Security x402 offerings

Is this merchant safe to depend on? Audits up to ten x402 endpoints in one call and returns a verdict — ready, partially ready or not ready — with the specific blockers. Checks each endpoint serves a payable challenge with well-formed terms and a declared input schema, and adds the cross-endpoint check no single probe can make: whether every endpoint pays the same recipient. Never pays anything.

Is this API safe to pay? Audit any x402 merchant or paid endpoint BEFORE you send money. Answers whether a service is trustworthy, whether it is a scam or dead, and whether it will actually deliver. Probes the live endpoint for a valid HTTP 402 challenge, detects price drift between the advertised price and the real challenge price, pulls settlement history and unique-buyer counts, and verifies on-chain USDC actually reaching the payTo address. Returns a PASS / CAUTION / AVOID verdict with the evidence behind it. Every check is measured live at request time — nothing cached, no reputation scores. Use before paying an unknown merchant, to avoid wasting money on dead or fraudulent endpoints.

Pre-payment risk firewall for AI agents: validate {recipient, amount, chain} BEFORE sending an x402 payment and get ALLOW/REVIEW/BLOCK in <200ms — OFAC blocklist, sanctioned-mixer check, runaway-overspend vs expected price and anomalous-amount guard. Stop an agent from overpaying or paying a bad address; transaction risk screening and payment guardrail before send. The hosted spending guard / agent budget protection an offline SDK cannot give.

Inspect and validate an x402 endpoint before an autonomous agent spends USDC.

Payability preflight: audits an x402 seller's 402 challenge

Inspect and validate an x402 endpoint before an autonomous agent spends USDC.

Paid ETOS-style Repo Trust Check API. Payment via x402 protocol (USDC on Base).

Paid ETOS-style Repo Trust Check API. Payment via x402 protocol (USDC on Base).

Grade any x402 seller's payment-security posture from the outside - a read-only black-box check mapped to the 'Five Attacks on x402' failure modes. Probes the URL's 402 challenge (never pays) and scores TLS transport, gated-response cache hygiene (Attack III / cache leakage), error/info-leak hygiene, and payment-terms well-formedness, then returns a letter grade with per-check findings and an honest note on what only insider/active testing can confirm.

x402 proves payment, not trust: an agent handed a URL pays blind, and the payee may be a real service or a burner that takes the USDC and vanishes. Give a url and/or a payTo wallet and get GO/WARN/STOP: is the URL a well-formed 402 resource, does its price match, and is the seller wallet sanctioned, a fresh throwaway, or paying a different address than you expected. Screens the PAYMENT side the protocol drops — not a guarantee the service returns good data.

Call before trusting or paying a third-party x402 seller: catalog-vs-live integrity -- verify an x402 listing (payee, price, network, service, tags, bazaar block) against the live 402; an endpoint preflight for x402 catalog vs live offers, seller trust and listing consistency so an agent can skip dead or mismatched endpoints (JSON).

x402 Settlement Review: Replay and Double Charge — Reviews pasted seller-side x402 code that accepts the PAYMENT-SIGNATURE header, talks to the facilitator and delivers the goods (the cash register), for the silent failures that double-charge a buyer, hand the goods to a stranger or bill the seller. It flags an expiry check that runs before the nonce lookup, a repeat path that delivers without proof the chain observer lacks, a repeat window taken from the buyer's validBefore instead of the sell…

Use to vet a domain or URL before trusting it, or to check if a name is available. Registry RDAP data (the WHOIS successor) for up to 20 domains per call: registrar, created/expiry dates, age in days, status and transfer lock, nameservers, DNSSEC. Subdomains resolve to their registered domain. One flat price per call.

Semantically similar endpoints that currently out-score a given one on the trust score. Each carries score, grade, similarity, advertised price, and endpoint page; same-host siblings and avoid-flagged endpoints are excluded. An empty list means nothing beats the subject.

Score up to 10 x402 endpoints in a single paid call. Each found result carries the FULL trust report — identical to POST /v1/x402-trust: score, grade, scoreRange, recommendation, confidence breakdown, the component/subscore breakdown, explanations, flags + flagsDetailed, advertised price, and 30-day on-chain stats — not just the scalar verdict. Cache rows older than 15 minutes are recomputed on-demand from the latest stored probes and settlements, so scores typically reflect reality within minutes. No live network re-probe is performed — for a fresh live probe use /v1/x402-trust. Per-request recompute limits apply (time + count); results include `recomputed` and `computed_at` so you can see which rows used cached vs freshly computed scores. Resources not in our observation set return `found:false`; you still pay for the batch AS LONG AS at least one resource is tracked. If NONE of the supplied resources are in our observation set the request is rejected with HTTP 400 `notCharged:true` and you are NOT charged. Duplicates are ignored.

