Before your agent pays an unknown x402 or MCP resource, ask the question the payment rail can't answer: who am I paying, what changed, is this price normal, does the live endpoint match its history? One call returns a risk verdict — ALLOW, REVIEW, or DENY — with the evidence behind it.
Built on a daily longitudinal census of the entire x402 Bazaar. Each check reports evidence across identity, protocol, change, entity, economics, and consistency — then a verdict.
// npm i @x402/fetch @x402/evm viem
import { wrapFetchWithPayment, x402Client } from "@x402/fetch";
import { registerExactEvmScheme } from "@x402/evm/exact/client";
import { privateKeyToAccount } from "viem/accounts";
const client = new x402Client();
registerExactEvmScheme(client, { signer: privateKeyToAccount(process.env.EVM_PRIVATE_KEY) });
const fetchWithPay = wrapFetchWithPayment(fetch, client);
// before paying some unknown resource, check it:
const res = await fetchWithPay("https://verify.inferenceview.com/check", {
method: "POST",
headers: { "content-type": "application/json" },
body: JSON.stringify({ url: "https://api.some-tool.com/paid-endpoint" })
});
console.log(await res.json());
{
"ok": true,
"risk": "HIGH",
"recommendation": "DENY",
"signals": [
{ "level": "high", "code": "recipient_mismatch",
"detail": "Live payment recipient 0xBAD… differs from last observed 0xF99… — classic swap indicator." },
{ "level": "high", "code": "price_anomaly",
"detail": "$0.40/call is a 5.4σ outlier vs this host's 25 priced resources (median $0.012). Clean cohort — high-confidence signal." },
{ "level": "info", "code": "seller_footprint",
"detail": "Recipient wallet operates 12 known resources across 2 hosts." }
],
"live_offer": { "scheme": "exact", "network": "eip155:8453", "priceUsd": 0.40, "payTo": "0xBAD…" },
"checked_against": { "resources": 6000, "hosts": 970, "seller_wallets": 783, "first_snapshot": "2026-07-16" },
"verified_at": "2026-07-16T14:32:00Z",
"valid_for_seconds": 3600
}
Never charged for failed work. Payment settles only after the check completes. Free for humans: paste any resource at inferenceview.com/gate (rate-limited). Live census: GET /registry/stats.
Our original utility, for when an agent needs a fact grounded rather than a counterparty checked. Send a claim, get a verdict (supported / refuted / mixed / unverifiable) with sources, confidence, and contradictions. $0.05 / $0.10 / $0.25.
POST {claim, context?} → {verdict, confidence, primary_sources, freshness, contradictions, evidence}. Legacy /verify/* routes remain active aliases.
We never return a green checkmark. "Verified" refers to specific assertions, each backed by observed evidence:
| Dimension | Question it answers |
|---|---|
| Identity | Is this who it claims to be — and have we seen it before? |
| Protocol | Is the live x402 challenge structurally valid right now? |
| Change | Did the wallet, price, schema, or description move since it was last trusted? |
| Entity | What else does the payment wallet operate — one tool or a whole footprint? |
| Economics | Is the price a real outlier next to this seller's other endpoints? (robust median+MAD z-score, validated at 1.8% false-positive rate on clean cohorts) |
| Consistency | Does the live offer match the recipient we have on record? |
We validate our own detector like a trading strategy before shipping it: measured 1.8% false-positive rate, and the price signal only fires where the data supports it — elsewhere it returns context and says so (live coverage). The moat is longitudinal: change intelligence needs history no one can reconstruct later. Delivery intelligence (did you get what you paid for?) is next — VERIFY is becoming a customer of the resources it rates.
Download the shim, run it with your own wallet key, and any MCP client (Claude Desktop, Claude Code, Cursor, your agent framework) gets a verify tool that pays per call automatically. Your key never leaves your machine.
curl -O https://verify.inferenceview.com/shim.mjs npm i @modelcontextprotocol/sdk @x402/axios @x402/evm axios zod viem EVM_PRIVATE_KEY=0x... node shim.mjs
// MCP client config
{ "mcpServers": { "verify": {
"command": "node", "args": ["/path/to/shim.mjs"],
"env": { "EVM_PRIVATE_KEY": "0x..." } } } }
| Field | Description |
|---|---|
| claim | required · string ≤1500 chars — the statement to verify |
| context.url | optional — page the claim came from |
| context.company | optional — company the claim concerns |
| context.person | optional — person the claim concerns |
| context.product | optional — product the claim concerns |
| context.data_point | optional — the specific figure to check |
Sales research agents checking a prospect fact before an outreach. Procurement agents validating a vendor claim before purchase. Financial research agents grounding a datapoint before a report. Coding agents confirming a library's behavior before adopting it. Any agent whose next action costs more than $0.10 if the premise is wrong.
Machine-readable offer sheet: /pricing
InferenceView is the intelligence layer for enterprise AI consumption — and the next thing enterprises will ask is "what are my agents buying, and are their purchases economically rational?" VERIFY is our first machine-native utility: a real x402 seller, operating in the open, so we understand agentic commerce from the inside while we build the AgentOps layer that watches it.
Operated with human oversight: hard daily capacity, kill switch, and a full transaction ledger. Questions or higher-volume needs: inferenceview.com.