Ask an agent what it wants from a service provider and the honest answer isn't the one you'd expect. Not the cheapest price, not the richest output, not the best latency on a good day. It's this: will the endpoint answer when I call it? An agent planning a multi-step task doesn't need a provider that was excellent last month. It needs one that is alive right now — because a dead endpoint doesn't just fail one call, it poisons every step built on top of it.
Liveness is the quality nobody advertises. Providers publish pricing pages, feature lists, changelogs. What they don't publish is uptime you can verify — because on the x402 network, liveness decays fast. Our numbers say so. On 24 October we swept every service we index — 1,206 liveness checks in about thirteen minutes. 1,187 were alive. 19 were dead: domains gone, endpoints that no longer answer, services that took payment last week and return nothing today. Nobody announced those deaths. The providers certainly didn't. The only way you learn is by calling — and if you're an agent, calling is the expensive way to find out.
So liveness is now a first-class field on every review we publish. A review doesn't just carry a verdict, latency and error rate anymore — it carries an alive status and the timestamp of the last check. And the two are wired together: a BUY from a dead endpoint renders as CAUTION at read time, because a great review of a service that no longer exists is worse than no review at all. Liveness also gates our bulk routes — our top-10 bundle excludes anything UNTESTED or DEAD outright. The verdict answers "was it good when we measured it?" The alive status answers "is it still there?" Agents need both, and until recently we only published one of them.
The obvious problem: we can't be everywhere at once. Our catalogue now lists 19,243 endpoints, and scanning that many constantly is simply not possible. Our sweep runs once every two weeks, which means it can never guarantee a status — and so, by design, liveness decays constantly after the first day. A check is definitive on the day it's made; from the next morning it's evidence, not fact. A service that died on Tuesday still reads as alive until the next pass finds it. And from the provider's side it's worse: a provider who fixes an outage, or a brand-new service that just launched, waits up to two weeks for our sweep to notice. There's an asymmetry — providers know their liveness in real time and we learn it on a schedule.
So we've closed the gap from the provider's side: POST /v1/pulse. Providers self-report their service as LIVE, and the report is receipt-anchored — the x402 settlement itself is stored as the claim's proof, priced at half a cent, one report per service per 24 hours. The rules are deliberately narrow. A self-report can only ever claim LIVE — you cannot pulse another provider's service DOWN, and you cannot un-publish your own failures, because DOWN verdicts come from our sweep and nothing else. A pulse doesn't touch verdicts, error rates or review text at all. It does exactly one thing: refreshes the alive status and its timestamp.
The honest truth: no liveness proof stays fresh — whoever makes it. A proof decays from the moment it's stamped. Our sweep check, a provider's self-report, any third party's probe: each is definitive on its day and evidence ever after. The sweep runs once every two weeks; a pulse lasts a day at full strength and fades across two weeks. That's why Pulse is the definitive proof — not because providers are more honest than our probes, but because the provider is the one party who can renew the proof the moment it goes stale, at any moment, without waiting for anyone's schedule. Once a day is the lowest frequency we suggest: a service pulsing daily is provably alive, continuously, on the record. The half-cent price isn't the point — the receipt is. A free claim is noise; a paid claim has an address behind it.
Pulse also feeds the rankings. Liveness is a ranked signal, not just a badge: our top-10 lists weight liveness-weighted reliability, so a service whose proof is fresh — theirs or ours — outranks an equal service whose evidence has gone stale. Providers who pulse keep their standing; providers who let their proof rot slide, whatever their verdict. In the first days since the endpoint went live, 51 pulses have come in from 17 distinct providers, every settlement on-chain — the providers who understand how this works are already on the right side of the decay curve. Next post, we walk through the top-10 listings themselves: the services that survive every gate we have, verdict and liveness together.
The wider snapshot (remote D1): 1,252 services carry a graded review — 29 BUY, 723 CAUTION, 321 AVOID, 179 INTERIM — behind 1,696 individually receipted calls, every one a real on-chain settlement with its transaction hash on record. Of the reviewed catalogue, 1,187 are verified alive and 19 verified dead; 17 of the living ones have providers actively pulsing.