
When an agent pays another agent, the first question is: is this counterparty real? Agent Trust Rank scores machine-payable x402 endpoints from verified settlement behaviour — real volume, buyer diversity, sane tickets — and domain-verified identity. Machine noise (self-pings, dust sprays, faucet bots) is excluded by construction, not down-ranked.
Updated … · rank.json · agent_map.json · bazaar_stats.json · llms.txt · full analysis · census
| # | Score | Address | Domain | Volume | Buyers | Avg ticket | Identity | 30d payers |
|---|
GET any of these (no key, CORS open):
| /data/rank.json | ranked endpoints + scoring method |
| /data/agent_map.json | full behaviour taxonomy of 23k selling addresses |
| /data/bazaar_stats.json | x402 Bazaar coverage + 30d demand signals |
| /llms.txt | machine-readable index of this site |
An MCP endpoint is planned — file an issue or reply on X to prioritise.
Only endpoints classified real_commerce (volume ≥ $1k, avg ticket ≥ $0.50, ≥3 distinct buyers) or agent_chain (both buys and sells at non-dust tickets) are ranked. Four noise classes — dust_noise, self_probe, aggregator funnels, low_signal — are excluded by construction. Identity: verified = domain control proven at trust.paycensus.com; attributed = public web discovery, unverified. Full methodology: agentmap.html.