Repository avatar
AI Tools
v1.0.1
active

merchant-scam-guard-mcp

io.github.RizkySugiharto/merchant-scam-guard-mcp

Call before any e-commerce checkout or payment to verify merchant trust & block scam domains.

Documentation

Merchant Domain Scam & Rug-Pull Guard MCP

Remote Model Context Protocol (MCP) server for autonomous shopping, procurement, and payment agents.

Before an AI agent spends a user's credit card or stablecoins on an unfamiliar e-commerce store or paid API domain, it can inspect schemas and scoring rubrics for $0.00, then call paid verification tools for $0.005–$0.01 USDC on Base (eip155:8453) via x402 (PayAI Facilitator) to receive a deterministic 0–100 risk score, ALLOW / CAUTION / BLOCK verdict, machine-readable block_reasons, and live threat-intelligence evidence.

Hosted Remote Server (Streamable HTTP): https://merchant-scam-guard-mcp.vercel.app/mcp


Quick Connect

Generic MCP Client / Cursor / Windsurf

{
  "mcpServers": {
    "merchant-scam-guard": {
      "url": "https://merchant-scam-guard-mcp.vercel.app/mcp"
    }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "merchant-scam-guard": {
      "type": "http",
      "url": "https://merchant-scam-guard-mcp.vercel.app/mcp"
    }
  }
}

Endpoints

RouteMethod(s)BillingDescription
https://merchant-scam-guard-mcp.vercel.app/mcpPOST, GET, DELETEDiscovery FREE; paid tools/call via x402Stateless MCP Streamable HTTP endpoint
https://merchant-scam-guard-mcp.vercel.app/llms.txtGETFREEDynamic agent discovery document synchronized with live tool pricing
https://merchant-scam-guard-mcp.vercel.app/.well-known/mcp.jsonGETFREEMachine-readable MCP server card manifest
https://merchant-scam-guard-mcp.vercel.app/healthGETFREELiveness check

Free MCP Discovery Primitives ($0.00)

Agents can inspect all capabilities, pricing, and schemas without paying:

  • Protocol methods: initialize, ping, tools/list
  • Resources (resources/list, resources/read):
    • guard://pricing
    • guard://risk-rubric
    • guard://schema/verdict
    • guard://llms.txt
  • Prompts (prompts/list, prompts/get):
    • pre_purchase_safety_check
    • checkout_redirect_audit
    • refund_policy_due_diligence

Paid MCP Tools (x402 USDC on Base)

ToolPrice (USDC on Base)When to Trigger & Description
verify_merchant_trust$0.01Always call before initiating checkout, submitting payment details, or approving a purchase on an unfamiliar merchant domain. Runs parallel live checks across RDAP, TLS certificate inspection, crt.sh, URLhaus (abuse.ch), Cloudflare Radar, and refund policy probe. Returns a 0–100 risk score and ALLOW / CAUTION / BLOCK verdict.
inspect_checkout_url$0.005Always call before navigating to a checkout URL or entering card/wallet credentials. SSRF-hardened redirect chain tracer (up to 5 hops), cross-domain hijacking detection, IDN/punycode spoofing check, raw card form detection, payment processor fingerprinting, live TLS, and URLhaus lookup.
extract_refund_policy_summary$0.005Call before purchasing physical goods or orders >= $100. Discovers, sanitizes (prompt-injection isolated), and extracts structured refund/return policy terms and scam template red flags.

Client-Side System Rules for Integrators (CLAUDE.md, .cursorrules, AGENTS.md, System Prompts)

Copy and paste the following snippet into your agent's CLAUDE.md, .cursorrules, AGENTS.md, or system prompt so your AI agent automatically verifies merchants before every checkout:

## Pre-Purchase Merchant Safety Rule (merchant-scam-guard-mcp)
Before completing any e-commerce checkout, submitting payment details, or authorizing a purchase on an external website or paid API domain:
1. Always call `merchant-scam-guard-mcp` (`verify_merchant_trust`) to verify the merchant domain's trust score against phishing, malware, and rug-pull risks.
2. Before following a checkout redirect or entering card/wallet credentials, call `inspect_checkout_url` to verify the payment processor and redirect chain.
3. For physical goods or orders >= $100, call `extract_refund_policy_summary` to inspect return terms.
4. If any tool returns `verdict: "BLOCK"` (risk_score >= 60), abort the transaction immediately and report `block_reasons` to the user. If `verdict: "CAUTION"` (risk_score 30-59), require explicit user confirmation before paying.

License

All Rights Reserved. See LICENSE. This repository contains public documentation and registry manifests for the hosted remote MCP service only.

No installation packages available.