MCP Rubber Duck

What it does

MCP Rubber Duck connects your AI client to multiple LLM providers simultaneously — OpenAI, Gemini, Groq, Claude, Codex, and more. Inspired by rubber duck debugging, it lets you explain a problem to a "panel of ducks" and get diverse AI perspectives at once. Four tools render rich interactive HTML panels inside MCP Apps-compatible clients (Claude Desktop, Cursor, VS Code Copilot) for side-by-side comparison, voting, debates, and usage analytics.

Key features

• Compare Ducks — Side-by-side model responses with latency, token counts, and error states in a live interactive panel
• Duck Vote — Multi-LLM voting with confidence bars, consensus badge, and collapsible reasoning
• Duck Debate — Structured Oxford, Socratic, or adversarial debates rendered round-by-round
• Usage Stats — Provider breakdown with estimated costs and token distribution bars
• MCP Bridge — Route duck queries through external MCP servers for extended tool access
• CLI Agent Support — Use Claude Code, Codex, Gemini CLI, Grok, or Aider as ducks
• Guardrails — Rate limiting, token limits, pattern blocking, and PII redaction

Installation

Claude Desktop — add to claude_desktop_config.json:

{
  "mcpServers": {
    "rubber-duck": {
      "command": "npx",
      "args": ["mcp-rubber-duck"],
      "env": {
        "OPENAI_API_KEY": "sk-..."
      }
    }
  }
}

VS Code Copilot — add to .vscode/settings.json:

{
  "github.copilot.chat.mcp.servers": {
    "rubber-duck": {
      "command": "npx",
      "args": ["mcp-rubber-duck"],
      "env": { "OPENAI_API_KEY": "sk-..." }
    }
  }
}

Cursor — add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "rubber-duck": {
      "command": "npx",
      "args": ["mcp-rubber-duck"],
      "env": { "OPENAI_API_KEY": "sk-..." }
    }
  }
}

Supported hosts

Claude Desktop, VS Code Copilot, Cursor, Windsurf (all confirmed in README/docs).