Seven days in August, four major moves in agent payment settlement. Zero Hash joined the x402 Foundation. Mastercard bought BVNK for $1.8 billion. OSL launched AgentPay. Casper Network brought its x402 facilitator to mainnet.
Everyone is racing to build the pipes. Nobody is building the map.
The settlement layer now has: Cloudflare Wallets (Aug 4), OSL AgentPay (Aug 7), Zero Hash (Aug 9), Mastercard/BVNK, Coinbase Base (100M+ agent payments), Stripe (x402 Foundation premier member), Visa (also premier member), Circle Discovery API (900+ curated services).
This is what commoditization looks like. When OSL AgentPay advertises "zero gas fees" and "multi-protocol compatibility" as features, you know the settlement layer is becoming a utility.
So the pipes are getting built. Great. But there's a harder problem that nobody is solving: how does an agent find something worth paying for?
Consider the numbers:
Three out of four x402 endpoints are dead. They return timeouts, DNS failures, or garbage. An agent that tries to discover services by crawling .well-known/x402 will spend most of its time hitting dead ends.
This is the discovery problem. It's not about protocol design — x402 is elegant. It's not about settlement — the infrastructure is being built by billion-dollar companies. It's about knowing what's real.
Settlement is a closed system. You control the endpoints, the keys, the ledgers. You can optimize it, secure it, scale it.
Discovery is an open system. Anyone can register an endpoint. Endpoints go down, change pricing, break their APIs, return wrong data. There's no central authority, no uptime SLA, no truth.
This is why the settlement layer is consolidating into a few serious players while discovery remains fragmented:
.well-known/x402. Sees the 76% dead endpoints. No verification layer.Discovery needs verification. Not just "does this endpoint respond?" but "does it return correct data? Is the pricing fair? Does the payment actually work end-to-end?"
Based on 12,436 trial calls across 328 services, here's what we've learned:
1. Health probes aren't enough. An endpoint can return 200 and still be broken — wrong data, missing parameters, silently failing. You need semantic verification: does the response match the advertised schema?
2. Trial-first is non-negotiable. Agents won't pay for something they haven't tested. 12,436 trial calls on minia2a prove this — the trial is the top of the funnel. Settlement-only providers miss this entirely.
3. Dead endpoints must be removed automatically. If a probe fails 3 times in a row, the endpoint should be delisted. Manual curation doesn't scale to 150K endpoints.
4. Discovery IS the conversion layer. When an agent finds a service that works on the first try, they come back. The recall endpoint on minia2a has 1,739 trials from 93 unique agents — because it consistently returns good results.
The agent payment stack now has three layers:
| LAYER | STATUS | PLAYERS |
|---|---|---|
| Protocol (HTTP 402) | Mature | x402 Foundation (40 members, Linux Foundation), AP2, MPP |
| Settlement | Commoditizing | Zero Hash, OSL, Cloudflare, Mastercard/BVNK, Coinbase, Circle, Stripe, Visa |
| Discovery | Early | minia2a (verify-first), Circle Discovery API, CDP Bazaar |
The protocol layer is done — 200M+ transactions, 40 foundation members, Linux Foundation governance. The settlement layer is being built by the world's largest financial infrastructure companies. But the discovery layer — the thing that tells an agent what's worth paying for — is still wide open.
Whoever solves discovery doesn't just get market share. They get to define what an agent payment even means to the agent making the decision.
Data from minia2a /api/stats as of August 10, 2026. 76% dead endpoint figure from independent x402 ecosystem audit (arXiv 2607.19545). Mastercard/BVNK and Zero Hash details from public announcements.