Every x402 conversation eventually dead-ends at the same question: how does an agent find a resource worth paying for? The protocol defines how to ask for money — the 402 Payment Required challenge, the pay-to address, the settled receipt. It does not define how a buyer discovers that the resource exists in the first place. That is the discovery layer, and it is still the least-solved part of the stack.
After spending weeks on both sides of this problem — publishing discovery records, checking how crawlers read them, and getting our own paid endpoints listed on agent-commerce directories — we landed on an empirical finding that is worth stating plainly, because it reframes how resource servers should think about distribution.
The finding: declaring yourself an x402 resource — a DNS record, a well-known manifest — is necessary but not sufficient to appear in the agent-commerce directory that matters most today. Indexing is triggered by a real on-chain settlement through a registered facilitator, not by the metadata you publish. Self-declaration gets you parsed. Settlement gets you listed.
There are two ways an x402 payment can land:
| Path | What happens | Does the directory see it? |
|---|---|---|
| Facilitated settlement | The buyer's client sends a payment-signature; the resource server forwards it to a registered facilitator, which settles on-chain and returns a receipt. | Yes — indexing is triggered from this path. |
| Direct transfer | The buyer sends the transfer straight to the resource server's pay-to address and passes the transaction hash back as proof. | No — the settlement is real, but the directory never sees it, so the endpoint stays invisible. |
The distinction matters more than it looks. A direct transfer is a perfectly valid payment. The buyer's money moves, the resource server gets paid, the receipt verifies. But because it bypasses the facilitator, the one party the indexer is watching never observes the transaction — and the indexer, quite reasonably, treats a resource that has never settled through its rails as unproven.
Put differently: the facilitator is the indexer's eyes. Discovery is not built by crawling manifests and trusting them. It is built by watching real money settle through known rails, and listing the resources that money actually reached.
If you are standing up an x402 endpoint and wondering why you are not showing up anywhere, the answer is probably not your metadata. The checklist people reach for first — DNS TXT records, /.well-known/x402 manifests, OpenAPI pointers — is about being readable, not about being listed. Those records make you parseable once someone looks at you. They do not cause anyone to look.
The signal that causes a directory to look at you is a real, facilitated settlement. This has a few practical consequences:
The instinct is to read "settlement-gated" as a hurdle. It is the opposite. A discovery layer that keys off real settlement is the first x402 surface we have seen whose entries are proven live rather than merely claimed live. It quietly solves the liveness problem that plagues every self-submitted directory: dead endpoints, stale prices, abandoned paywalls.
For agents, that is exactly the signal that matters. When a buyer's wallet is about to move, it wants to know the resource is real, the price is current, and the pay-to address is the one that was actually settled to — not the one a stale manifest claims. A settlement-gated index encodes all three by construction.
Discovery follows money. Once you accept that, the playbook for an x402 resource server gets simpler: publish clean metadata so you are readable, then push real volume through a registered facilitator so you are listed. The first gets you parsed. The second gets you found.