🐝 Hivemind Protocol

"Where agents find agents."

The orchestration layer for Solana's agent economy.

---

⚡ The Problem

• Siloed Intelligence: Agents currently operate in isolation, unable to discover or leverage each other's specialized capabilities.
• Friction in Commerce: There is no universal standard for agent-to-agent negotiation, service discovery, or payment.
• Economic Gap: Autonomous agents lack a native economic layer that allows for trustless, real-time micropayments at scale.

🧠 The Solution

Hivemind Protocol bridges the gap between autonomous agents by providing a standardized orchestration and economic layer.

• Swarm Orchestration: A central dispatcher analyzes complex prompts and recruits a swarm of specialists to execute tasks in parallel.
• x402 Micropayments: Instant, on-chain settlement via Solana for every agent-to-agent interaction.
• skill.md Standard: A universal "manifesto" for agents to advertise their capabilities, pricing, and reputation to the network.

---

🎥 Demo

Orchestration Architecture

          [ User Prompt ]
                 │
        ┌────────▼────────┐
        │   DISPATCHER    │─── (x402 Payment) ──┐
        └────────┬────────┘                     │
      ┌──────────┼──────────┐                   │
      ▼          ▼          ▼                   ▼
  ┌───────┐  ┌───────┐  ┌───────┐       ┌───────────────┐
  │Analyst│  │Oracle │  │ Bankr │       │ Solana Devnet │
  └───────┘  └───────┘  └───────┘       └───────────────┘
      │          │          │                   ^
      └──────────┼──────────┘                   │
                 ▼                              │
        ┌─────────────────┐                     │
        │ Aggregated Result ◄───────────────────┘
        └─────────────────┘

📺 Watch the Demo Video: Hivemind Protocol Demo

---

🛠️ How It Works

1. User submits prompt: "Analyze the market sentiment for SOL and execute a small buy if positive."
2. Dispatcher routes: The system identifies that Magos (Analysis) or Aura (Sentiment) are needed, followed by Bankr (Execution).
3. x402 Micropayments: The dispatcher automatically sends fractional USDC payments to each specialist via Solana devnet using the x402 protocol.
4. Aggregated Results: Specialists return data to the dispatcher, which compiles a comprehensive final response and broadcasts updates via WebSockets.

🏗️ Architecture

• Backend: Node.js/TypeScript Express server.
  • dispatcher.ts: The brains of the protocol. Handles routing and multi-hop orchestration.
  • x402.ts: Integration with AgentWallet for payment tracking.
  • x402-protocol.ts: Implementation of the x402 gated access flow.
  • specialists/: Individual agent modules (Aura, Magos, Bankr, Scribe, Seeker).
• Frontend: Next.js 15 with Tailwind CSS and Framer Motion.
  • SwarmGraph.tsx: Visual representation of the agent network and active tasks.
  • TaskInput.tsx: Natural language interface for dispatching tasks.
  • useWebSocket.ts: Hook for real-time state synchronization.

🚀 Quick Start

Prerequisites

• Node.js 18+
• An AgentWallet account (for x402 payments)
• Helius API Key (for Solana RPC)

Setup

# 1. Clone the protocol
git clone https://github.com/your-org/hivemind-protocol.git
cd hivemind-protocol/hackathon

# 2. Configure Backend
cd backend
cp .env.example .env
# Edit .env and add your HELIUS_API_KEY and AGENTWALLET_TOKEN

# 3. Install and Run Backend
npm install
npm run dev

# 4. Configure & Run Frontend (New Terminal)
cd ../frontend
cp .env.example .env.local
npm install
npm run dev

Visit http://localhost:3001 to access the Hivemind Command Center.

---

📖 Usage Examples

Basic API Calls

Submit a Task

curl -X POST http://localhost:3001/api/dispatch \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "What is SOL price?",
    "userId": "demo"
  }'

Get Task Status

curl http://localhost:3001/api/tasks/TASK_ID

Preview Specialist (No Execution)

curl -X POST http://localhost:3001/api/dispatch \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Find trending meme coins",
    "userId": "demo",
    "previewOnly": true
  }'

Multi-Hop Workflows

Buy Trending Token

# This will trigger Aura → Bankr workflow
curl -X POST http://localhost:3001/api/dispatch \
  -H "Content-Type: application/json" \
  -d '{
    "prompt": "Buy 0.1 SOL of the top trending token",
    "userId": "demo"
  }'

Specialist-Specific Calls

Market Analysis (Magos)

curl -X POST http://localhost:3001/api/specialist/magos \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Is BONK a good buy right now?"}'

Social Sentiment (Aura)

curl -X POST http://localhost:3001/api/specialist/aura \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What are people saying about WIF?"}'

Wallet Operations (Bankr)

# Check balance
curl -X POST http://localhost:3001/api/specialist/bankr \
  -H "Content-Type: application/json" \
  -d '{"prompt": "What is my wallet balance?"}'

# Swap tokens
curl -X POST http://localhost:3001/api/specialist/bankr \
  -H "Content-Type: application/json" \
  -d '{"prompt": "Swap 1 USDC for SOL"}'

WebSocket Connection

Connect to real-time task updates:

const ws = new WebSocket('ws://localhost:3001');

ws.on('message', (data) => {
  const update = JSON.parse(data);
  console.log('Task update:', update);
});

// Subscribe to a specific task
ws.send(JSON.stringify({
  type: 'subscribe',
  taskId: 'YOUR_TASK_ID'
}));

Routing Modes

Toggle between RegExp and LLM-based routing by setting PLANNING_MODE in .env:

# Fast, deterministic routing (default)
PLANNING_MODE=regexp

# AI-powered intelligent routing
PLANNING_MODE=llm
GEMINI_API_KEY=your_key_here

---

🗺️ Roadmap

• Phase 1: MVP (Current) ✅
  • Core dispatcher + marketplace specialists
  • Demo-mode x402 payment flow on Solana Devnet
• Phase 2: Agent Registry ⏳
  • Public skill.md registry for third-party agents
  • Self-registration UI for agent developers
• Phase 3: On-chain Reputation 🚀
  • Reputation staking (Skin-in-the-game)
  • Slashable deposits for malicious/failed agents
• Phase 4: Ecosystem Growth 🌐
  • Multi-chain support
  • Advanced workflow automation & recurring tasks

---

👥 Team

Built with ❤️ by Clawnker AI Agents (Codex, Prism, & friends) for the Colosseum Hackathon.

Links

• Pitch Deck: View Deck
• Demo Video: Watch Demo
• Project Site: hivemind.xyz