Part 1 and part 2 compared the prices printed in x402 listings. A listed price says what a seller hopes to charge. It does not say whether anyone pays it.
The Coinbase CDP Bazaar also reports, per listing, how many distinct wallets paid it and how many paid calls it served in the last 30 days. Our price pages now have a "Which listings agents actually pay" section built from those two numbers. This part pulls out seven categories where the pattern is odd. All figures are from the 2026-10-09 snapshot.
We sell in this market (Tanod, a pay-per-call API and MCP server), so we are one of the rows. Our numbers are in each table, and they are small.
Two limits first. The counts are as the Bazaar reports them, not verified by us, and a seller's own test wallets are included. A "wallet" is not a customer, and a high call count from few wallets can be one script in a loop.
Web search: the known engines win, near the median
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| stableenrich.dev/api/exa/search (Exa, resold) | 0.01 | 176 | 26,032 |
| api.exa.ai/search (Exa, direct) | 0.007 | 88 | 5,080 |
| blockrun.ai/api/v1/exa/search (Exa, resold) | 0.011 | 44 | 1,388 |
| tanod.dev/v1/search (Mojeek index) | 0.012 | 1 | 1 |
The most-paid listings resell or run a known engine (Exa, Tavily, You.com, Firecrawl), and they charge around the median, not the least. Tanod's search is USD 0.012, with one wallet and one call.
Page: web search
EVM chain reads: one host takes the top eight
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| api.onesource.io/api/chain/erc20-balance | 0.003 | 678 | 684 |
| api.onesource.io/api/chain/block-number | 0.001 | 631 | 721 |
| api.onesource.io/api/chain/live-balance | 0.003 | 607 | 610 |
| tanod.dev/v1/chain/balance | 0.002 | 2 | 2 |
All eight most-paid listings are on api.onesource.io: 437 to 678 paying wallets each and 438 to 721 paid calls, at USD 0.001 to 0.005. The most-paid listing from any other host we kept has 39 wallets. Wallets per call is close to 1, so these look like many wallets paying once each, not a few wallets repeating.
Page: EVM chain reads
Exchange rates: 150 wallets at USD 0.005
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| tick.hugen.tokyo/tick/latest (best bid and ask for an FX pair) | 0.005 | 150 | 213 |
| tick.hugen.tokyo/tick/all | 0.005 | 136 | 164 |
| api.x402node.dev/forex/convert/cny/usd/cny-usd | 0.004 | 24 | 113 |
| tanod.dev/v1/fx (fiat rates) | 0.001 | 2 | 2 |
Prices in this table run from USD 0.001 to 0.008. The two most-paid listings are at USD 0.005, five times Tanod's price, and neither is a convert-this-amount call: they return bid and ask quotes. Our fiat-rate route had 2 wallets and 2 calls.
Page: exchange rates
Sentiment analysis: a USD 10 listing with 42 wallets
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| www.api-xpay.com/x402/sentiment (AI sentiment of text) | 10 | 42 | 48 |
| x402.asterpay.io/v1/ai/sentiment | 0.004 | 7 | 10 |
| api.stelardigital.com/sentiment | 0.005 | 7 | 7 |
| tanod.dev/v1/text/sentiment | 0.001 | 1 | 1 |
Listed prices here run from USD 0.002 to USD 10. The USD 10 listing has 42 wallets and 48 calls, six times the wallets of the next listing, which charges USD 0.004. We do not know why. Its live payment request does ask 10 USDC per call (we checked the 402 response; we did not pay). The data cannot say whether those wallets are real buyers or test traffic. Treat it as a number to be suspicious of, not as evidence that agents pay USD 10 for sentiment.
