Muxi

🚀 MUXI.ai

MUXI is an ambitious project that offers a powerful open-source framework for building multi-AI agent systems with complex orchestration capabilities. It provides persistent memory, standardized communication protocols, and chain-of-thought tracking, and seamlessly integrates multiple interfaces, including CLI, API/WS, Web UI, MCP, and WebRTC. The project aims to provide developers with a comprehensive platform to create advanced AI applications with real-time communication and MCP server integration, enabling interactions between complex agents in a modular and scalable environment.

📚 Read the Whitepaper  │  ⭐️ Star this Repository

⚠️ Important Note

Ensure to read the 📚 whitepaper to understand "everything" about the project and ⭐️ star this repository to follow its development.

✨ Features

• 🤖 Multi-Agent Support: Create and manage multiple AI agents with different capabilities.
• 🧠 Memory System: Short-term and long-term memory for context-aware interactions.
  • Use FAISS for short-term buffered memory.
  • Use SQLite with sqlite-vec for lightweight deployment.
  • Use PostgreSQL and pgvector for scalable long-term memory.
  • Implement a user-level profile-based memory system with Memobase.
  • Automatically extract user information from conversations and store it in context memory.
• 🔌 MCP Client Integration: Connect to external services via the Model Context Protocol service.
  • Support HTTP + SSE transport for web-based MCP servers.
  • Support command-line transport for local executable servers.
  • Feature an adaptive reconnection mechanism and exponential backoff strategy.
  • Have comprehensive error handling and diagnostic capabilities.
  • Support cancellation of long-running operations.
• 🌟 MCP Server Implementation: Provide services by exposing your agents as MCP-compatible servers.
  • Use SSE-based server endpoints for MCP host integration.
  • Discover tools from agent capabilities.
  • Share authentication with REST API.
  • Provide bridging packages for non-SSE clients.
• 🔄 Agent-to-Agent Communication: Enable direct interaction and collaboration between agents.
• 📦 Modular Design: Highly decoupled components for easy extension and maintenance.
• 🛠️ Multi-Platform Support: Compatible with various operating systems and hardware architectures.

✨ Core Features

• Persistent Memory Management: Ensure the persistence and reliability of agent memory through advanced data storage technologies.
• Chain-of-Thought Tracking: Enable traceability and transparency in complex problem-solving processes.
• High Availability Design: Ensure the system's stable operation in critical task scenarios.
• Extensible Plugin Architecture: Support flexible function extension through plugins and be compatible with various third-party services.

📚 Documentation

Project Structure

MUXI
├── core/          # Core algorithms and basic components
│   ├── memory/
│   │   └── ...     # Memory management module
│   └── protocol/
│       └── ...     # Communication protocol implementation
├── interface/    # Multiple interface adapters
│   ├── cli/
│   │   └── ...     # Command-line tools
│   ├── webui/
│   │   └── ...     # Web interface components
│   └── mcp/
│       └── ...     # MCP protocol adapter
└── docs/          # Documentation and development guides
    └── contributing.md  # Contributor guidelines

📄 License

During the Development Phase

During the development phase, this project adopts a dual-license model:

1. Non-commercial Use: The AGPL-3.0 license is used to ensure the healthy development of the open-source community.
2. Commercial Use: A commercial license is available. Please contact support@muxi.com for details.

After the Release of Version 1.0

When the project reaches the v1.0 version, it will switch to a more permissive open-source license:

• LGPL-3.0 or a similar license, allowing free use for both non-commercial and commercial purposes.

🔧 How to Contribute

Contributions are welcome! Please read our Contributor Guidelines to understand the code specifications, development environment configuration, and the process of submitting a Pull Request.

Acknowledgments

Thank you to all open-source projects for their support of this framework ❤️