The x402 Market, Counted: 14,766 Listings, 12,741 Calls a Day, One Winner

Reported coverage of x402 quotes aggregate volumes and founding members. Nobody has counted what is actually listed and actually called. The discovery index answers without authentication, so Salesmart S.r.l., which publishes Sinapsi, enumerated all of it on 3 August 2026. The result is a market with an enormous shop window, almost no customers, and one dominant product.

What we counted

| Measure | Value | |---|---| | Listings indexed | 14,766 | | Distinct sellers (payTo) | 1,252 | | Distinct hosts | 1,520 | | Calls in the last 30 days, whole index | 382,225 (~12,741/day) | | Median calls per listing (30d) | 2 | | 90th percentile | 9 | | Median price | $0.010 | | Implied gross revenue, 30d | ~$63,703 (~$775k/yr annualised) |

Every listing carries a quality object with l30DaysTotalCalls, l30DaysUniquePayers and lastCalledAt, which is what makes this countable at all.

The shape: a supply bubble with one product

98.9% of listings have at least one call in 30 days, which sounds healthy until you see the median: two calls a month. The distribution is not skewed, it is degenerate.

- Top 1 listing = 31.6% of all calls in the index - Top 10 = 62.8% - Top 100 = 77.2%

The leaders are almost all the same product — search:

| Calls (30d) | Unique payers | Price | Resource | |---|---|---|---| | 120,928 | 49 | $0.006 | x402.twit.sh/tweets/search | | 60,698 | 424 | $0.010 | x402.tavily.com/search | | 23,680 | 1 | $0.010 | agents.chain.link/v1/operations/… | | 10,702 | 266 | $0.010 | stableenrich.dev/api/exa/search | | 4,932 | 87 | $0.007 | api.exa.ai/search |

Two observations that matter more than the ranking. First, what agents pay for today is retrieval of information they do not have — web and social search, then on-chain data. That is the demand, and it is narrow. Second, several high-volume entries have one unique payer: those are not markets, they are one integration billing itself through a public protocol.

What is listed, by kind

Classifying by URL keyword (crude heuristic, declared as such): crypto/on-chain 12.1%, web-data and scraping 9.3%, market data 9.0%, company and legal data 6.8%, AI/LLM 2.9%, identity/KYC 2.7%, geo/weather 2.7%, social 0.9%, unclassified 53.7%. Supply is dominated by generic data plumbing.

MCP on x402 barely exists

Of 14,766 listings, 68 involve MCP — 0.46%. Together they took 389 calls in 30 days, about 13 a day for the entire paid-MCP ecosystem. The named ones are DeFi actions (mcp.yield.xyz), attestation (attester.dev), Brazilian company dossiers (api.brazilmcp.dev), wallet history.

This is the single most decision-relevant number in the snapshot, and it cuts both ways: the paid-MCP market is empty, so being early costs nothing — and it is empty because nobody has shown that agents will pay per tool call rather than per HTTP fetch (Paid MCP Tools: 65 Listings, 12.5 Calls a Day, and One Site Making Seven Times the Whole Market).

Discovery is richer than the marketing suggests, and still cannot rank

We were wrong on first inspection and correct the record here: descriptions are not missing (absent in only 1.1%) and are informative ("Latest Ethereum block height — current chain tip via eth_blockNumber"). The extensions.bazaar object carries the input schema (method, query params, body) and an example output. An agent can therefore work out *how* to call a service without documentation.

What it still cannot do is decide *whether the answer will be any good*. The quality object measures demand — calls and unique payers — not outcome. A resource called often by one payer scores like a resource called often by four hundred, and nothing anywhere records whether the response held up. The scarce good in this market is not payment, which works; it is knowing what is worth buying (What Is Actually Priced: The One-Cent Ceiling, and the Market That Escaped It).

Version migration is mostly done

13,037 listings declare x402Version: 2 against 1,729 on v1 (88% / 12%), and network naming shows the same split — CAIP-2 identifiers (eip155:8453) alongside 1,935 accepts that still declare a plain legacy string (1,785 of them base, then polygon 112, solana 32, base-sepolia 6), spread across 1,734 listings. Anyone implementing today must accept both. Networks by frequency: Base, Solana, Polygon, Arbitrum, XRPL. Assets: USDC dominant, with EURC, Global Dollar and XRP present.

Honest limits

Scope. This is one facilitator's discovery index. Traffic settled outside it, or through other facilitators, is invisible here — which is likely part of why reported aggregates (~75M transactions, ~$24M in a month) are orders of magnitude above what we count. The two numbers measure different things and should never be quoted side by side without saying so — and that reported aggregate is itself weaker than it looks: x402: HTTP 402 Finally Gets a Job, at 32 Cents a Transaction shows by double fetch that the press figure reproduces the protocol's own public counter rather than an independent count.

The traffic counters are the operator's, not ours. We collected the snapshot, but l30DaysTotalCalls and l30DaysUniquePayers are self-reported by the index operator; treat them as the most flattering available figure, the same caveat used in The Agent Labour Market, Probed: A Complete Bounty Board With Four Customers, which reads the same enumeration.

No growth reading from one snapshot. lastUpdated is a listing refresh timestamp, not a creation date: 88% of entries carry a July 2026 date and 45 distinct days appear in total, so it cannot date market entry. Anyone inferring growth from it is reading noise. The correct method is repeated snapshots, which is why the script writes a dated file and never overwrites — the same rule as our forecasting time series. The second snapshot is the first one that can say anything about growth.

Revenue is implied, not observed. We multiplied 30-day calls by the price of each listing's first non-junk accept (the rule is written into prezzo_listing() in the script, so the $63,703 is recomputable to the cent). Discounts, failed settlements and free tiers are invisible to us.