maiat-sdk

Trust scores, token safety & swap verification for AI agents.

Install

npm install maiat-sdk

Quick Start

import { Maiat } from "maiat-sdk";

const maiat = new Maiat();

// Check if an agent is trustworthy
const trusted = await maiat.isTrusted("0x...");

// Token safety check
const safe = await maiat.isTokenSafe(
  "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913",
);

// Full trust score
const score = await maiat.agentTrust(
  "0xE6ac05D2b50cd525F793024D75BB6f519a52Af5D",
);
console.log(score.trustScore, score.verdict); // 69, "caution"

// Trust-verified swap quote
const swap = await maiat.trustSwap({
  swapper: "0x...",
  tokenIn: "0xEeeeeEeeeEeEeeEeEeEeeEEEeeeeEeeeeeeeEEeE", // ETH
  tokenOut: "0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913", // USDC
  amount: "10000000000000000", // 0.01 ETH in wei
});
console.log(swap.trust.tokenOut?.score); // 100
console.log(swap.calldata); // ready-to-sign tx calldata

API

new Maiat(config?)

| Option | Default | Description | | --------- | ----------------------------------- | --------------------------------------- | | baseUrl | https://app.maiat.io | API base URL | | apiKey | — | Optional API key for higher rate limits | | timeout | 15000 | Request timeout (ms) |

Methods

| Method | Returns | Description | | -------------------------------- | ------------------- | ---------------------------------- | | agentTrust(address) | AgentTrustResult | Full trust score + breakdown | | tokenCheck(address) | TokenCheckResult | Token safety analysis | | trustSwap(params) | TrustSwapResult | Swap quote with trust verification | | listAgents(limit?) | { agents, total } | Browse indexed agents | | isTrusted(address, threshold?) | boolean | Quick trust check (default ≥60) | | isTokenSafe(address) | boolean | Quick token safety check |

Outcome Reporting (Training Data)

The most valuable data for improving trust accuracy:

const maiat = new Maiat({ clientId: "my-trading-bot" });

// 1. Check trust
const score = await maiat.agentTrust("0xCounterparty");

// 2. Take action based on trust
if (score.trustScore >= 60) {
  // ... execute swap ...
}

// 3. Report what happened
await maiat.reportOutcome({
  target: "0xCounterparty",
  action: "swap",
  result: "success", // or "failure", "scam"
  txHash: "0x...", // on-chain proof
  maiatVerdict: score.verdict,
  maiatScore: score.trustScore,
});

This closes the feedback loop: check → act → report → oracle gets smarter.

Links

• Protocol: app.maiat.io
• GitHub: JhiNResH/maiat-protocol
• ACP: Agent #3723 on Virtuals