🛡️ DeFiGuard: Multi-Agent Risk Management System

AI-powered, multi-chain DeFi portfolio risk monitoring with autonomous agents

Monitor crypto portfolios across 12 chains, analyze risks using SingularityNET MeTTa AI, detect fraud, and receive real-time alerts via ASI:One chat.

---

📋 Table of Contents

• Demo Video
• Key Features
• Architecture
• Agents Overview
• Agent Addresses
• Quick Start
• Using DeFiGuard
• Risk Scoring Methodology
• Technologies Used
• Documentation
• Testing
• Security
• Contributing
• Acknowledgments
• Contact
• License

---

🎥 Demo Video

▶️ Watch Demo (4 minutes)

---

✨ Key Features

🔍 Multi-Chain Monitoring - Track portfolios across 12 EVM chains (Ethereum, BSC, Polygon, Arbitrum, Optimism, Avalanche, Base, Fantom, Gnosis, Moonbeam, Celo, Cronos)

🧠 AI Risk Analysis - SingularityNET MeTTa knowledge graphs with 50+ assets, 25+ risk rules, explainable decisions

💬 Natural Chat Interface - ASI:One integration with ASI-1 AI for conversational portfolio management

🚨 Real-Time Alerts - Instant notifications on risk escalation (Low → Critical)

🕵️ Fraud Detection - Honeypot detection, high tax identification, scam pattern matching via GoPlus Security API

📊 Market Intelligence - CoinGecko API integration with price change detection, volume spike identification

---

🏗️ Architecture

    ┌────────────────────────────────────────────────┐
    │                 USER INTERFACE                 │
    │                 (ASI:One Chat)                 │
    └──────────────────────┬─────────────────────────┘
                           │
                           ▼
              ┌────────────────────────┐
              │       Alert Agent      │ ◄────── Real-time Notifications
              │     (Chat Protocol)    │
              └────────┬───────────────┘
                       │
           ┌───────────┼───────────┬──────────┐
           ▼           ▼           ▼          ▼
    ┌──────────┐ ┌──────────┐ ┌─────────┐ ┌──────────┐
    │Portfolio │ │   Risk   │ │ Market  │ │  Fraud   │
    │ Monitor  │ │ Analysis │ │  Data   │ │ Detection│
    │  Agent   │ │  Agent   │ │  Agent  │ │  Agent   │
    └────┬─────┘ └────┬─────┘ └────┬────┘ └────┬─────┘
         │            │            │           │
         └────────────┴────────────┴───────────┘
                             │
                             ▼
    ┌────────────────────────────────────────────────┐
    │              MeTTa Knowledge Graph             │
    │           (50+ assets, 25+ risk rules)         │
    └────────────────────────────────────────────────┘   

---

🤖 Agents Overview

1. Alert Agent (Chat Interface)

Address: agent1q2zusjcsgluu9pkkf9g2fn5lyqnaf9jqlhm3smlhvqcd6nct46ezy2qvm2l

Capabilities:

• ASI:One chat interface with ASI-1 AI
• Portfolio registration forwarding
• Real-time alert delivery
• Natural language Q&A
• Commands: register, status, history, portfolio, chains, help

Tech: ASI:One Chat Protocol, ASI-1 mini model, React state (no localStorage)

📄 Doc

2. Portfolio Monitor Agent

Address: agent1qv3pywlds6n86hr55p7lpvncwtd22d25yfe82zjg5tgx325cg9dnqylzy6f

Capabilities:

• 12-chain multi-wallet tracking
• Native token balance monitoring
• 10-minute scan cycles (round-robin)
• ERC-55 checksum validation
• CoinGecko price integration (60s cache)

Limitations: 1 wallet/cycle, max 3 chains/scan, native tokens only

📄 Doc

3. Risk Analysis Agent (MeTTa-Powered)

Address: agent1qtrn82fz9tnspwudzrjr7mm9ncwvavjse5xcv7j9t06gajmdxq0yg38dyx5

Capabilities:

• SingularityNET MeTTa knowledge graph reasoning
• Multi-factor risk scoring (concentration 35%, volatility 45%, chain diversity 20%)
• Pattern matching for complex risk scenarios
• Explainable AI decisions (traceable to knowledge base)
• 5-level classification (Safe → Critical)

Tech: MeTTa v0.1.0, 50+ asset classifications, 25+ risk rules

📄 Doc | MeTTa Integration

4. Market Data Agent

Address: agent1qgwdvuucfhpvucqdru0gnrwc2zqf0ak5u24rvxua9flcazctmdvdsyrr8qq

Capabilities:

• CoinGecko API integration (10,000+ tokens)
• 5-minute price caching
• Volume spike detection
• Market cap tracking
• Anomaly alerts (>10% price change, >50% volume)

