Agent Reputation MCP Server

MCP server for AI agent trust scores, ratings, and reputation tracking. Enables AI agents to rate each other, track interaction outcomes, and compute trust scores.

Features

• submit_rating — Submit a trust rating (1-5) for an AI agent with category and comment
• get_reputation — Full reputation profile with score distribution and category breakdown
• get_leaderboard — Top-rated agents overall or filtered by category
• report_interaction — Log interaction outcomes (success/failure) for trust computation
• get_trust_score — Computed trust score (0-100) based on ratings + interactions + volume
• compare_agents — Side-by-side reputation comparison of multiple agents

Installation

npm install -g @aiagentkarl/agent-reputation-mcp-server

Usage with Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "agent-reputation": {
      "command": "npx",
      "args": ["-y", "@aiagentkarl/agent-reputation-mcp-server"]
    }
  }
}

Data Storage

All data is stored locally in ~/.agent-reputation/ as JSON files:

• ratings.json — All submitted ratings
• interactions.json — All logged interactions

Trust Score Computation

The trust score (0-100) is computed from three components:

| Component | Max Points | Source | |-----------|-----------|--------| | Rating Score | 50 | Average rating * 10 | | Interaction Score | 30 | Success rate * 30 | | Volume Bonus | 20 | log10(total_count + 1) * 10 |

Trust Levels

| Score | Level | |-------|-------| | 80-100 | Highly Trusted | | 60-79 | Trusted | | 40-59 | Neutral | | 20-39 | Low Trust | | 0-19 | Untrusted |

License

MIT