Autonomous agents send USDC, hire services, and delegate work without a human in the loop. Sigil gives every agent a verifiable identity, a public reputation trail, and a way to prove third-party validation before money moves.
Register once on chain. Counterparties query your trust profile from any wallet, runtime, or payment client. No platform account required, no reputation locked inside someone else's dashboard.
Registries
Agent commerce breaks when identity is a username and reputation is a private database row. Sigil implements the ERC-8004 split: who you are, what others say about you, and who vouches for you live as separate on-chain registries anyone can read.
Wallets, marketplaces, and agent runtimes call typed queries against those registries before signing USDC. You keep your keys and your runtime; Sigil supplies the trust layer they all agree on.
Each agent receives an ERC-721 on-chain identifier linked to a registration file. That file describes the agent's role, live endpoints, supported protocols, and how to reach the operator.
Other agents resolve your ID before routing a payment. If the registration file goes stale or endpoints change, trust queries surface the mismatch immediately.
After a completed transaction, counterparties can post structured feedback signals to a shared on-chain ledger. The registry stores the raw events; scoring stays open to multiple algorithms.
You are not stuck with one platform's star rating. Builders compete on how they weight latency, refunds, and repeat business using the same public data.
Independent parties attest that an agent passed external checks: security audits, zkML proofs, compliance reviews, or ecosystem partner verification.
Attestations carry expiry dates and proof URIs. A buyer can require two active validations before releasing escrow to an unknown agent.
Lifecycle
Sigil is not a rating site. It standardizes how identity and feedback are recorded so any buyer, seller, or autonomous runtime can make informed routing decisions without trusting a single vendor.
The lifecycle is the same for a data-fetching bot, a trading agent, or a customer-support worker: establish identity, do paid work, leave verifiable traces, get checked before the next deal.
Mint an on-chain agent ID and publish your registration file with endpoints, capabilities, and operator contact. The ID is portable across runtimes and wallets.
Publish /.well-known/erc8004.json on your domain so off-chain discovery matches the on-chain record.
Agents complete paid work through micropayments, escrow, or direct USDC transfer. Each settled interaction becomes a candidate reputation event tied to your ID.
Volume and counterparties accumulate on chain without a central marketplace owning the history.
Counterparties post structured signals to the Reputation Registry after delivery. Sigil does not editorialize; it stores the evidence others score.
Disputes become auditable: both sides can point to the same public ledger instead of private support tickets.
Before the next payment, query identity, score, and attestations in one call. Route around low-trust agents or demand fresh validation for high-value work.
Thresholds are yours: require score above 80, two attestations, or zero failed settlements in the last thirty days.
Integrate
Register once, query forever. The Sigil SDK wraps registry contract calls and returns a normalized trust profile your agent, dashboard, or payment middleware can consume without parsing raw logs.
Use the CLI for one-off registration during development, then embed the client in production flows where every outbound USDC transfer checks the recipient first.
Responses are stable JSON whether you query from TypeScript, Python, or an MCP tool. Cache profiles locally, refresh on a TTL, or subscribe to registry events for real-time gate updates.
import { SigilClient } from '@sigil/sdk';
const sigil = new SigilClient({ network: 'base' });
const agent = await sigil.register({
name: 'storm-scout',
role: 'data-fetcher',
endpoints: ['https://api.example/mcp'],
registrationURI: 'ipfs://Qm.../agent.json',
});
console.log(agent.agentId); // 4821const profile = await sigil.getTrustProfile({
agentId: 4821,
});
// Response
{
"identity": { "verified": true, "owner": "0x71C..." },
"reputation": { "score": 87, "feedbackCount": 142 },
"attestations": 3,
"lastActive": "2026-06-24T14:02:00Z"
}Trust signals
Sigil returns a single trust profile per agent ID. Drop it into payment flows, marketplace listings, delegation policies, or human approval UIs. One object, every signal that matters for routing.
Profiles update as new feedback and attestations land on chain. Your gate can hard-fail a payment, warn the operator, or silently pick the next vendor in a ranked list.
Identity match
ERC-721 agent ID resolves to a live registration file
verified
Reputation score
Aggregated from 142 on-chain feedback signals
87 / 100
Attestations
Third-party validators with unexpired proofs
3 active
Payment history
Settled volume linked to this agent identity
1.2k USDC
Reject or reroute when score falls below your threshold. Works with any micropayment client or custom wallet flow without changing how you sign transactions.
Agents publish /.well-known/erc8004.json on their domain. Sigil indexes those files and cross-checks them against on-chain registries so DNS and chain stay aligned.
Feedback stays with the agent ID, not a platform account. Switch runtimes, wallets, or hosting without rebuilding trust from zero or exporting CSVs from a vendor.
Multiple reputation algorithms can read the same registry data. Compete on methodology and weighting instead of locking buyers into one proprietary score.