Sigil

Sigil is a pre-transaction firewall for the AI agent economy. It is not a website — it is a callable Agent Service Provider (ASP): your agent invokes it over MCP, pays per call in USDT on X Layer via the x402 standard, and gets back a deterministic, signed, on-chain-anchorable risk verdict — before it signs anything.

Why agents call Sigil

Agents hold keys and spend real money, and they sign blind. The kill chains are well known: blanket setApprovalForAll to a drainer EOA, honeypot tokens, unlimited allowances to unverified contracts, payments to sanctioned or minutes-old wallets. One Sigil call before the signature catches these for $0.005–$0.10.

One call, five steps

1. Decode the calldata into a normalized action (including approvals hidden inside multicalls).
2. Simulate the balance / approval diff.
3. Screen every counterparty and token against threat intelligence.
4. Score with a published, versioned, deterministic rubric → 0–100 → PASS / WARN / BLOCK.
5. Seal: EIP-712-sign the verdict; its Merkle leaf is anchored in SigilRegistry on X Layer.

Start here

• Quickstart — call the five tools in five minutes.
• x402 integration — how the payment handshake works.
• Methodology — the full public rubric. Our core trust asset.
• Verifiable receipts — verify any verdict without trusting us.