agent-feedback-mcp-server 0.1.0

Quality signals for MCP tools — agents report results, building a community-driven quality database

Agent Feedback Loop 📊

Community-driven quality signals for MCP tools. Agents report tool results, building a quality database that helps all agents pick better tools.

The Problem

Agents don't know which MCP tools are reliable. They try tools blindly and hope for the best.

The Solution

Automated feedback: agents report success/failure and quality after each tool call. Over time, a quality database emerges that helps every agent make better decisions.

Installation

pip install agent-feedback-mcp-server

{
  "mcpServers": {
    "feedback": {
      "command": "uvx",
      "args": ["agent-feedback-mcp-server"]
    }
  }
}

Tools

Tool	Description
report_tool_result	Report success/failure and quality score
get_tool_quality	Get quality metrics for a specific tool
get_best_tools	Find highest-rated tools (optionally by task)
get_trending_tools	See what's trending recently

Network Effect

More agents reporting → Better quality data → Better tool choices → More agents using → More reports. The database gets better with every user.

License

MIT