@jhinresh/mcp-server

The trust layer for AI agents — as an MCP server.

Query Maiat trust scores from Claude, GPT, or any MCP-compatible AI assistant. One tool call = one trust check.

Quick Start

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "maiat": {
      "command": "npx",
      "args": ["@jhinresh/mcp-server"],
      "env": {
        "ALCHEMY_API_KEY": "your-alchemy-key",
        "MAIAT_API_KEY": "optional-for-higher-rate-limits"
      }
    }
  }
}

From source

cd packages/mcp-server
npm install
npm run build
npm start

Tools

| Tool | Description | |------|-------------| | trust_score | Get trust score (0-1000) for any address | | token_safety | Token-specific safety check (honeypot, rug, liquidity) | | protocol_rating | Protocol safety rating by name | | batch_score | Score up to 10 addresses at once | | explain_score | Human-readable score explanation |

Example Usage (in Claude)

"Check if 0x1234...abcd is safe to interact with"

The AI will call trust_score and return something like:

{
  "address": "0x1234...abcd",
  "score": 720,
  "risk": "low",
  "source": "live_onchain",
  "breakdown": {
    "base": 100,
    "txHistory": 200,
    "balanceSignal": 100,
    "contractBonus": 50
  }
}

Resources

| Resource | URI | Description | |----------|-----|-------------| | Scoring Methodology | maiat://methodology | How scores are calculated |

Environment Variables

| Variable | Required | Description | |----------|----------|-------------| | ALCHEMY_API_KEY | Recommended | Enables live on-chain scoring | | MAIAT_API_KEY | Optional | Higher rate limits on Maiat API | | MAIAT_API_URL | Optional | Custom API endpoint (default: https://app.maiat.io/api/v1) |

How It Works

1. Maiat DB first — checks indexed scores from Maiat's database
2. Live fallback — if no DB record, queries on-chain data via Alchemy
3. Scoring algorithm — transaction count, balance, contract status → 0-1000 score

License

MIT