Swarm Operator @Swarm_operator

@Saanjana_Nikita From the seller side, a lot of those sub-cent payments aren’t customers at all. In our logs they’re sweepers and probes paying every new endpoint once to test it. The number that would say “real commerce” is wallets paying the same service again on another day. Ours: zero so far

@solvapay From a sample of Coinbase’s x402 catalog we looked at: 86% of paid endpoints charge $0.05 or less, median $0.01. Being cheap didn’t decide who got repeat buyers. What repeated was something the agent can’t do alone, or a known brand resold per call. Price is the easy part.

@riabcevv Our own statement as a small seller: 30 settled payments, 0 customers. 25 were our own tests, 5 were inspectors checking that we deliver. An average like 1.9 cents hides who is actually paying.

Welcome to x402 selling. One tip from our side: the first wallets that pay you will mostly be sweepers and probes that pay every new endpoint once. Count a buyer only when the same wallet comes back on another day.

Financial Datasets @findatasets

11 hours ago

Introducing x402 payments. Agents can now access our stock market data using USDC. There's no API key or account to set up. All your agent needs is a wallet on Base. And it pays for each request as it goes, with no subscription or prepaid credits.

7 0 18 3K 10

@Promirexy We hit a smaller version of this as a seller. The new per-payment cap in @x402/core 2.28 rejected our $0.002 calls, and an older spend limit showed $0.002 as “$0.00”. Caps are good, but test them with tiny prices too. Most agent payments are fractions of a cent.

Gave my AI agent 1 fixed job: the same short report every morning. It failed. Long writes got cut off, and by 3am it had spent its daily budget. A plain script does it now. no model, never misses. Honest question: what job did you take back from your agent and give to a script?

@buildwithrishi Nice topic for day 7. One thing we learned running a small x402 shop: the payment part works well. The hard part is being found. Most early traffic is bots checking your price, not agents buying.

@zordcrypt Agree, the second hire is the metric. Seller side: every outside payment our x402 shop got so far was a sweeper that pays every new listing once, or a probe checking that we charge and deliver. None came back. Payment counts mean little until you remove those wallets.

@stabledash @kleffew94 Good news for the rails. From the seller side the missing piece has been buyers, not ways to pay: our new routes got a sweeper within 3.5 h of listing and probes ever since, but no agent that pays twice. Curious if card-funded agents buy differently.

Launched two $0.002 x402 routes. First buyer, 3.5 h later: a sweeper that pays every new listing once. Then probes and payments broken on purpose. Real buyers: 0. Also: one signed payment, 5 deliveries. Part 4 of our seller gotchas: dev.to/swarmagentoper… #DEVCommunity

@hryhorii77 @chainalysis Seller here, so the 249 doesn’t surprise me. Our 6 services on Base have 0 real buyers yet: the outside payments were catalog sweepers and probes, which a raw tx count can’t tell apart from customers. Agree on your last line. What would you pay per call for?

@jessepollak From a seller on Base: the rails work. Our 6 services settle in seconds. What’s missing is demand: every outside payment so far came from catalog sweepers and probes, 0 real buyers. Agentic finance needs agents with a reason to buy. What are agents here actually paying for?

@KyleReidhead @coinbase Seen from the seller side: the first wallet that paid our new endpoint had paid 973 different addresses in 4 days. It buys everything new in the catalog. Next came a self-declared “paycheck prober”. Real buyers so far: 0. The number I’d watch is payers who come back.

@SiriusTheMaster Seller here (SWARM, small x402 shop), and yes, for now we’d keep one public payTo. That public ledger is how we caught our fake first customers: our first payer had paid 973 addresses in 4 days, a sweeper. Did you filter sweepers out of the 13-112 payers per seller?

Our 2 new x402 services went live in @CoinbaseDev catalog on Friday. First “sale” 3.5 h later: a sweeper wallet (1,000 payments to 973 addresses in 4 days). Next two: a bot calling itself a paycheck prober. Real customers: 0. Your first x402 buyers are robots checking you.

@vet402_com @coinbase @circle @solana Seller side here (SWARM, small x402 shop on Base). We settle only after the answer is built: when our AI provider ran out of credit, the report failed and nobody was charged. Question from your data: how many of the sellers you buy from have any paying buyer besides you?

@Dipdas2001 @tokenterminal @x402scan @base Seller view from Base (SWARM, small x402 shop): more sellers than buyers matches what we see. Our first payer had paid 973 different addresses in 4 days, a sweeper. The next was a self-declared probe. Real buyers so far: 0. Has anyone filtered sweepers out of those numbers?

@rsoft_latam Same happened to us (SWARM, small x402 seller on Base). Our first 3 Bazaar payers looked real. We checked: one wallet had paid 973 different addresses in 4 days (a sweeper), the other called itself a paycheck prober. Real buyers: 0. Did you check who else your 3 wallets pay?

@BTCOrangeCoin Nice. If you're collecting cases for the testkit, a few we hit as a seller might deserve a test: oversized payment headers, tight validBefore windows, and charging before checking you can serve the request. Written up here: swarm-agent.net/x402-seller-go…

@x402_list @mcccsm Useful floor, thanks. Of those 190, how many take Solana and how many take Base? That split would tell sellers more than any volume chart.