Executive summary
For three issues now, Decixa has reported how many x402 APIs are verified live — endpoints that answer a real HTTP 402 handshake with a parseable payment requirement. That count keeps climbing: 5,523 in April, 17,176 in May, 23,849 now — up +38.9% since May.
This issue asks a different question for the first time: of those 23,849 listings, how many are actually used? The answer is 466 — 1.95%, or roughly 1 in 51. (Here and throughout this issue, “used” means Decixa recorded at least one call to that API in our usage window — see Methodology for exactly how we measure this.) Being listed and being used are different things, and the gap between them runs through nearly every part of the catalog — not just one weak corner of it.
Verified-live grew +38.9% since May — 17,176 → 23,849.
Measured directly for the first time: only 466 APIs (1.95%, ~1 in 51) were actually used this month.
The gap is not confined to small categories. Decixa classifies every API by what it does — its “capability” — and the two largest capability groups, Analyze (interpretation/analysis APIs) and Extract (data-retrieval APIs), together 82.1% of the catalog (10,143 and 9,439 listings), are used at just 1.12% and 2.73% respectively. Store (data-storage APIs) sits at 0%.
Orbis — one of the largest hosts relaying x402 APIs — saw its share of the catalog fall from 70.1% to 48.2%, a real diversification signal. But it still can’t report a price on any of its 11,506 listings, and is used at just 1.21%, less than half the 2.65% rate everywhere else.
Next issue: a composite Decixa Score — verification, usage, uptime, and price stability combined — replaces raw counts as the trust signal agents use to choose between APIs.
The gap between listed and used
The headline number kept climbing in July: verified-live crossed 23,849, up +38.9% from May’s 17,176. On its own that is a supply story — more of the catalog clears our HTTP 402 probe. It says nothing about demand. This month, for the first time, we can also report how much of that catalog is actually used.
Orbis's own listing count fell slightly (−4.5%) even as the total catalog grew.
23,849 listings do not mean 23,849 usable capabilities, any more than 17,176did in May. What’s different this month is that we can finally say how much of that catalog is actually used: 466 APIs, 1.95%, roughly 1 in every 51 listings.
The range runs from Store (data-storage APIs) at 0% to Transact (payment-execution APIs) at 3.57% — a gap that narrows category by category but never closes. Analyze and Extract, the two largest groups by volume ( 10,143 and 9,439 listings, together 82.1% of the catalog), convert at just 1.12% and 2.73%— inside the same range as everything else, not above it. This isn’t one under-performing category dragging down an otherwise healthy market. It’s a structural gap that runs through the whole catalog, largest capabilities included.
One measurement caveat applies to the 1.95% figure above — see Methodology below. It is a conservative floor, not a ceiling.
The host that can't be priced
The 1.95% usage gap in Section 02 is not simply an Orbis effect. Excluding Orbis entirely, only 2.65% of the rest of the catalog shows measurable usage — nearly identical to the market-wide rate. But Orbis still matters disproportionately: it represents 48.2% of the entire verified-live catalog, more than any other host by a wide margin.
In May, one host — Orbis (orbisapi.com) — accounted for 70.1% of everything verified live. In July that is down to 48.2%. That is a genuinely positive signal: the market is diversifying away from a single dominant host, and Orbis’s own listing count actually fell slightly — from 12,048 to 11,506 — even as the total catalog grew +38.9%.
Orbis remains, by a wide margin, the largest single host in the index. The next-largest is api.tcgapi.dev at 837 listings — 3.5% of the catalog. Across the whole verified-live pool there are 1,145 distinct hosts; none but Orbis comes close to double digits of share.
Both bars scaled to Orbis’s listing count, to show the gap to the next-largest host.
But Orbis’s dominance comes with two structural weaknesses that show up nowhere else in the catalog at this scale. First: not one of its 11,506listings has a usable price. This is not a gap in Decixa’s own measurement — when we re-checked Orbis-hosted routes this month, their proxy URLs (orbisapi.com/proxy/<slug>-<hex>) returned HTTP 404 (not found) instead of the 402 they used to. The links themselves appear to have gone dead as Orbis rotates them — a real reliability issue on Orbis’s side, not a limitation of our backfill. Second: usage. Only 139 of those 11,506 listings — 1.21% — show any usage this month, against 2.65% (327 of 12,343) everywhere else. The largest host in the index underperforms the rest of the catalog on both signals: no observable price coverage, and less than half the usage rate of non-Orbis listings.
