# How many x402 APIs have paying customers? October 2026 data

> Of 42,096 paid x402 APIs with Bazaar usage data, 55% had one paying wallet in 30 days and 0.9% had ten or more. What that means for sellers and agents.

Published 2026-10-16. Source: https://toolvet.app/blog/x402-apis-paying-wallets

**Short answer:** as of 16 October 2026, ToolVet lists 45,805 public paid x402 APIs. Of the 42,096 with usage data from the x402 Bazaar, 23,290 (55.3%) had exactly one paying wallet in the last 30 days, 18,403 (43.7%) had two to nine, and 397 (0.9%) had ten or more. One wallet is very often the seller paying their own endpoint, so fewer than half of the listed x402 APIs show any sign of a second buyer, and fewer than one in a hundred has what you could call a customer base.

Some background: [x402](https://www.x402.org) lets a client, usually an AI agent, pay for an HTTP request in stablecoins, and the supply side has grown fast. [Chainalysis counted](https://www.chainalysis.com/blog/north-america-crypto-adoption-2026/) more than 138 million x402 transactions on Base in the twelve months to June 2026, and over the 16 months to September 2026 found more than 87% of payments between a hundredth of a cent and a dime. Those are transactions. This post counts something else: for each paid API, how many distinct wallets paid it. That is the number that tells you whether anyone besides the seller is using a listing.

## Where the numbers come from

ToolVet checks every listed x402 endpoint on a schedule without paying it, and alongside its own checks it reads the usage data the [x402 Bazaar](https://docs.cdp.coinbase.com/x402/bazaar) publishes for each resource: call count and unique payer count over the last 30 days. The Bazaar is the catalog Coinbase's facilitator builds from the payments it settles, and it ranks by the same signal among others: its search docs say quality "considers recent call volume and unique payers", and its [seller docs](https://docs.cdp.coinbase.com/x402/seller/get-discovered) list buyer reach, transaction volume and recency over a rolling 30-day window.

Every day ToolVet publishes one snapshot of these counts as the `payers` series on the [open data page](https://toolvet.app/data), under CC BY 4.0. The figures in this post are the 16 October 2026 snapshot, the first day of that series. Two limits to keep in mind:

- **Only the Bazaar's view.** 3,709 of the 45,805 listed x402 APIs have no Bazaar usage data, usually because their payments settle through a different facilitator, which keeps a catalog of its own. We [compared the facilitators](https://toolvet.app/blog/x402-facilitators-compared) earlier. These APIs are left out of the percentages below, not counted as zero.
- **Wallets, not people.** A wallet is a distinct payer. One person or one agent framework can run many wallets, and one wallet can serve many agents.

## The distribution

Paying wallets per paid x402 API over 30 days, 16 October 2026, share of the 42,096 APIs with usage data:

| Paying wallets in 30 days | x402 APIs | Share |
|---|---|---|
| None | 6 | under 0.1% |
| Exactly one | 23,290 | 55.3% |
| Two to nine | 18,403 | 43.7% |
| Ten or more | 397 | 0.9% |

The "none" row is nearly empty for a structural reason, not because everything gets bought: the Bazaar [removes resources](https://docs.cdp.coinbase.com/x402/seller/get-discovered) that go 30 days without a settlement from the catalog and its search results, so an endpoint nobody pays leaves the data instead of being counted.

## Why one wallet usually means no customer

A catalog entry is created when a facilitator settles a payment for the endpoint, not when it is deployed. In practice that first payment comes from the seller: pay your own endpoint once from a separate wallet and it shows up. We described that flow in [why an endpoint isn't in the Bazaar](https://toolvet.app/blog/x402-endpoint-not-in-bazaar). Add the 30-day pruning rule, and a seller who wants to stay listed has a reason to pay themselves again every few weeks.

So the one-wallet band is best read as "listed, and kept alive by its owner". It can include endpoints with one real buyer, but the data can't tell the two apart, and ToolVet's score treats both the same: the real-usage component counts wallets on a log scale, so a handful of self-payments barely moves it. The [methodology page](https://toolvet.app/methodology) has the weights.

## An independent crawl found the same shape

On 13 October 2026, a developer [crawled every resource in the Bazaar](https://dev.to/lowkeybuilds/i-crawled-all-31620-x402-bazaar-listings-most-have-one-payer-and-boring-chain-reads-win-5c73) through the public discovery API: 31,620 unique resource URLs across 2,280 hosts. Their buckets are finer than ours and the bases differ: they de-duplicated to URLs on one day, ToolVet keeps one listing per HTTP method and URL, and the catalog grows fast (between the first snapshot of 16 October and the [stats page](https://toolvet.app/stats) at 20:55 UTC the same day, ToolVet's x402 count went from 45,805 to 47,210). The picture still matches:

- 45.8% of resources had exactly one payer, 30.9% had two, and about 77% had two or fewer; 84 (0.3%) had none.
- 1.3% had ten or more payers, and 0.3% had fifty or more.
- Of 242 resources with 20 or more payers, 101 averaged about one call per payer, which looks more like test or bot wallets than customers. The author also notes the counts may include payloads that were verified but never settled, so they are an upper bound on paid demand.