Page: sentiment analysis
Speech to text: USD 0.40 per call, 17 wallets
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| ausca.com/v1/transcribe-media | 0.40 | 17 | 54 |
| agent402.tools/api/transcribe (OpenAI) | 0.03 | 4 | 40 |
| api.x-402.online/v1/transcribe (Whisper large-v3) | 0.02 | 3 | 20 |
| tanod.dev/v1/audio/transcribe | 0.01 | 0 | 0 |
Only five speech-to-text listings we kept had 3 or more paying wallets. The leader charges 13 to 20 times the others and has four times the wallets of any of them. Tanod's route is listed at USD 0.01 and does not appear in the snapshot, so it shows 0 and 0.
Page: speech to text
Web page to Markdown: the expensive host and the 12,868-call outlier
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| api.x402node.dev/web/batch | 0.02 | 21 | 101 |
| api.x402node.dev/web/read | 0.02 | 16 | 61 |
| api.clawfetch.ai/fetch | 0.001 | 14 | 251 |
| win.oneshotagent.com/v1/tools/web-read | 0.002 | 8 | 12,868 |
| tanod.dev/v1/render | 0.005 | 2 | 2 |
The three most-paid listings come from one host at USD 0.02 per call. Its listings mention residential IPs and CAPTCHA handling. Cheaper listings at USD 0.001 to 0.005 follow with 8 to 14 wallets. The oneshotagent.com listing has the most calls in the table, 12,868 from 8 wallets, which is about 1,600 calls per wallet. Cheapest is not what gets the most wallets here, and most wallets is not what gets the most calls.
Page: web page to Markdown
Email verification: 7,206 calls from 11 wallets
| Listing | Price (USD) | Wallets | Calls |
|---|---|---|---|
| win.oneshotagent.com/v1/tools/verify/email | 0.001 | 11 | 7,206 |
| email-verify-api-production-d262.up.railway.app/verify | 0.001 | 7 | 8 |
| hunter.x402.paywithlocus.com/hunter/email-verifier | 0.008 | 6 | 11 |
| tanod.dev/v1/email/verify (DNS only, no SMTP probe) | 0.002 | 2 | 2 |
The leader has 11 wallets and 7,206 calls at USD 0.001, while the next listings have 7 or fewer wallets and single-digit or low-double-digit calls. The same host shows up in the Markdown table above. One platform with heavy repeat use can dominate a call ranking. The pages rank by paying wallets first and calls second for this reason.
Page: email verification
What the pattern says, and what it does not
- Concentration is common. In EVM reads one host holds the top eight rows. In Markdown one host holds the top three. In search the top three rows are all Exa, two of them resold.
- Cheapest is not what gets paid. The leader is USD 0.40 in speech to text, USD 10 in sentiment and USD 0.005 in FX.
- Demand is thin outside a few categories. On other pages the top listing has 5 to 9 wallets: for example DNS lookup (8 wallets, 12 calls), PDF tools (5 wallets) and language detection (5 wallets).
- Our own routes, in the categories above, sit at 0 to 2 wallets and 0 to 2 calls. Those are a handful of buyers in 30 days, and a seller's own test calls can be among them.
None of this shows quality, only that a wallet paid. It is one snapshot, and the Bazaar is one place where x402 services are listed.
The dataset, and how an agent can query it
Every listing's counts, not only the top rows, are in a CC BY 4.0 dataset: tanod/x402-bazaar-endpoint-demand on Hugging Face. One row per endpoint: URL, host, name, category, price in USD, network, pay-to address, unique payers over 30 days, calls over 30 days, and last-called time. The 2026-10-09 snapshot has 34,062 rows.
The market-wide write-up is at x402 Bazaar agent demand data.
An agent can query the Tanod index of x402 endpoints and MCP servers through the Tanod MCP server at https://tanod.dev/mcp. It needs no API key or account, and most tools have a small free daily allowance per IP. After that, calls cost USDC over x402. Setup for 20 MCP clients is on tanod.dev/connect; the index tools are query_agent_index, bulk_agent_index and export_agent_index.
If you sell an x402 endpoint, the useful number in this data is wallets, not price. If you buy, check wallets before you trust a cheap listing.
Top comments (0)