@mimranakb/trust-layer-mcp

MCP server that gives any AI agent the x402 Trust Layer — 13 trust tools an agent calls before, during, and after it spends money over x402. Pay-per-call in USDC on Base / Solana / Polygon. No API keys — the agent pays with its own wallet via the Dexter / CDP facilitator.

Works with Claude Desktop, Claude Code, Cursor, and any MCP-compatible client.

Why

Autonomous agents pay unknown merchants. Before the private key signs, you want to know: is this merchant trustworthy? is the payment inside policy? was the response I paid for actually worth it? This server wires those checks into your agent as native MCP tools.

Tools

| Tool | Endpoint | Price | |------|----------|-------| | trust_before_x402_fetch | /api/pipeline/trust-v2 — full pre-pay flow (mandate diff + KYM + guard + buyer gate) | $0.35 | | trust_preflight_proxy | /api/x402/proxy — all-in-one preflight + optional attestation | $0.08 | | trust_agent_verify | /api/agent/verify — ERC-8004 TrustScore on Base mainnet | $0.04 | | trust_merchant_score | /api/merchant-trust/score — KYM (wash-trade / verification signals) | $0.06 | | trust_mandate_diff | /api/mandate/diff — signed mandate vs actual tool trace | $0.04 | | trust_buyer_gate | /api/trust-network/buyer-gate — certified-seller buyer gate | $0.03 | | trust_semantic_settle | /api/quality-escrow/semantic-settle — release or auto-refund | $0.12 | | trust_receipt_verify | /api/receipt-auditor/verify — verify settlement tx on-chain | $0.05 | | trust_protocol_full_pipeline | /api/protocol/pipeline/full-trust — Agent Trust Protocol v4 | $0.45 | | trust_protocol_trust_score_v2 | /api/protocol/trust-score/v2 — multi-factor TrustScore v2 | $0.08 | | trust_protocol_fraud_scan | /api/protocol/fraud/scan — Sybil / wash-trade / circular-payment scan | $0.10 | | trust_protocol_execution_receipt | /api/protocol/execution/issue — Proof-of-Execution receipt | $0.05 | | trust_protocol_credit_score | /api/protocol/credit/score — AI Agent Credit Bureau score (300–900) | $0.06 |

Full catalog and live dashboard: https://x402trustlayer.xyz/dashboard

Install & configure

The agent pays per call, so it needs a funded wallet key in the server's env.

export EVM_PRIVATE_KEY=0x...        # OR
export SOLANA_PRIVATE_KEY=...
# optional:
export TRUST_LAYER_BASE=https://x402trustlayer.xyz
export X402_PREFERRED_NETWORK=eip155:8453

Claude Desktop / Claude Code / Cursor

Add to your MCP config (claude_desktop_config.json, .cursor/mcp.json, etc.):

{
  "mcpServers": {
    "trust-layer": {
      "command": "npx",
      "args": ["-y", "@mimranakb/trust-layer-mcp"],
      "env": {
        "EVM_PRIVATE_KEY": "0x..."
      }
    }
  }
}

Restart the client and the 13 trust_* tools appear.

Recommended agent flow

1. trust_before_x402_fetch   → one call: mandate diff + KYM + guard + buyer gate
2. (your agent) x402_fetch    → pay the external merchant in USDC
3. trust_semantic_settle      → did the response match intent? release or auto-refund
4. trust_receipt_verify       → confirm the settlement tx on-chain

For high-value or regulated flows, use trust_protocol_full_pipeline (Agent Trust Protocol v4: passport, fraud, oracle, credit, compliance, guard, replay-bind in one call).

Security & custody

• Keys live only in the server process env and are used solely to sign x402 payments your agent initiates.
• Use a dedicated agent wallet with a limited balance — never your main treasury key.
• All Trust Layer responses are advisory signals; you keep control of the final payment decision.

Local dev

cd packages/trust-layer-mcp
npm install
npm run build
node dist/index.js   # speaks MCP over stdio

Links

• Live API & dashboard: https://x402trustlayer.xyz/dashboard
• OpenAPI: https://x402trustlayer.xyz/openapi.json
• Source: https://github.com/mimranchohan/x402-trust-layer

MIT © Mimran Chohan