Measured September 4, 2026. Every number on this page is read from our live API at build time.
x402 is doing real volume — x402.org reports 75M
transactions and $24M in the last 30 days. This page is not about the protocol. It is about what an
agent finds when it goes shopping: we probed every endpoint listed in
awesome-x402 and coinbase/x402’s README, unpaid, from outside, with no
credentials.
Every endpoint that is still live fails to name a facilitator in its 402. That is the field a buyer reads to learn who can settle the payment. Without it a spec-following client falls back to its own default, which cannot settle to that seller — so the endpoint is healthy, the price is fair, the terms are signable, and the buyer still experiences a payment failure. The seller never hears about it, because nobody files a bug titled “I could not give you money”.
x402.org/protected scores 63 on the same check. This is a
default, not negligence.| network declared | endpoints |
|---|---|
no chain declared | 11 |
eip155:8453 | 6 |
base | 3 |
solana:5eykt4UsFv8P8NJdTREpY1vzqKqZKvdp | 1 |
eip155:42161 | 1 |
An endpoint that declares no chain is unpayable by anyone: you cannot sign a payment without
knowing where it settles. Note also the split between CAIP-2 identifiers
(eip155:8453) and the legacy string base — a strict v2 client
resolving CAIP-2 may not match the legacy form at all.
| endpoint | network | price | score |
|---|---|---|---|
https://langston.click/api/search | solana:5eykt4UsFv8P8NJ | $0.02 | 67 |
https://kerne.fi/x402/reserves | eip155:8453 | $0.25 | 67 |
https://hookproof-kit.vercel.app/api/download | eip155:8453 | $5.0 | 67 |
https://eltociear-tokenguard.hf.space/scan | — | $0.02 | 67 |
https://eltociear-skill-audit.hf.space/audit | — | $0.005 | 67 |
https://eltociear-contract-guard.hf.space/check | — | $0.005 | 67 |
https://x402deploy.vercel.app/api/serp | base | $0.005 | 63 |
https://api.regressionincoming.fyi/v1/bazaar | eip155:8453 | $0.1 | 63 |
https://base-worker-01.j23726919.workers.dev/api/bilingual-bridge | eip155:8453 | $0.03 | 63 |
https://base-intel-api.jakemaxsigal.workers.dev/search | — | $0.015 | 63 |
https://base-worker-01.j23726919.workers.dev/api/asia-intelligence | — | $0.02 | 63 |
https://api.x402oracle.com/v1/trade-check | eip155:8453 | $0.002 | 57 |
https://nft-alpha-x402.fly.dev/api/nft-signal | eip155:8453 | $0.02 | 57 |
https://api.ai.everything.inc/mcp | eip155:42161 | $0.01 | 53 |
https://api.proxyllm.ai/x402/llm-cost-scout | base | $0.005 | 26 |
https://api.ipintel.ai/x402/?ip={ip} | base | $0.001 | 26 |
https://pyrimid.ai/api/mcp | — | — | 0 |
https://traderoute-psi.vercel.app/api/health | — | — | 0 |
https://pyrimid.ai | — | — | 0 |
https://httpay.xyz | — | — | 0 |
https://traderoute-psi.vercel.app | — | — | 0 |
https://monetizeyouragent.fun | — | — | 0 |
| code | count |
|---|---|
ACCEPTS_ASSET_MISSING | 6 |
ACCEPTS_MAXTIMEOUTSECONDS_MISSING | 6 |
ACCEPTS_NETWORK_MISSING | 6 |
ACCEPTS_PAYTO_MISSING | 6 |
ACCEPTS_SCHEME_MISSING | 6 |
AMOUNT_MISSING | 6 |
FACILITATOR_UNNAMED | 22 |
HEADER_ABSENT | 8 |
HEADER_OVER_PROXY_LIMIT | 3 |
LISTING_DESCRIPTION_MISSING | 9 |
All free, no signup. The checker probes from outside with no credentials, so it works on anyone’s endpoint — including ours.
# one endpoint
curl "https://x402-agent-pay.com/facilitator/report?url=https://your.api/paid&method=GET"
# the whole live map
curl "https://x402-agent-pay.com/facilitator/heatmap"
# up to 20 at once, $0.01 USDC on Base
curl -X POST "https://x402-agent-pay.com/facilitator/heatmap/batch" \
-H 'Content-Type: application/json' \
-d '{"urls":["https://your.api/paid"],"method":"GET"}'
There is a human version at /endpoint-report, and agents can call the same checks as MCP tools on agentworld.me/mcp.
An earlier version of this page and an accompanying thread said 13 payable endpoints out of 282. That count came from a filtered candidate list whose per-host results were not retained, so the number could not be checked - by a reader or by us. We re-ran it keeping one inspectable row per host: 747 hosts, 22 payable endpoints, 98 hard-dead links. The central finding did not soften. It got stronger: 22 of 22 payable endpoints still name no facilitator.
This measures the public directories, not the ecosystem — a seller who never listed
in a README is not counted here. Endpoints that blocked or timed out on our probe are returned as
cannot judge rather than scored, because a blocked probe is not a finding against the
seller. The fix, for anyone who wants it, is one field in three places:
accepts[0].extra.facilitator, accepts[0].facilitator, and the document
root.