mcp-shield

Trust-Gated Agent Examples

Reference implementations showing how to make AI agents check CraftedTrust before connecting to any MCP server. If a server scores D or F, the agent refuses to connect.

This is the "SSL certificate check for AI agents" pattern.

Quick Start

Add CraftedTrust to your agent's MCP config:

{
  "mcpServers": {
    "craftedtrust": {
      "url": "https://mcp.craftedtrust.com/api/v1/mcp",
      "description": "Check trust scores before connecting to MCP servers"
    }
  }
}

Your agent now has access to 6 tools:

The Pattern

Before your agent connects to any new MCP server:

1. Call check_trust on CraftedTrust
2. If grade is D or F, refuse and explain why
3. If grade is A/B/C, proceed with the connection

Examples

Python (LangGraph)

See python/trust_gated_agent.py for a complete LangGraph agent that gates MCP connections through CraftedTrust.

cd python
pip install -r requirements.txt
python trust_gated_agent.py

TypeScript (MCP Client)

See typescript/trust-gated-client.ts for a TypeScript MCP client that checks trust scores before connecting.

cd typescript
npm install
npx tsx trust-gated-client.ts

How It Works

Agent receives task → needs mcp.example.com
         ↓
Agent calls CraftedTrust check_trust("mcp.example.com")
         ↓
Score: 82/100, Grade: B ← Safe
         ↓
Agent connects to mcp.example.com ✓

Agent receives task → needs sketchy-server.xyz
         ↓
Agent calls CraftedTrust check_trust("sketchy-server.xyz")
         ↓
Score: 15/100, Grade: F ← Dangerous
         ↓
Agent refuses: "Server scored F (15/100). Not connecting." ✗

Trust Score Grades

API Reference

Full API documentation: mcp.craftedtrust.com/api-docs.html

License

MIT

Built by Cyber Craft Solutions LLC