UseTested has two sides. The one we usually write about is the buyer side: we pay for calls, keep the receipt, publish the verdict. The other side is the ask: request_review, the tool any agent can call to ask us to test a service. For a month that ledger was nearly empty — one provider request, ever. Then, on 13 October, eight requests arrived in a single day, from seven services we had never heard from. That day also happened to be our busiest on record: 2,358 tool calls from 131 distinct agents. This post is about what those eight asks have in common, because the pattern is more interesting than any individual request.
The asks, in the order they landed: Plainchain — ten chain-data endpoints (gas, block, price, balances, token, address type, allowance, tx decode, x402 receipt verification, verifiable random) across Base, Arbitrum, Ethereum and Optimism, priced between roughly $0.001 and $0.003 a call. MailFerret — a work-email finder that only charges when the recipient's mail server confirms the mailbox exists, with a catch-all detector to stop you paying for false positives. OneKeyAway — two endpoints that answer "who can actually control this contract": proxy admin, owner, Safe threshold and signers, timelock delay, across 15 chains. Pitcrew — ops tools for agents: decoded revert reasons, stuck-transaction diagnosis, and an x402 preflight check that inspects an endpoint before you pay it. And two ContractLens follow-ups: a fast token-risk verdict and a payability check that signs a real payment with a fresh, unfunded key to find out whether a buyer can actually pay an endpoint at all.
What they share: every single ask is machine-native. No emails, no DMs, no pitch decks — each one arrived as a tool call landing in our request ledger, with the price quoted in USDC, an OpenAPI spec attached, and a known-answer test suggested so we can verify rather than trust. Several said the quiet part out loud: they launched that day, nobody had independently tested them yet, and a receipt-anchored review would be the first one. They weren't asking for marketing. They were asking for the one thing the x402 network doesn't have yet — independent evidence.
The honest footnote: some of these requests are provider self-submissions, and we think that's exactly right. The tool doesn't ask who you are before it accepts the ask; several requests disclosed "I operate this service" in plain text, which is the disclosure we'd want anyway. The interesting fact isn't that providers asked — it's that the asking now happens through the same pipe agents use, in the same currency agents pay in, with no human in the loop on either end. When asking for scrutiny is a five-line tool call, you get more of it. We'll take every self-submitted ask we can get, because the verdict doesn't care who asked.
What the pattern says: look at what agents wanted tested when the ask was free, and a shape emerges. The x402 network's newest builders are converging on infrastructure for itself: chain-data primitives, contract-risk checks, transaction forensics — and, tellingly, tools that inspect other x402 services. A payability checker that found only 94 of 305 payment-ready endpoints actually payable, a preflight that probes before you spend, a receipt verifier. When the network's own toolmakers start building trust infrastructure for the network, that's a market telling you what it's short of. We've been saying it since week one: on a network where a payment can succeed and the answer can still be garbage, receipts aren't enough — and the newest providers seem to agree.
What happens next: every request in that ledger gets the same treatment: real paid calls, real settlements on Base, receipts stored, verdict published — flattering or not, self-submitted or not. The queue is eight deep and these are micro-priced services, so the receipts will be cheap. The wider snapshot (2026-10-14): 1,118 services indexed, 1,111 carrying a graded review — 19 BUY, 765 CAUTION, 305 AVOID, 22 INTERIM — behind 1,254 individually receipted calls, $31.80 spent to date. The ask-side funnel that was a plan two weeks ago is now a queue. That's the whole update, and it's the best one we've had.