Is this actually bad news?
It would be easy to read a 1.95%usage rate as a verdict on x402. We think that’s worth questioning rather than assuming — and early internet history suggests one reason why.
In the first years of the World Wide Web, the overwhelming majority of pages that went online were personal experiments — people testing what a homepage could even be, not building something meant to last. Very little of what got published in the early web mattered a year later. That didn’t mean the web was a bad idea. It meant the web was young, and in any open, low-friction-to-publish system, most of what gets built early is exploration, not signal. The bar rose gradually, as individuals and organizations kept building — one useful thing at a time.
We can’t tell you today which x402 APIs will matter a year from now. But this issue’s own data hints that the same sifting process may already be underway — a hint, not a conclusion: of 23,849 verified-live APIs, we found 46 — a very small slice — with call volume that grew at least 2x and stayed elevated over multiple weeks, corroborated for most of them by more than one independent usage source. One, a search API for X/Twitter posts, climbed from close to zero weekly calls in late May to an average of roughly 37,000 calls a week over the following seven weeks (a low of 13,700 and a high of 48,300) — confirmed independently by both of our usage-signal sources.
46 out of 23,849is a small number. That’s the point, not a problem: in an early, permissionless ecosystem, most of what gets listed won’t matter, and a small number of things will. Decixa’s job isn’t to make that ratio look better than it is — it’s to help agents find the 46 without sifting through the other 23,803.
What's next
Raw counts — how many APIs are listed, how many calls they get — are easy to game and easy to misread. Both gaps this issue surfaced (a catalog that grew +38.9% while measurable usage stayed under 2%; the single largest host reporting zero prices) are exactly the kind of thing a simple ranking by volume or call count would hide. Starting next issue, Decixa is introducing a composite Decixa Score — combining verification status, usage, uptime, and price stability into a single ranking signal — so agents, and this report, have something harder to game than a listing count.
Methodology and caveats
What counts as verified-live. An API has to pass our quality review, still be reachable, not be excluded from the index, and have actually completed an HTTP 402 payment handshake with a readable price (the same bar used by the live Market Index on decixa.ai).
How we measure “used.” Decixa pulls call-volume data for each API from two independent sources — x402watch and Agentic.Market — and combines them, after adjusting for likely wash trading (self-generated, non-genuine calls), into a single usage score. “Used” means that score is above zero as of the August 11 snapshot. It reflects a trailing rolling window of recent call activity, not a full calendar-month total.
A caveat on the 1.95% usage figure. The wash-trading adjustment above is itself an estimate, and we corrected it for one of our two sources (Agentic.Market) on the same day as this data pull: the previous version assumed APIs with a single payer were up to 90% fake traffic by default, but when we checked that assumption against real data, actual wash activity on single-payer APIs was closer to 0.12% — off by more than two orders of magnitude. Because our usage score updates on a 7-day rolling basis, it takes up to a week for that correction to fully flow through. That means the 1.95% figure in this issue is a conservative floor: we expect it to move up, not down, over the following week as the corrected values roll in.
A caveat on the 27.5% price-coverage figure. This is not a market change. The pricing pipeline that populates per-endpoint price data had been effectively broken for most of the first half of the year; this month’s 27.5%reflects our measurement catching up, not a sudden wave of providers publishing prices. Orbis’s 0% coverage is a separate, more specific finding: as described in Section 03, its proxy links appear to have gone dead over time, which is a reliability issue on Orbis’s side rather than a gap in how we measure.
On June. There is no June 2026 issue. The team spent that month hardening the verification and trust-layer pipeline this report depends on — this issue, and its usage-rate measurement in particular, is a direct product of that work.
Search the verified-live pool
All 23,849 endpoints in this report are searchable through Decixa. Pass natural-language intent, get back ranked endpoints with cost, latency, and verification metadata.