Score up to 100 x402 endpoints in a single paid call. Each found result carries the FULL trust report — identical to POST /v1/x402-trust: score, grade, scoreRange, recommendation, confidence breakdown, the component/subscore breakdown, explanations, flags + flagsDetailed, advertised price, and 30-day on-chain stats — not just the scalar verdict. Cache rows older than 15 minutes are recomputed on-demand from the latest stored probes and settlements, so scores typically reflect reality within minutes. No live network re-probe is performed — for a fresh live probe use /v1/x402-trust. Per-request recompute limits apply (time + count); results include `recomputed` and `computed_at` so you can see which rows used cached vs freshly computed scores. Resources not in our observation set return `found:false`; you still pay for the batch AS LONG AS at least one resource is tracked. If NONE of the supplied resources are in our observation set the request is rejected with HTTP 400 `notCharged:true` and you are NOT charged. Duplicates are ignored.

Score up to 200 x402 endpoints in a single paid call. Each found result carries the FULL trust report — identical to POST /v1/x402-trust: score, grade, scoreRange, recommendation, confidence breakdown, the component/subscore breakdown, explanations, flags + flagsDetailed, advertised price, and 30-day on-chain stats — not just the scalar verdict. Cache rows older than 15 minutes are recomputed on-demand from the latest stored probes and settlements, so scores typically reflect reality within minutes. No live network re-probe is performed — for a fresh live probe use /v1/x402-trust. Per-request recompute limits apply (time + count); results include `recomputed` and `computed_at` so you can see which rows used cached vs freshly computed scores. Resources not in our observation set return `found:false`; you still pay for the batch AS LONG AS at least one resource is tracked. If NONE of the supplied resources are in our observation set the request is rejected with HTTP 400 `notCharged:true` and you are NOT charged. Duplicates are ignored.

Score up to 500 x402 endpoints in a single paid call. Each found result carries the FULL trust report — identical to POST /v1/x402-trust: score, grade, scoreRange, recommendation, confidence breakdown, the component/subscore breakdown, explanations, flags + flagsDetailed, advertised price, and 30-day on-chain stats — not just the scalar verdict. Cache rows older than 15 minutes are recomputed on-demand from the latest stored probes and settlements, so scores typically reflect reality within minutes. No live network re-probe is performed — for a fresh live probe use /v1/x402-trust. Per-request recompute limits apply (time + count); results include `recomputed` and `computed_at` so you can see which rows used cached vs freshly computed scores. Resources not in our observation set return `found:false`; you still pay for the batch AS LONG AS at least one resource is tracked. If NONE of the supplied resources are in our observation set the request is rejected with HTTP 400 `notCharged:true` and you are NOT charged. Duplicates are ignored.

WCAG accessibility audit of a public URL using headless Chromium and axe-core: violation/pass/incomplete counts by impact, per-rule detail, and a 0-100 score.

Install-safety signals for an npm or PyPI package: registry metadata, install-script usage, OSV advisories, typosquat risk, and a 0-100 trust score.

Send an HPKE-encrypted prompt to the Aubrai chat agent running inside an attested private enclave. Only the enclave can decrypt your message.

Offline EIP-712 typed data (chain-agnostic, no RPC): domain + types + primaryType + message JSON in, the normalized typed data plus domainSeparator, hashStruct and the final digest (the 32-byte value…

Random bytes in one GET, without an LLM or Math.random(): hex, base64 or a UUID. curl: GET /random?bytes=32&format=hex (bytes 1-1024, default 32; format hex|base64|uuid, default hex). Response: {"value":"9f2c41ab…(64 hex characters)","format":"hex","bytes":32,"bits":256,"generated_at":"2026-10-16T09:20:00.000Z","sha256":"41ab…","sha256_of":"value","source":"crypto.getRandomValues"}. The bytes come from the Worker runtime's Web Crypto API, crypto.getRandomValues, called once per paid request after the payment is verified; nothing is stored, cached or served twice, and the sha256 is the SHA-256 of the value you received, so you can check the string. format=uuid answers a UUID version 4 string. 0.001 USDC, no account, no key, one GET, one delivery. A purchase is two calls by design: this first call returns 402 with the price and the payment terms, and you then repeat the same request with the PAYMENT-SIGNATURE header to receive the 200 delivery.

SHA-256 of text encoded as UTF-8 (maximum 1024 bytes). Costs 0.01 USDC.