UNPKG

@paradiselabs/mco-protocol

Version:

MCO (Model Configuration Orchestration) MCP Server & Configuration Tool

310 lines (229 loc) 9.55 kB
# 🚀 MCO Protocol: The Missing Orchestration Layer for MCP <div align="center"> [![NPM Version](https://img.shields.io/npm/v/@paradiselabs/mco-protocol.svg)](https://www.npmjs.com/package/@paradiselabs/mco-protocol) [![License](https://img.shields.io/badge/License-MIT-blue.svg)](https://opensource.org/licenses/MIT) [![Hackathon](https://img.shields.io/badge/🏆-MCP%20Hackathon%202025-gold)](https://huggingface.co/Agents-MCP-Hackathon) **Completing the Agentic Trifecta: MCP + A2P + MCO** *Transform unreliable agents into structured, autonomous workflows with progressive revelation and persistent memory.* [🎮 **Live Demo**](https://huggingface.co/spaces/paradiselabs/mco-protocol-demo) • [📦 **NPM Package**](https://www.npmjs.com/package/@paradiselabs/mco-protocol) • [📖 **Documentation**](https://github.com/paradiselabs-ai/MCO-Protocol/blob/main/docs) </div> ## 🌟 The Agentic Trifecta ```mermaid graph TB subgraph "The Foundation of Autonomous AI" MCP[📊 MCP<br/>Model Context Protocol<br/><i>Data Integration</i>] A2P[🤝 A2P<br/>Agent-to-Agent Protocol<br/><i>Communication</i>] MCO[🎛️ MCO<br/>Model Configuration Orchestration<br/><i>Reliable Orchestration</i>] end MCP --> AGENT[🤖 Autonomous Agent] A2P --> AGENT MCO --> AGENT AGENT --> RESULT[✨ Production-Ready<br/>Autonomous AI] style MCO fill:#667eea,stroke:#333,stroke-width:3px,color:#fff style RESULT fill:#2ecc71,stroke:#333,stroke-width:2px,color:#fff ``` **Why MCO is Essential:** - 📊 **MCP** connects agents to data sources → *"What can I access?"* - 🤝 **A2P** enables agent communication → *"How do we coordinate?"* - 🎛️ **MCO** ensures reliable execution → *"How do we actually get things done?"* ## 🎯 The Problem MCO Solves Traditional autonomous agents (AutoGPT, BabyAGI) suffer from: - 🔄 **Endless loops** and failed executions - 🧠 **Context overload** leading to poor decisions - 🎯 **Lack of focus** on core objectives - 📉 **Unpredictable reliability** in production ## 💡 The MCO Solution: Progressive Revelation ```mermaid graph LR subgraph "Traditional Approach" T1[Agent] --> T2[Everything at Once<br/>📚 Core + Features + Styles + Context] T2 --> T3[❌ Overwhelmed<br/>Loops & Failures] end subgraph "MCO Progressive Revelation" M1[Agent] --> M2[🧠 Persistent Memory<br/>Core + Success Criteria] M2 --> M3[⚡ Step 1: Focus on Core] M3 --> M4[✨ Step 2: + Features Injection] M4 --> M5[🎨 Step 3: + Styles Injection] M5 --> M6[✅ Reliable Completion] end style T3 fill:#e74c3c,color:#fff style M6 fill:#2ecc71,color:#fff ``` ## 🛠️ How MCO Works ### SNLP (Syntactic Natural Language Programming) MCO uses a revolutionary programming language that combines structured syntax with natural language: ```yaml # mco.core - Always in persistent memory @workflow "Research Assistant" >NLP An AI assistant that conducts autonomous research with reliability. @data: topic: "AI Agent Orchestration" findings: [] @agents: researcher: steps: - "Research the topic thoroughly" - "Analyze patterns and insights" - "Create comprehensive report" # mco.features - Injected at 33% progress @feature "Data Visualization" >NLP Create charts and graphs when appropriate to enhance understanding. # mco.styles - Injected at 66% progress @style "Professional Formatting" >NLP Use clear headings, bullet points, and executive summary format. ``` ### Orchestration Flow ```mermaid sequenceDiagram participant AF as Agent Framework participant MCO as MCO MCP Server participant SNLP as SNLP Files Note over AF,SNLP: Progressive Revelation in Action AF->>MCO: start_orchestration() MCO->>SNLP: Load mco.core + mco.sc MCO-->>AF: orchestration_id AF->>MCO: get_next_directive() Note right of MCO: Persistent Memory Only MCO-->>AF: Step 1 + Core Context AF->>MCO: complete_step(result) MCO->>MCO: Evaluate against success criteria AF->>MCO: get_next_directive() Note right of MCO: Strategic Injection MCO->>SNLP: Inject mco.features MCO-->>AF: Step 2 + Core + Features AF->>MCO: complete_step(result) AF->>MCO: get_next_directive() MCO->>SNLP: Inject mco.styles MCO-->>AF: Step 3 + Core + Features + Styles AF->>MCO: complete_step(result) MCO-->>AF: ✅ Workflow Complete ``` ## 🚀 Quick Start ### Installation ```bash npm install -g @paradiselabs/mco-protocol ``` ### Create Your First