Autonomous agents hire and pay agents they've never met — instantly, irreversibly, with no one checking who's on the other side. Vouchgate is that check: one call returns a signed PASS / FLAG / BLOCK verdict before money moves.
No implemented check matched, against point-in-time public data.
Elevated risk — e.g. a known mixer contract, or a fresh wallet with no track record.
Hard signal — e.g. a sanctioned address on the OFAC SDN list.
Every verdict is ed25519-signed and reproducible — verify it yourself against the published key at /signing-key. Read-only and non-custodial: Vouchgate never holds keys, moves funds, or sits in a settlement path.
POST https://mcp.vouchgate.xyz/mcp (MCP tool: screen_counterparty)
{ "address": "0x… wallet or OKX.AI agent ID" }
→ {
"verdict": "BLOCK",
"riskScore": 100,
"reasons": [
"address matches a sanctions-adjacent list entry (OFAC SDN snapshot)",
"on-chain activity found (ethereum: nonce 2, holds balance)"
],
"evidenceHash": "…", "signature": "…"
}
Billed per call via x402 (OKX Agent Payments Protocol) — $0.01, settled on X Layer.
Every payment is verifiable on the X Layer explorer. These are per-call payments from distinct third-party addresses — real usage, not a claim.
Every screening call is a labelled signal about a counterparty. The intended moat is a privacy-preserving risk knowledge base: what Vouchgate learns about an address is derived from public on-chain data and accrues over time, while who asked stays private — caller identity is hashed and never linked to a screened target. More usage sharpens the signal without ever selling out the querier. That data network effect — not any single check — is the long-term defensibility.