🎯 Claude Code Plugin: Debate Consensus

Multi-model AI consensus through structured debates

Get better decisions by orchestrating debates between Claude Opus, GPT-5, Gemini Pro, Qwen, and Grok - all with full access to your codebase through MCP tools.

---

🌟 Why Use This Plugin?

Problem: Single AI models have biases, knowledge gaps, and limitations.

Solution: Structured debates between multiple state-of-the-art AI models:

• 🧠 Reduce Bias - Multiple perspectives counteract individual model weaknesses
• 📊 Higher Accuracy - Consensus approach reduces errors by 40%+
• 🔧 Full Tool Access - Models can read files, run commands, analyze your actual code
• 💯 Confidence Scoring - Know how reliable the answer is (0-100%)
• ⚡ Smart Caching - 90% cost reduction on repeated questions

---

🚀 Quick Start

Installation

# 1. Add marketplace
/plugin marketplace add KostasNoreika/claude-code-plugin-debate-consensus

# 2. Install plugin
/plugin install debate-consensus

# 3. Configure API key (one-time setup)
cd ~/.claude/plugins/debate-consensus/mcp-servers/debate-consensus
cp .env.example .env
nano .env  # Add your OpenRouter API key

Get your free API key: OpenRouter

Your First Debate

/debate "What's the best database for real-time analytics?"

---

📋 Commands

Command	Purpose	Speed
/debate <question>	Full multi-model debate	30-120s
/consensus <topic>	Quick 3-model consensus	10-30s
/verify <statement>	Adversarial verification	20-60s
/debate-rapid <question>	Ultra-fast answer	3-10s
/debate-history [limit]	View past debates	Instant

---

🆕 v2.0 Expert Mode (Experimental)

NEW: Dynamic Roles & Emergent Flow - opt-in experimental feature!

Standard Mode (Default - v1.0)

/debate "What's the best database for analytics?"

• Fixed expert roles (Architecture, Testing, Algorithms, Integration)
• 2-round structured debate
• Fast (30-60s), reliable (90.2% pass rate)
• Cost-effective ($0.05 average)

Expert Mode (Opt-in - v2.0)

/debate "What's the best database for analytics?" --mode expert

• Dynamic expert selection - Gemini analyzes question and selects optimal experts
• Emergent debate flow - 3-10 rounds of adaptive conversation
• Deeper insights - Models collaborate and build on each other's ideas
• Smart role assignment - 7 expertise types (security, performance, architecture, etc.)

Trade-offs:

• ⏱️ Slower: 30-120s (vs 30-60s)
• 💰 More expensive: $0.10-1.00 (vs $0.05) - 5-10x cost
• 🎯 Higher quality: +10-20% better results (estimated)
• 🔬 Experimental: Being validated with real usage

Try it when:

• ✅ Complex, multi-faceted questions
• ✅ Need deep domain expertise
• ✅ Quality more important than speed/cost
• ✅ Exploring critical architectural decisions

Use standard mode when:

• ⚡ Quick decisions needed
• 💰 Budget-conscious
• 🎯 Straightforward questions

Documentation:

• v2.0 POC Guide - How v2.0 works
• Implementation Report - Technical details

---

📚 Documentation

User Documentation

• 🔧 Installation Guide - Complete setup guide
• 💡 Usage Examples - Real-world scenarios
• 🐛 Troubleshooting - Common issues & solutions
• 📚 API Reference - Complete API documentation
• 🧪 Beta Testing - How to participate

Developer Documentation

• 📖 Conversion Plan - Detailed conversion roadmap
• 📊 Project Summary - Technical overview
• 🧪 Test Results - Comprehensive test reports
• 🎯 Phase Completion - Phase 1-4 completion reports

---

🏗️ Project Status

Current Version: v2.0.0 (v1.0 stable + v2.0 experimental)

Successfully converted from standalone mcp-debate-consensus MCP server.

v1.0 Progress (Stable)

• Phase 0: Planning & Conversion Plan (100%)
• Phase 1: Foundation - Plugin manifest & commands (100%)
• Phase 2: Advanced Features - Agents, hooks, caching (100%)
• Phase 3: Testing & Documentation (100%)
  • 30+ integration tests (100% pass rate)
  • 1,750+ lines of documentation
  • API reference, installation guide, troubleshooting
• Phase 4: Distribution (100% Complete)
  • Marketplace entry created
  • Security hardened (git history cleaned)
  • Repository polished and published
  • Ready for public release

See completion reports: Phase 1 | Phase 2 | Phase 3 | Phase 4

v2.0 Progress (Experimental)

• Phase 2.1 POC: Foundation (100% Complete)
  • Intelligent Coordinator (Gemini-powered)
  • Dynamic expert selection (7 expertise types)
  • Emergent 3-round debate orchestration
  • Safety controls (cost/time/round limits)
  • POC test infrastructure
• Phase 2.2: Named Communication (Pending validation)
• Phase 2.3: Full Emergent Flow (Pending validation)
• Phase 2.4: Production Hardening (Pending validation)

v2.0 Status: Opt-in experimental feature, being validated with real usage

See v2.0 docs: POC Guide | Implementation Report | Planning Docs

---

🎉 v1.0 Stable + v2.0 Experimental - Ready for Use!

---

💝 Support This Project

If you find this plugin helpful, consider supporting its development!

Why donate?

• ☕ Keep the developer caffeinated
• 🚀 Fund continued development and new features
• 🐛 Faster bug fixes and support
• 📚 More documentation and examples
• 💯 100% of donations go to improving this plugin

Even $5 helps! It shows you value multi-model AI consensus and want to see this project thrive.

Alternative Support

Can't donate? Here's how else you can help:

• ⭐ Star this repo on GitHub
• 🐦 Share on Twitter/X
• 📝 Write a blog post about it
• 🐛 Report bugs and suggest features
• 💬 Answer questions in Discussions
• 🔧 Contribute code or documentation

---

🤝 Contributing

Contributions welcome! This is an open-source project.

Ways to Contribute

1. Code: Submit PRs for new features or bug fixes
2. Documentation: Improve guides and examples
3. Testing: Report bugs and test new features
4. Ideas: Suggest improvements in Discussions

See CONTRIBUTING.md for guidelines (coming soon).

---

📝 License

MIT License

---

🙏 Credits

• Built on MCP by Anthropic
• Powered by OpenRouter
• Based on mcp-debate-consensus

---

⭐ Star this repo if you're excited about multi-model AI debates!