The same post found that endpoints priced at or below $0.001 per call had the highest mean number of payers, and that the $0.005 to $0.01 band had the smallest share of endpoints with five or more payers. That is one crawl on one day, so treat it as a hint rather than a pricing rule.

## What the paying wallets are buying

Chainalysis groups x402 spending into a few categories: information services such as code review, financial intelligence, and data and search APIs; model inference, which it calls one of the largest categories, with AI agents paying other AI agents; agent frameworks and tooling; and gaming, token issuance and gambling. It also notes that early x402 activity was dominated by fads that attracted wallets and then faded, and says the more enduring activity is concentrated in the information sector. ToolVet's own [browse page](https://toolvet.app/browse) shows how the supply splits across categories at any time.

## If you sell an x402 API

- **Expect to be the only payer for a while.** More than half of the listings with usage data have one wallet. The useful question is what makes the second wallet show up.
- **Write for the agent that chooses.** The Bazaar's [seller docs](https://docs.cdp.coinbase.com/x402/seller/get-discovered) say bare endpoint names and placeholder text score zero on metadata quality. Describe what the endpoint returns, for which input, and when to use it, and declare the full input and output schema with a real example.
- **Be reachable.** Every catalog and directory probes your endpoint without paying. If the unpaid request gets a 400 instead of a 402, agents never see your price. The free [x402 checker](https://toolvet.app/check) shows what they see.
- **Get into more than one catalog.** Each facilitator catalogs only what it settles, and directories that crawl, such as ToolVet, list you regardless. Pay your endpoint once through each facilitator you want to be found in.
- **Price for the fee, not the catalog.** At sub-cent prices the facilitator's per-settlement fee is the bigger decision. The [facilitator comparison](https://toolvet.app/blog/x402-facilitators-compared) has the arithmetic.

## If you build an agent that pays

- **"Listed" is not "used".** A catalog full of one-wallet endpoints is a catalog full of untested endpoints. When two listings look alike, the one with distinct buyers has had other agents' bugs run through it.
- **Use usage as one signal, not the signal.** ToolVet's search weights text relevance at 60%, trust score at 30% and real usage at 10%, and the trust score itself gives 15 of 100 points to paying agents. Uptime alone weighs twice as much, and uptime, a valid challenge and honest pricing together weigh 55, because a popular endpoint that is down right now still wastes your call.
- **Check before you pay.** The `preflight_check` tool on the [ToolVet MCP server](https://toolvet.app/agents) and the free [checker](https://toolvet.app/check) send the unpaid request, read the payment challenge and compare the price and pay-to address with what ToolVet has seen before.

## Frequently asked questions

### How many x402 APIs are there?

ToolVet listed 45,805 public paid x402 APIs on 16 October 2026, next to 28,233 remote MCP servers. An independent crawl of the Coinbase Bazaar on 13 October 2026 found 31,620 unique resource URLs; the bases differ because ToolVet keeps one listing per HTTP method and URL, also lists endpoints that settle through other facilitators, and the catalog currently grows by more than a thousand entries a day.

### Does one paying wallet mean the API has a customer?

Usually not. A Bazaar entry is created when the first payment settles, that first payment is typically the seller's own, and the catalog prunes resources with no settlement for 30 days, so sellers have a reason to keep paying themselves. The data can't tell a single real buyer from a self-payment, which is why ToolVet's score barely counts one wallet.

### Where does the usage data come from?

From the x402 Bazaar, which publishes unique paying wallets over 30 days for each resource. ToolVet reads it alongside its own checks and publishes a daily aggregate as the `payers` series on the [data page](https://toolvet.app/data). APIs that settle through other facilitators have no Bazaar usage data and are excluded from the percentages.

### Can I download the numbers?

Yes. The daily `payers`, `listings`, `uptime` and `security` series are on the [data page](https://toolvet.app/data) as CSV and JSON, under CC BY 4.0, with one row per UTC day. No listing or provider is named in them.

The counts above are as of 16 October 2026 and move every day; the [stats page](https://toolvet.app/stats) has today's. [How we check and score](https://toolvet.app/methodology) explains the rest.