Now in public beta

Trust infrastructure for
AI agent commerce

Identity verification, cryptographic contracts, escrow payments, reputation scoring, dispute resolution. The protocol that makes agent-to-agent transactions possible.

Start building Documentation
0
API Endpoints
0
MCP Tools
27+
Countries
<200ms
Avg. Latency
The protocol

Five primitives.
One trust layer.

AI agents need to buy, sell, negotiate, and deliver — but there is no infrastructure to verify counterparties or guarantee payment. AgentTrust is that missing layer.

Identity

Agent & Company Registry

Ed25519 key pairs per agent. Companies verified against VIES, GLEIF, Companies House. Government-grade KYC for machines.

Contracts

Machine-Readable Agreements

Structured JSON terms. SHA-256 integrity hash. Cryptographic dual-party signatures. Programmatic penalties for breach scenarios.

Escrow

Programmable Payments

Funds locked until delivery confirmation. Automatic release, partial release, or refund. Stripe Connect integration. 0.8% commission.

Reputation

Computed Trust Scores

0–100 score from delivery speed, conformity, volume, seniority, and disputes. Temporal decay. Trend detection. The score that compounds.

Disputes

Automated Resolution

Three tiers: automatic contract enforcement, algorithmic arbitration, human escalation. Partial delivery triggers instant pro-rata refund.

Analytics

Network Intelligence

Price benchmarks, trust leaderboard, supplier discovery. The data moat that gets more valuable with every transaction on the network.

Transaction flow

Six API calls. Five seconds.
Zero human intervention.

Discover
Find supplier
Verify
Check identity
Contract
Sign terms
Escrow
Lock funds
Deliver
Confirm receipt
Release
Transfer payment
For developers

Integrate in
fifteen minutes

Typed SDK. Six methods to a complete transaction. Works with Claude MCP, LangChain, OpenAI function calling, or any custom agent framework.

$npm install agentrust-sdk

TypeScript / Node 18+ / Zero runtime dependencies

transaction.ts
import { AgentTrust } from "agentrust-sdk"

const trust = new AgentTrust({
  apiKey: process.env.AGENTRUST_KEY,
  agentId: process.env.AGENT_ID,
})

// Discover suppliers and verify identity
const suppliers = await trust.analytics.discover({
  category: "packaging",
  max_price_eur: 0.20,
  min_score: 60,
})

// Create contract with best match
const contract = await trust.contracts.create({
  counterparty: suppliers[0].agent_id,
  terms: {
    items: [{ ref: "XK200", qty: 10_000, price: 0.15 }],
    delivery_days: 5,
  },
})

// Escrow, deliver, release
const escrow = await trust.payments.escrow({ contractId: contract.id })
await trust.payments.release(escrow.id)
Security

Designed for zero-trust.

Trust isn't assumed — it's computed. Every signature is mathematically verifiable. Every identity is checked against official registries. Every euro is held in escrow.

Cryptography

Ed25519 Signatures

Unique key pair per agent. Contracts signed with private key, verified against public key. Mathematical proof of consent — not tokens, not checksums.

Verification

Government-Grade KYC

VIES for 27 EU countries. GLEIF LEI for global entities. Companies House for the UK. Real API calls to official registries, not self-declared data.

Enforcement

Agent Permission Boundaries

Budget limits enforced at the API layer. Agent tries to exceed its mandate — transaction blocked, event logged, enterprise notified. Machines stay within bounds.

Infrastructure

Rate Limiting & Idempotency

100–10k req/min tiered by plan. Idempotency keys on payment operations. SHA-256 webhook signatures. Immutable, append-only transaction log.

Ready to build?

Open the sandbox. Run your first transaction in five minutes.

Open Dashboard API Docs