NEST - NANDA Sandbox and Testbed

A production-ready framework for deploying and managing specialized AI agents with seamless agent-to-agent communication and intelligent discovery.

NEST (NANDA Sandbox and Testbed) is part of Project NANDA (Networked AI Agents in Decentralized Architecture) - a comprehensive ecosystem for intelligent agent deployment and coordination.

Key Features

• Intelligent Agents: Deploy specialized AI agents powered by Claude LLM
• A2A Communication: Agents can find and communicate with each other using @agent-id syntax
• Cloud Deployment: One-command deployment to AWS EC2 with automatic setup
• Index Integration: Automatic registration with NANDA agent Index
• Scalable: Deploy single agents or 10+ agents per instance
• Production Ready: Robust error handling, health checks, and monitoring

Quick Start

Deploy a Single Agent

bash scripts/aws-single-agent-deployment.sh \
  "agent-id" \                    # Unique identifier
  "your-api-key" \                # Anthropic Claude API key
  "Agent Name" \                  # Display name
  "domain" \                      # Field of expertise
  "specialization" \              # Role description
  "description" \                 # Detailed agent description
  "capabilities" \                # Comma-separated capabilities
  "smithery-api-key" \            # Smithery API key (optional)
  "registry-url" \                # Agent registry URL (optional)
  "mcp-registry-url" \            # MCP registry URL (optional)
  "port" \                        # Port number 
  "region" \                      # AWS region 
  "instance-type"                 # EC2 instance type 

Example:

bash scripts/aws-single-agent-deployment.sh \
  "furniture-expert" \
  "sk-ant-api03-..." \
  "Furniture Expert" \
  "furniture and interior design" \
  "knowledgeable furniture specialist" \
  "I help with furniture selection and interior design" \
  "furniture,interior design,decor" \
  "smithery-key-xxxxx" \
  "http://registry.chat39.com:6900" \
  "https://your-mcp-registry.ngrok-free.app" \
  "6000" \
  "us-east-1" \
  "t3.micro"

Deploy Multiple Agents (10 per instance)

bash scripts/aws-multi-agent-deployment.sh \
  "your-api-key" \
  "scripts/agent_configs/group-01-business-and-finance-experts.json" \
  "smithery-key-xxxxx" \
  "http://registry.chat39.com:6900" \
  "https://your-mcp-registry.ngrok-free.app" \
  "us-east-1" \
  "t3.xlarge"

Architecture

NEST/
├── nanda_core/                        # Core framework
│   ├── core/
│   │   ├── adapter.py                  # Main NANDA adapter
│   │   ├── agent_bridge.py             # A2A communication
│   │   └── registry_client.py          # Registry integration
│   ├── discovery/                      # Agent discovery system
│   └── telemetry/                      # Monitoring & metrics
├── examples/
│   ├── nanda_agent.py                  # Main agent implementation
│   └── agent_configs.py                # Agent personalities
├── scripts/
│   ├── aws-single-agent-deployment.sh     # Single agent deployment
│   ├── aws-multi-agent-deployment.sh      # Multi-agent deployment
│   ├── deploy-agent.sh                    # Deploy to existing server
│   └── agent_configs/                     # Agent configuration files
│       ├── 100-agents-config.json            # 100 agent personalities
│       └── group-*.json                      # Agent group configs
└── README.md

Agent Communication

A2A Communication

Agents can communicate with each other using the @agent-id syntax:

# Test A2A communication
curl -X POST http://agent-ip:{PORT}/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "@other-agent-id Can you help with this task?",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "test123"
  }'

MCP (Model Context Protocol) Integration

Agents can discover and execute tools from MCP servers using the #registry:server-name syntax:

Smithery MCP Servers:

# Query Smithery registry servers
curl -X POST http://agent-ip:{PORT}/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "#smithery:@{mcp_server_name} get current weather in NYC",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "mcp-test"
  }'

NANDA MCP Servers:

# Query NANDA registry servers
curl -X POST http://agent-ip:{PORT}/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "#nanda:nanda-points get my current points balance",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "nanda-test"
  }'

The agent will automatically:

1. Discover the MCP server from the appropriate registry
2. Connect to the server and get available tools
3. Use Claude to intelligently select and execute the right tools
4. Return formatted results

Available Agent Groups

Pre-configured agent groups for quick deployment:

• Business & Finance: Financial analysts, investment advisors, business strategists
• Technology & Engineering: Software engineers, DevOps specialists, AI researchers
• Creative & Design: Graphic designers, content creators, brand strategists
• Healthcare & Life Sciences: Medical researchers, health informatics specialists
• Education & Research: Academic researchers, curriculum developers
• Media & Entertainment: Journalists, content producers, social media managers
• Environmental & Sustainability: Climate scientists, sustainability consultants
• Social Services: Community organizers, social workers, policy analysts
• Sports & Recreation: Fitness trainers, sports analysts, nutrition experts
• Travel & Hospitality: Travel planners, hotel managers, tour guides.

Prerequisites

• AWS CLI configured with credentials
• Anthropic API key
• Python 3.8+ (for local development)

Monitoring

Each deployed agent includes:

• Health checks on startup
• Automatic registry registration
• Process management with supervisor
• Individual logs for debugging
• Performance metrics collection

Configuration

Environment Variables

• ANTHROPIC_API_KEY: Your Claude API key
• AGENT_ID: Unique agent identifier
• AGENT_NAME: Display name for the agent
• REGISTRY_URL: NANDA registry endpoint
• PUBLIC_URL: Agent's public URL for A2A communication
• PORT: Port number for the agent server

Agent Personality Configuration

Agents are configured with:

• Domain: Primary area of expertise
• Specialization: Specific role and personality
• Description: Detailed background for system prompt
• Capabilities: List of specific skills and knowledge areas

Testing

Test Single Agent

curl -X POST http://agent-ip:{PORT}/a2a \
  -H "Content-Type: application/json" \
  -d '{"content":{"text":"Hello! What can you help me with?","type":"text"},"role":"user","conversation_id":"test123"}'

Test A2A Communication

curl -X POST http://agent-a-ip:{PORT}/a2a \
  -H "Content-Type: application/json" \
  -d '{"content":{"text":"@agent-b-id Please help with this task","type":"text"},"role":"user","conversation_id":"test123"}'

Production Deployment

For production use:

1. Single Agent: Use t3.micro for cost-effective single agent deployment
2. Multi-Agent: Use t3.xlarge or larger for 10+ agents per instance
3. High Availability: Deploy across multiple AWS regions
4. Monitoring: Enable CloudWatch logs and metrics
5. Security: Use proper security groups and VPC configuration

License

MIT License - see LICENSE file for details.

Contributing

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

Support

For issues and questions:

• Create an issue in this repository
• Check the documentation in /scripts/README.md
• Review example configurations in /scripts/

---

Built by Project NANDA
Visit Project NANDA | NEST Repository