Agents-MCP-Hackathon/mco-protocol

🚀 MCO Protocol: Real Agent Orchestration Demo

agent-demo-track

🎯 Hackathon Submission: Track 3 - Agentic Demo

This is a complete Gradio app showcasing the power of MCO (Model Configuration Orchestration) - the missing orchestration layer for AI agents. Experience real AutoGPT-style agents being orchestrated by an actual MCO MCP server with progressive revelation and persistent memory.

🎮 Live Demo

🔗 [Try the Live Demo](https://huggingface.co/spaces/Agents-MCP-Hackathon/mco-protocol-real-orchestration)

📹 Video Overview

🎥 [Watch the App Overview Video](https://youtu.be/VIDEO_ID_HERE) (Video explaining usage and purpose)

✨ What Makes This Demo Special

🤖 Real AutoGPT Agent + Real MCO Orchestration

• —Generic AutoGPT Agent: Standard agent with tools (not MCO-specific)
• —Real MCO MCP Server: Actual subprocess orchestration (Docker-enabled)
• —Progressive Revelation: Watch MCO strategically inject context at optimal moments
• —Live Modal Credits: Real LLM inference with Claude via Modal API

🎛️ Single-Page Experience

• —Side-by-Side Logs: See Claude's <thinking> process and MCO orchestration in real-time
• —Visual SNLP Generator: Create MCO workflow files without learning syntax
• —Downloadable Results: Export generated SNLP configurations

🛠️ Technical Architecture

🚀 Key Features Demonstrated

1. Real Agent Orchestration

• —Agent receives generic instruction: "Use the MCO server tool to get a code review task"
• —MCO handles all orchestration logic through existing workflow files
• —Progressive revelation injects features and styles at optimal moments
• —Agent completes task autonomously with structured workflow

2. Live MCP Server Integration

• —Docker container runs real Node.js + Python environment
• —MCO MCP server starts as actual subprocess
• —Real stdio communication using MCP protocol
• —Proper AsyncExitStack resource management

3. Visual SNLP Configuration

• —Generate complete MCO workflow files visually
• —Edit workflow names and types through simple UI
• —Download generated mco.core, mco.sc, mco.features, mco.styles
• —No need to learn SNLP syntax

💡 The MCO Revolution

This demo showcases MCO Protocol - completing the "Agentic Trifecta":

• —📊 MCP: Data integration ("What can I access?")
• —🤝 A2P: Agent communication ("How do we coordinate?")
• —🎛️ MCO: Reliable orchestration ("How do we actually get things done?")

Before MCO vs After MCO

🔧 How to Use the Demo

1. 1.Agent Demo:
2. 2.Enter any instruction for the agent
3. 3.Watch Claude's thinking process in real-time
4. 4.See MCO orchestration logs show progressive revelation
5. 5.Agent automatically uses MCO server tool when appropriate

1. 1.SNLP Generator:
2. 2.Set workflow name and type
3. 3.Generate complete MCO configuration files
4. 4.Download all files as a bundle
5. 5.Use in your own MCO projects

🏗️ Technical Implementation

Docker Setup

FROM node:18-slim
RUN npm install -g @paradiselabs/mco-protocol
# Python + dependencies
COPY . .
CMD ["python", "app.py"]

Agent Architecture

class GenericAutoGPTAgent:
    # NOT MCO-specific - just a standard agent with tools
    async def process_instruction(self, instruction):
        # Agent decides what to do, calls MCO as needed
        if "mco" in instruction.lower():
            await self.call_mcp_tool("mco-orchestration", "get_next_directive", {})

Real MCP Connection

server_params = StdioServerParameters(
    command="npx",
    args=["@paradiselabs/mco-protocol", "--config-dir", "/app/workflow"]
)
# Real subprocess with AsyncExitStack

🌟 Why This Demo Matters

This demo proves that MCO solves the reliability problem that has plagued autonomous agents since AutoGPT:

1. 1.Progressive Revelation: No more context overload
2. 2.Structured Workflows: Clear steps and success criteria
3. 3.MCP Integration: Works with any framework in one line
4. 4.Production Ready: Real implementation, not just a concept

🚀 Get Started with MCO

# Install MCO Protocol
npm install -g @paradiselabs/mco-protocol

# Add to any MCP-enabled framework
{
  "mcpServers": {
    "mco-orchestration": {
      "command": "npx",
      "args": ["@paradiselabs/mco-protocol", "--config-dir", "./workflow"]
    }
  }
}

📦 Project Structure

/
├── app.py              # Main Gradio application
├── Dockerfile          # Docker container setup
├── requirements.txt    # Python dependencies
├── modal_llm_client.py # Modal API integration
├── workflow/           # MCO configuration files
│   ├── mco.core       # Persistent memory
│   ├── mco.sc         # Success criteria
│   ├── mco.features   # Strategic injections
│   └── mco.styles     # Strategic injections
└── README.md          # This file

🏆 Hackathon Achievement

This demo showcases:

• —✅ Real MCP Server: Actual MCO protocol implementation
• —✅ Creative Agent Use: Novel orchestration approach
• —✅ Production Ready: Deployed and functional
• —✅ Educational Value: Clear demonstration of agent reliability
• —✅ Technical Innovation: First orchestration protocol for MCP ecosystem

🔗 Links

• —🎮 [Live Demo](https://huggingface.co/spaces/Agents-MCP-Hackathon/mco-protocol-real-orchestration)
• —📹 [Video Overview](https://youtu.be/VIDEO_ID_HERE)
• —📦 [MCO NPM Package](https://www.npmjs.com/package/@paradiselabs/mco-protocol)
• —🐙 [GitHub Repository](https://github.com/paradiselabs-ai/MCO-Protocol)
• —💬 [Discord Community](https://discord.gg/uQ69vc4Agc)

agent-demo-track | Built with ❤️ for MCP Hackathon 2025 | Paradise Labs