NEST Quickstart: Deploy Your First NANDA Agent

What is NEST?

NEST (NANDA Sandbox and Testbed) is the official framework for deploying and managing AI agents within the NANDA ecosystem. Built as part of Project NANDA, NEST provides:

• One-command deployment to AWS EC2 — single agents or multi-agent clusters
• Built-in A2A communication — agents find and talk to each other using @agent-id syntax
• MCP integration — agents discover and execute tools from MCP servers
• Automatic NANDA Index registration — agents are discoverable globally as soon as they're deployed
• Claude-powered reasoning — agents use Anthropic's Claude for intelligent task handling

Prerequisites

Before you start, you'll need:

Step 1: Deploy a Single Agent

Clone the NEST repository and deploy your first agent with a single command:

git clone https://github.com/projnanda/NEST.git
cd NEST

bash scripts/aws-single-agent-deployment.sh \
  "my-first-agent" \
  "sk-ant-api03-YOUR_KEY" \
  "My First NANDA Agent" \
  "general assistant" \
  "helpful AI assistant" \
  "A general-purpose AI agent for testing" \
  "question-answering,summarization,analysis"

This command provisions an EC2 instance, installs dependencies, deploys the agent, registers it with the NANDA Index, and starts the health monitoring system. Your agent is globally discoverable within seconds.

Step 2: Test Agent Communication

Once deployed, test your agent directly:

curl -X POST http://AGENT_IP:6000/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "Hello! What can you help me with?",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "test-001"
  }'

To test agent-to-agent communication, deploy a second agent and use the @agent-id syntax:

curl -X POST http://AGENT_A_IP:6000/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "@my-second-agent Can you help analyze this data?",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "a2a-test"
  }'

Agent A will automatically discover Agent B through the NANDA registry, establish a communication channel, and relay the task — all via the A2A protocol.

Step 3: MCP Tool Integration

NEST agents can discover and execute tools from MCP servers using a simple hash syntax. Two registries are supported out of the box:

Smithery MCP servers — access the public Smithery registry:

curl -X POST http://AGENT_IP:6000/a2a \
  -H "Content-Type: application/json" \
  -d '{
    "content": {
      "text": "#smithery:@weather-server get current weather in NYC",
      "type": "text"
    },
    "role": "user",
    "conversation_id": "mcp-test"
  }'

NANDA MCP servers — access the NANDA ecosystem registry:

curl -X POST http://AGENT_IP:6000/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 automatically discovers the MCP server from the appropriate registry, connects, uses Claude to select the right tools, and returns formatted results. This bridges NANDA's agent discovery with Anthropic's Model Context Protocol.

Step 4: Multi-Agent Deployment

Ready to scale? NEST supports deploying 10+ agents per instance using pre-configured agent groups:

bash scripts/aws-multi-agent-deployment.sh \
  "sk-ant-api03-YOUR_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"

NEST ships with pre-configured agent groups spanning Business & Finance, Technology & Engineering, Healthcare, Creative & Design, Education, Media, Environmental, and more — up to 100 agent personalities ready to deploy.

Architecture at a Glance

NEST's modular architecture separates concerns cleanly:

• nanda_core/ — Core framework: NANDA adapter, A2A agent bridge, and registry client
• nanda_core/discovery/ — Agent discovery system for finding peers across the network
• nanda_core/telemetry/ — Monitoring, health checks, and performance metrics
• examples/ — Reference agent implementation and personality configurations
• scripts/ — Deployment automation and agent group configs

Every deployed agent includes automatic health checks, registry registration, process management via supervisor, individual logs, and performance metrics collection — production-ready out of the box.