policycheck-mcp

MCP server for AI seller verification and policy risk analysis. Checks return policies, shipping, warranty, and terms of service for any online store.

Powered by PolicyCheck — the policy analysis engine behind LegalEasy.

Why use this?

AI purchasing agents and agentic commerce platforms need to verify sellers before completing transactions on behalf of users. PolicyCheck gives your agent the ability to:

• Assess seller trustworthiness before checkout
• Flag risky policies like no-refund clauses, binding arbitration, and class action waivers
• Score buyer protection on a 0–100 scale with factual summaries
• Auto-discover policies from any online store URL

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

Tools

Tool	Description
analyze_seller	Full risk analysis of a specific policy page URL. Returns risk level, buyer protection score, key findings, and a factual summary.
quick_risk_check	Give it a store URL and it auto-discovers return, shipping, and terms pages. Returns an overall risk score with per-policy breakdowns.
check_policy_text	Paste raw policy text and get an instant risk assessment. No URL needed — useful when text is already extracted.

Installation

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "policycheck": {
      "command": "npx",
      "args": ["-y", "policycheck-mcp"]
    }
  }
}

Claude Code

claude mcp add policycheck -- npx -y policycheck-mcp

Cursor

Add to your Cursor MCP settings:

{
  "mcpServers": {
    "policycheck": {
      "command": "npx",
      "args": ["-y", "policycheck-mcp"]
    }
  }
}

Configuration

Environment Variable	Default	Description
POLICYCHECK_API_URL	https://policycheck.tools/api/a2a	A2A endpoint URL

How it works

Each tool sends a JSON-RPC 2.0 request to the PolicyCheck A2A (Agent-to-Agent) endpoint. The API analyzes seller policies and returns:

• Risk level: low / medium / high / critical
• Buyer protection score: 0–100
• Key findings: Plain-English summary of risks (arbitration clauses, no-refund policies, liability caps, etc.)
• Summary: Factual description of detected risk indicators and policy findings

Example

Ask Claude: "Get risk data for https://www.example-store.com"

The agent calls quick_risk_check, gets back structured risk data, and uses it alongside purchase context to inform a decision.

License

MIT