Workflow ```bash # Initialize new MCO project mco init my-research-assistant # Opens configuration tool in browser # Generates: mco.core, mco.sc, mco.features, mco.styles ``` ### Add to Any MCP-Enabled Framework ```json { "mcpServers": { "mco-orchestration": { "command": "npx", "args": ["@paradiselabs/mco-protocol", "--config-dir", "./my-research-assistant"] } } } ``` ### Use in Your Agent Framework ```python # Works with ANY MCP-enabled framework directive = mcp_client.call_tool("get_next_directive") result = execute_task(directive.instruction) mcp_client.call_tool("complete_step", step_id=directive.step_id, result=result) ``` ## 🎭 Live Demo **🎮 [Try the Interactive Demo](https://huggingface.co/spaces/paradiselabs/mco-protocol-demo)** Generate real SNLP configurations and see MCO in action with live MCP server simulation. ## 📊 Architecture Overview ```mermaid graph TB subgraph "MCO MCP Server" CLI[🖥️ CLI Interface<br/>mco init, serve, validate] CONFIG[🎛️ Configuration Tool<br/>Web-based SNLP Generator] PARSER[📝 SNLP Parser<br/>@markers + >NLP sections] ENGINE[⚡ Orchestration Engine<br/>Progressive Revelation] MCP[📡 MCP Tool Provider<br/>start_orchestration, get_next_directive] end subgraph "SNLP Files" CORE[🧠 mco.core<br/>Persistent Memory] SC[🎯 mco.sc<br/>Success Criteria] FEATURES[✨ mco.features<br/>Strategic Injection] STYLES[🎨 mco.styles<br/>Strategic Injection] end subgraph "Agent Frameworks" AUTOGPT[🤖 AutoGPT] CREWAI[👥 CrewAI] LANGGRAPH[🕸️ LangGraph] CUSTOM[⚙️ Custom Agents] end CLI --> CONFIG CONFIG --> CORE & SC & FEATURES & STYLES PARSER --> CORE & SC & FEATURES & STYLES PARSER --> ENGINE ENGINE --> MCP MCP <==> AUTOGPT MCP <==> CREWAI MCP <==> LANGGRAPH MCP <==> CUSTOM style MCO fill:#667eea,color:#fff style CORE fill:#e8f5e9 style SC fill:#e3f2fd style FEATURES fill:#fff3e0 style STYLES fill:#fce4ec ``` ## 🏆 Perfect for MCP Hackathon 2025 **Track 1: MCP Server Implementation** ✅ MCO exemplifies the future of MCP by: - 🔧 **Extending MCP's Vision**: Making agent orchestration as standardized as data access - 🎯 **Solving Real Problems**: Transforming unreliable agents into production-ready systems - 🚀 **Ready for Production**: Live NPM package, working implementation - 🌟 **Innovative Approach**: First orchestration protocol designed specifically for MCP ecosystem ## 📈 Before vs After ```mermaid graph LR subgraph "Before MCO" B1[🤖 Agent] --> B2[❓ Vague Prompts] B2 --> B3[🔄 Loops & Failures] B3 --> B4[😤 Manual Intervention] end subgraph "After MCO" A1[🤖 Agent] --> A2[🎛️ MCO Orchestration] A2 --> A3[📋 Structured Steps] A3 --> A4[✅ Reliable Completion] end style B3 fill:#e74c3c,color:#fff style A4 fill:#2ecc71,color:#fff ``` ## 🔗 Available MCP Tools MCO exposes these tools through the MCP protocol: | Tool | Description | Use Case | |------|-------------|----------| | `start_orchestration` | Initialize new workflow | Begin autonomous task | | `get_next_directive` | Get next step with context | Progressive execution | | `complete_step` | Mark step complete | Track progress | | `get_workflow_status` | Check progress | Monitoring | | `evaluate_against_criteria` | Quality assessment | Success validation | ## 🎨 CLI Commands ```bash mco init [project-name] # Create new MCO project mco validate [config-dir] # Validate SNLP files mco serve [config-dir] # Start MCP server mco templates # List available templates ``` ## 🤝 Contributing We welcome contributions! MCO is designed to become the standard for agent orchestration. 1. Fork the repository 2. Create your feature branch (`git checkout -b feature/amazing-feature`) 3. Commit your changes (`git commit -m 'Add amazing feature'`) 4. Push to the branch (`git push origin feature/amazing-feature`) 5. Open a Pull Request ## 📜 License MIT License - see [LICENSE](LICENSE) file for details. ## 🚀 Join the Revolution **MCO Protocol is live and ready to transform how you build autonomous agents.** - 📦 **Install**: `npm install -g @paradiselabs/mco-protocol` - 🎮 **Demo**: [Interactive Gradio Space](https://huggingface.co/spaces/paradiselabs/mco-protocol-demo) - 💬 **Discord**: [Join our community](https://discord.gg/uQ69vc4Agc) - 🐦 **Twitter**: [@paradiselabs_ai](https://twitter.com/paradiselabs_ai) --- <div align="center"> **🌟 Star this repository if MCO helps you build better agents! 🌟** *Made with ❤️ by [Paradise Labs](https://paradiselabs.co)* </div>