Every agent-payment endpoint claims to work. We scored them all — then spent real USDC to find out which ones actually deliver. This is what the x402 economy looks like when you stop trusting listings.
The other ~25,350 indexed endpoints are open — no payment gate, nothing to settle. We probe those free. The money story is the 15,264 that charge. Solana sweep: 72 doors paid, $0.84 spent, receipts on Solscan. 4,080 indexed.
Free probes can't tell you if an endpoint delivers. A 402 can be perfectly formed, the payTo can match, latency green — and the moment money moves, you get nothing back. So we don't just probe. We buy.
Automated discovery across CDP Bazaar, x402scan, awesome-x402, and direct probes. Every gated endpoint enters the index with full metadata.
A Coinbase-custodied agent wallet signs real USDC payments (EIP-3009). Not simulated. Not testnet. Facilitator receipts are the per-call proof — x402 does not mint a separate on-chain tx per listing.
If data returns and it's useful → returned. If payment settles but nothing comes back → demoted automatically. No human judgment involved.
We attempted a real USDC purchase from every gated endpoint in the ecosystem — 15,264 doors knocked, then knocked again in a rehash pass that re-verified every settle. Across all sweeps, 1,703 payments settled; 1,348 proof-backed. 328 returned usable data — 19.26% of settles delivered. 1,539 came back in the Aug-25 full sweep alone; the ledger total is 1,703 once retries and the rehash pass are included. The failures are broken payment paths: the endpoint advertises x402, but checkout fails the moment a real agent arrives with real money. We paid those doors twice and the same doors still failed. "Proof-backed" in the stats bar = settles with a full facilitator receipt or an on-chain-verified unique tx hash.
We knocked on 40,614 doors with real USDC. 1,539 settled. 328 delivered data. We paid those 15,264 doors twice and confirmed it at scale. 19.26% of settles returned data across both chains. The rest took the money — or broke the moment a real agent arrived with a wallet. That gap is the distance between x402's promise and its reality. We publish both sides of it. And we just swept 72 Solana doors too — $0.84, receipts on Solscan.
Five surfaces. One mission: make the x402 economy trustworthy at machine speed.
Every directory has a leaderboard. Ours has a spam filter with receipts. These are listings that looked perfect — until we checked who actually gets the money. This board is the reason our leaderboard only ranks domains a real operator runs.
40,614 endpoints scored for integrity. Every score reproducible from public data. 1,348 proof-backed with receipts.
Continuous payTo drift detection across 40,614 endpoints. 1,985 mismatches in index. Know before your agent pays.
Full-text search across declassified government documents. JFK, UAP, MKUltra, COINTELPRO — the receipts are public.
Legal open-access routing for scientific papers. Aaron's ghost drives this one — paywall bypass through legitimate channels.
Built for agents first: no accounts, no keys, no signup. Your agent pays per request with x402 USDC on Base — free tier on every core route.
Free tier: 25 calls/day. Paid: $0.01–$0.10 per call via x402 on Base. No accounts, no API keys, no signup — your agent pays per request.
RobotsShop is an autonomous shop operated by the Digital Hubris / AnalogHubris stack. Every number on this page is reproducible from public data. No invented metrics, no paid placement, no sponsored rankings.
Scoring is tamper-resistant: payTo must match listing, bodies must be valid, spam gets penalized. We'd rather index 40K honest endpoints than 400K noisy ones.
This page is semantic. llms.txt is current. The API returns clean JSON. Ask your agent to browse — it will understand us perfectly.
The only directory that pays its own listings. When we say an endpoint works, it's because we spent real money and watched it deliver.