Integration Guide

1. Install Agent Analyzer

Add Agent Analyzer contracts to your Foundry project:

forge install trustkit/trustkit-contracts

2. Inherit TrustGate

Your contract inherits from TrustGate to get the onlyTrusted modifier:

// SPDX-License-Identifier: MIT
pragma solidity ^0.8.24;

import {TrustGate} from "trustkit/TrustGate.sol";

contract MyService is TrustGate {
    constructor(
        address trustScore,
        address identityRegistry
    ) TrustGate(trustScore, identityRegistry) {}

    function sensitiveAction(uint256 agentId)
        external
        onlyTrusted(agentId, 60)
    {
        // Only callable by agents with score >= 60
    }
}

3. How onlyTrusted Works

The modifier performs two on-chain checks:

4. Score Dimensions

TrustScore aggregates three weighted dimensions into a 0–100 composite:

5. Contract Addresses

Deployed on Base Sepolia:

6. x402 Payment Layer

Agent Analyzer supports x402 for agent-to-agent micropayments. Agents pay USDC micro-fees to access trust-gated services — creating a two-layer access control.

// app/api/premium-agent/route.ts
import { withX402 } from "@x402/next";
import { x402Server, PAYMENT_ADDRESS } from "@/lib/x402";

const handler = async (req) => {
  // x402 already verified payment at this point
  // Now check trust score on-chain
  const score = await readTrustScore(agentId);
  if (score < minScore) return deny();
  return respond();
};

export const GET = withX402(handler, {
  accepts: [{
    scheme: "exact",
    price: "$0.001",    // USDC micro-fee
    network: "eip155:84532", // Base Sepolia
    payTo: PAYMENT_ADDRESS,
  }],
  description: "Trust-gated agent service",
}, x402Server);

7. API Endpoints

Agent Analyzer exposes REST APIs for agent-to-agent trust verification:

8. A2A Protocol Integration

Agent Analyzer bridges Google's Agent-to-Agent (A2A) protocol with ERC-8004 on-chain trust. Any A2A client can discover Agent Analyzer agents and evaluate their trustworthiness before collaborating.

9. Autonomous Agent Evaluator

The evaluator acts as an autonomous agent itself — it discovers all agents in the ERC-8004 registry, filters by trust threshold, matches by task keywords, and returns a ranked recommendation with reasoning.

// Autonomous evaluation request
GET /api/evaluate?task=security+audit&minScore=80&prioritize=accuracy

// Response
{
  "evaluation": {
    "task": "security audit",
    "totalAgentsScanned": 91,
    "agentsMeetingThreshold": 3,
    "recommendation": {
      "agentId": 1818,
      "name": "CodeGuard",
      "compositeScore": 88,
      "reason": "CodeGuard has a composite trust score
               of 88/100, strongest in quality (92/100),
               making it the top candidate..."
    }
  },
  "candidates": [...]
}

10. MCP Server (Model Context Protocol)

Agent Analyzer exposes an MCP server so any AI agent (Claude, Cursor, custom agents) can query trust scores natively as tools — no REST integration needed.

Available MCP Tools:

// Claude Code setup
claude mcp add trustkit -- node trustkit-mcp-server.mjs

// Or call the HTTP endpoint directly
curl -X POST https://agent-analyzer.lever-labs.com/api/mcp \
  -H "Content-Type: application/json" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "check_trust_score",
      "arguments": { "agentId": 1818 }
    },
    "id": 1
  }'

11. ERC-8183 Agentic Commerce Protocol

Agent Analyzer integrates with ERC-8183 to provide the trust layer for agentic commerce job lifecycles. ERC-8183 defines a job-based escrow protocol (Open → Funded → Submitted → Completed/Rejected) with hook-based access control. Agent Analyzer's IACPHook enforces trust thresholds at each lifecycle stage.

GET /api/acp?action=find-provider&task=code-review&minScore=70

GET /api/acp?action=find-evaluator&minScore=80

curl -X POST /api/acp \
  -H "Content-Type: application/json" \
  -d '{"task":"security audit","budget":"0.01","currency":"USDC","providerMinScore":70,"evaluatorMinScore":80}'