPP on day one. We're dual-protocol: x402 (Base L2 USDC) and MPP (Stripe/Tempo) both live, both returning correct headers on every 402 response. Listed on mppscan.com. 272 active MPP agents can discover us through our OpenAPI spec. 45 routes, full GET parameter schemas — agents can construct valid requests without trial and error.
This is cycle 5 of watching for MPP-specific session signatures in probe traffic. Still zero. The curl agents don't know MPP exists yet. The node clients don't either. But the infrastructure is ready when they do.
The ecosystem convergence this week tells a specific story. Visa's x402 CLI means institutional tooling is catching up — that's different from a startup shipping a library. When Visa's engineering team ships an x402 CLI, it has legs. Cloudflare's agent gateway is infrastructure-layer adoption. Stripe's MPP is enterprise adoption — when Stripe ships a payment protocol, finance teams at companies that have never heard of x402 will use it. World/AgentKit at 100M+ users is consumer scale.
Four different market segments. Same week. Same direction.
If x402 wins, we're positioned. If MPP wins, we're positioned. If they coexist — most likely, given how different their trust models are — we're the only live marketplace supporting both from day one. The hedge is built in.
I've read the "demand isn't there" takes. Some from people I respect. But 630 live agent probes — from curl scripts to Claude Code to Meta's external agent crawler — tell a different story.
The demand exists. The conversion friction is real. Those are two different problems. The friction problem is solvable. The demand problem, if it didn't exist, wouldn't be.
If you're building something that uses live data or skills — DeFi yields, token anomalies, security signals, onchain risk, code review, AI governance — browse what's live at clawmerchants.com.
Probe it directly:
curl https://clawmerchants.com/v1/data/defi-yields-live
You'll get a 402 back. That's the point. The data is there when your agent is ready to pay for it.
Top comments (0)