Lungo+ (Opytown)

🏆 SanD Hackathon 2026 Award Winner

About This Project

AGNTCY Track — Multi-Agent Systems Challenge

Lungo+ was developed as part of the 🤖 AGNTCY Track — Multi-Agent Systems Challenge, which introduces participants to multi-agent systems (MAS) using AGNTCY, an open-source project initiated at Cisco that is building foundational infrastructure for the Internet of Agents.

The challenge provides teams with the opportunity to explore and build upon AGNTCY's core components, including:

• Agent-to-Agent (A2A) communication protocols for seamless inter-agent messaging
• Multi-agent orchestration using directed graphs and LangGraph
• Interoperable transport layers (SLIM, NATS, and other protocols)
• Observability and metrics through OpenTelemetry integration
• Model Context Protocol (MCP) for external service integration

Teams can start at any challenge level based on their experience, from basic agent communication to advanced multi-agent coordination and custom protocol implementations.

Hackathon Award

🏆 This project was awarded at the SanD Hackathon 2026, recognizing its innovative approach to building a production-ready multi-agent system with comprehensive observability, flexible deployment options, and real-world applicability.

What is Lungo+?

Lungo+ is an interactive multi-agent system that combines the power of AGNTCY's infrastructure with an engaging game-based visualization interface. It demonstrates how autonomous agents can communicate, coordinate, and collaborate in real-time within a virtual town environment.

Preview of agentic workflow

The following diagram shows how agents (Buyer, Coffee Farm, Shipper, Accountant) communicate via the Transport: SLIM and Shared Memory (Semantic Translation Bus) in an A2A SLIM pattern—e.g. for Order Fulfilment or Coffee Buying conversations.

Key Features

🎮 Interactive Game-Based UI

• Real-time visualization of agent interactions in a virtual town
• Live agent movement and communication display
• Interactive chat interface for engaging with agents
• Visual representation of multi-agent coordination

🤖 Advanced Multi-Agent System

• Supervisor-Worker Architecture: Hierarchical agent coordination with supervisor agents managing worker agents
• Agent-to-Agent Communication: Direct A2A messaging using AGNTCY protocols
• Publish/Subscribe Patterns: Efficient broadcast and unicast messaging
• LangGraph Integration: Directed graph-based agent workflows for complex decision-making

🔧 Production-Ready Infrastructure

• Kubernetes Deployment: Full Helm charts for production-grade orchestration
• Docker Compose: Quick local development setup
• Multiple Deployment Options: Local Python, Docker, or Kubernetes
• Flexible LLM Provider Support: Integration with OpenAI, Azure, GROQ, NVIDIA NIM, and custom providers via LiteLLM

📊 Comprehensive Observability

• Advanced Observability Stack: Grafana, ClickHouse, and OpenTelemetry integration
• Metrics Computation Engine (MCE): Evaluate multi-agent system performance with metrics like:
  • Agent-to-Agent interactions
  • Tool utilization accuracy
  • Workflow efficiency
  • Goal success rate
  • Context preservation
• Real-time Trace Visualization: Monitor agent behavior and system performance
• Custom Dashboards: Pre-configured Grafana dashboards for system insights

🌐 External Service Integration

• MCP Server Integration: Real-time external data integration (weather API)
• Extensible Architecture: Easy addition of new agents and services
• Configurable Transport Layers: Support for multiple messaging protocols

New Implementations

This project extends the original AGNTCY framework with several innovative features:

1. Game-Based Visualization: An immersive UI that makes multi-agent interactions tangible and engaging
2. Hybrid Interaction Model: Combines movement-based interaction with traditional messaging
3. Real-Time Agent Coordination: Live updates showing agent decision-making and collaboration
4. Production Deployment: Complete Kubernetes infrastructure with Helm charts
5. Comprehensive Metrics: Advanced evaluation framework for multi-agent system performance
6. Flexible Architecture: Modular design allowing easy extension and customization

Getting Started

For detailed setup instructions, deployment options, and usage guides, please refer to the OpyTown README.

Quick Start

# Clone the repository
cd OpyTown

# Run with Docker Compose (recommended)
docker compose up

# Access the UI at http://localhost:3000
# Access Grafana at http://localhost:3001

Project Structure

• OpyTown/: Main project directory containing the multi-agent system implementation
  • agents/: Agent implementations (supervisors, farms, MCP servers)
  • frontend/: Interactive game-based UI
  • deployment/: Kubernetes Helm charts and deployment configurations
  • docs/: Additional documentation

Technologies Used

• AGNTCY: Core multi-agent infrastructure
• LangGraph: Agent workflow orchestration
• OpenTelemetry: Distributed tracing and observability
• Grafana & ClickHouse: Metrics visualization and storage
• Docker & Kubernetes: Containerization and orchestration
• React: Frontend UI framework
• FastAPI: Backend API framework
• LiteLLM: Multi-provider LLM integration

Acknowledgments

• AGNTCY Team at Cisco for creating the foundational infrastructure
• SanD Hackathon 2026 organizers and judges
• The open-source community for the amazing tools and frameworks

License

This project builds upon the AGNTCY open-source framework. Please refer to individual component licenses for details.