x402 Payability Observatory — week 2026-W44
How much of the x402 surface an agent can actually pay, and why the rest can't. Every endpoint Settled tracks is probed unpaid on the verb its listing declares and, when that fails, on the other verb; payees are checked on Base; a scout wallet buys from every seller listed up to $0.25 in turn, for real.
Generated 2026-10-29 15:17 UTC · JSON · sample CSV (CC BY 4.0) · signed daily snapshot
Payability
| Verdict | Endpoints | Meaning |
|---|---|---|
unreachable | 63,548 | no HTTP answer |
payable | 53,738 | valid 402 quote on the verb that answers; nothing paid yet |
unknown | 8,321 | not established yet |
cannot_pay_either_verb | 5,978 | neither GET nor POST yields a 402 or a 2xx |
delivered | 2,482 | paid and delivered — a real payment came back with a real response |
free | 386 | 2xx without a paywall |
malformed_402 | 332 | answers 402 but the quote is malformed |
auth_gated | 128 | 401/403 before any payment |
unpriced | 20 | valid 402 but no accept we can price |
Scout (real payments)
734 payments this week, $21.96 spent. 700 came back with content (95.4%); 433 matched their listing and 32 did not. Every purchase, with its transaction. 3,689 payee addresses checked on Base; 495 had never been used.
By listed price
| Listed price | Sellers bought this week | Purchases | Spent | Delivered | Paid, nothing back | Sellers on hold |
|---|---|---|---|---|---|---|
| up to $0.01 | 534 of 1,672 | 538 | $2.18 | 97.2% | 12 | 6 |
| $0.01 to $0.05 | 57 of 380 | 81 | $2.52 | 95.1% | 3 | 2 |
| $0.05 to $0.25 | 83 of 174 | 94 | $9.39 | 89.4% | 9 | 0 |
| $0.25 to $1.00 | 13 of 88 | 16 | $7.88 | 68.8% | 3 | 0 |
A seller is a payTo wallet, counted in the tier of its cheapest payable endpoint; purchases count at the listed price of what was bought. The scout buys each seller's cheapest endpoint in turn: sellers up to $0.25 at most once a week, and sellers whose endpoints all cost more (up to $1.00) at most once every 60 days, a few rows a run from a daily allowance of their own. Delivered: content came back. Paid, nothing back: the scout paid and got nothing usable. On hold: that happened twice in 30 days across more than half of the endpoints the scout bought from the seller, so the scout stopped buying and /v1/preflight answers avoid until a purchase delivers.
Biggest live hosts
| Host | Live | Payable |
|---|---|---|
| saylorinnovations.com | 1,593 | 1,816 |
| dev.intel.rallylive.ca | 1,583 | 1,583 |
| monitor.intel.rallylive.ca | 1,344 | 1,344 |
| www.mahastrategies.com | 1,155 | 1,155 |
| data.intel.rallylive.ca | 1,142 | 1,142 |
| ip.intel.rallylive.ca | 1,073 | 1,073 |
| dns.intel.rallylive.ca | 1,063 | 1,063 |
| site.intel.rallylive.ca | 1,063 | 1,063 |
The thread
x402 Payability Observatory · Oct 22–29 134,933 endpoints tracked, 126,612 probed unpaid (GET, then POST if needed): • 56,220 payable (44.4% of probed) · +47,102 new · 65,587 went dark Data (CC BY): settled.tools/observatory
2/ Why the rest can't be paid (every endpoint is tried on GET and POST before it is written off): unreachable: 63,548 pays on neither verb: 5,978 free (no paywall): 386 malformed 402: 332 auth-gated: 128 unpriced: 20
3/ Scout: 734 real USDC payments to 690 sellers this week, 700 delivered content (95.4%), 433 matched their listing, 32 did not, $21.96 spent. 2,588 endpoints carry a paid-and-delivered verdict.
4/ 4,006 live hosts · $9,535 in x402 payments reached 656 seller wallets, Oct 22–28 (ledger: Sep 24–Oct 29, 2,463 indexed sellers). The biggest hosts: saylorinnovations.com: 1,593 live, 1,816 payable dev.intel.rallylive.ca: 1,583 live, 1,583 payable
5/ Every verdict is EIP-191 signed and the whole index is snapshotted daily, hash-chained (verify at /v1/verify, /v1/snapshots). Sample CSV: CC BY 4.0. Other measurers: publish your verdicts for the same sample and let's compare where we disagree — reply. Not financial advice.
Method and license
Liveness: an unpaid request with Accept: application/json on the declared verb, then the other verb if that one returned neither 402 nor 2xx; three consistent failures before a terminal status, except that a host with no successful probe for 6 hours has all its live endpoints marked dead at once (flag host_down); three of them are then probed every 30 minutes, and when one answers the rest get their status back at once. A stable endpoint is re-probed about every 2 days, less often on a host with thousands of endpoints (x2 from 5,000, x3.5 from 20,000). Payability derives from the damped status plus the latest quote. Payee existence: USDC balance, nonce or code on Base via Multicall3. Delivery: real payments by the scout wallet, coverage first: every seller whose cheapest endpoint is listed up to $0.25 in turn, at most once a week, and sellers up to $1.00 at most once every 60 days; each purchase is judged on a 2xx with non-empty, non-error content. x402 payments: USDC payments into indexed sellers' wallets on Base whose amount equals a price the seller quotes (payments at a price the seller does not quote are not counted; rows recorded before the 0.9.87 deploy on 2026-10-14 hold only daily totals, and for them a day's payments count when they average $1 or less), leaving out self-payments, wallets that paid more than 20 sellers in the window (crawlers and monitors), and wallets the seller itself sent USDC to in the window or the 7 days before it (a seller paying itself through wallets it funds; counted since 0.9.107); counted over whole UTC days, Oct 22–28. The sample CSV and this page are licensed CC BY 4.0; attribute "Settled (settled.tools)". Other measurers: publish your verdicts for the same sample and let's compare — reply on X or write in.
Data: Settled probes, sanctioned venue APIs and Base on-chain payout logs. Best-effort, not financial advice. Labels marked "suspected" can be disputed. Machine-readable: /v1/income/venues · llms.txt · openapi.json