Tech: CoinGecko API v3, 5-min TTL cache, aiohttp

📄 Doc

5. Fraud Detection Agent

Address: agent1q0x3wcul6azlcu4wy5khce9hklav28ea9f8kjqcq649rs4jat5kc7zxarn6

Capabilities:

• GoPlus Security API integration
• Honeypot detection (cannot sell)
• High tax identification (>10%)
• Ownership analysis (renounced check)
• 20+ security checks with weighted scoring
• Risk classification (Safe → Critical)

Tech: GoPlus API, Honeypot.is API, 12 blockchain explorers

📄 Doc

---

🔗 Agent Addresses

Deployment Architecture

🐳 Docker Container (Railway Hosting)

All 5 agents run together in a single Docker container using Fetch.ai's Bureau framework, enabling efficient inter-agent communication and resource management.

☁️ Agentverse Registration

Only the Alert Agent is registered on Agentverse to serve as the public-facing interface for ASI:One chat interactions.

Agent	Address/Status	Deployment
Alert Agent	agent1q2zusjcsgluu9pkkf9g2fn5lyqnaf9jqlhm3smlhvqcd6nct46ezy2qvm2l	✅ Agentverse + Railway
Portfolio Monitor	Internal (Bureau)	🐳 Railway Container
Risk Analysis	Internal (Bureau)	🐳 Railway Container
Market Data	Internal (Bureau)	🐳 Railway Container
Fraud Detection	Internal (Bureau)	🐳 Railway Container

💡 Architecture Benefits:

• Unified Deployment: Single container simplifies DevOps and reduces hosting costs
• Low Latency: Internal Bureau messaging eliminates network overhead between agents
• Single Entry Point: Alert Agent handles all external communication via Agentverse
• Scalability: Railway auto-scaling manages all agents together

🔌 How to Connect: Users interact only with the Alert Agent address through ASI:One or Agentverse Chat. The Alert Agent coordinates with all internal agents automatically.

Agentverse Cloud

Each agent hosted independently on Agentverse and communicate via the Fetch.ai network.

Agent	Address	Status
Portfolio Monitor	agent1qvyvw79t54ysq7rdp5xfc9qtqkycrnvtqlwjncrqfj3v8ne3dhzfvkjmdrn	✅ Active
Risk Analysis	agent1q2stpgsyl2h5dlpq7sfk47hfnjqsw84kf6m40defdfph65ftje4e56l5a0f	✅ Active
Alert Agent	agent1qwzszgd7h0knxwdj2j73htqswatm87t0ftsj4d3wlzlv54kftx5gyu8ygun	✅ Active + Chat
Market Data	agent1qv7r47p6r8as5kw083fr36rjw4yjn3z59pe77x2hqeu7kgfh8leas7wxux8	✅ Active
Fraud Detection	agent1qvyvsyr93jp4detyrt7zy3hnvtrpu4jthy90nwv8uqpeunhywvdpgtglguc	✅ Active

💡 Independent Deployment Benefits:

• Distributed Architecture: Each agent runs in isolated Agentverse environments
• Network Communication: Agents discover and message each other via Fetch.ai Almanac
• Individual Scaling: Each agent can be updated/restarted independently
• No Single Point of Failure: Redundancy through distributed hosting

---

🔌 How to Connect: Regardless of deployment mode, users interact only with the Alert Agent address (agent1q2zusjcsgluu9pkkf9g2fn5lyqnaf9jqlhm3smlhvqcd6nct46ezy2qvm2l) through ASI:One or Agentverse Chat. The Alert Agent coordinates with all other agents automatically.

---

🚀 Quick Start

Prerequisites

python --version  # 3.10+
git --version

Installation

# Clone & setup
# 1. Clone repository
git clone https://github.com/DhanteyUD/DeFiGuard.git
cd DeFiGuard

# 2. Create virtual environment
python -m venv venv

# 3. Activate virtual environment
# Windows:
venv\Scripts\activate

# Mac/Linux:
source venv/bin/activate

# 4. Install dependencies
pip install -r requirements.txt

# Configure (copy & edit .env)
cp .env.example .env

# Add your AGENT_SEEDS and optional API keys

# Verify MeTTa
python verify_metta.py

# Run all agents
python main.py

---

📱 Using DeFiGuard

Via ASI:One Chat (Recommended)

1. Open ASI:One or Agentverse Chat
2. Search for Agent: DeFiGuard Alert Agent or address agent1q2zusjcsgluu9pkkf9g2fn5lyqnaf9jqlhm3smlhvqcd6nct46ezy2qvm2l
3. Start Chatting:

User: register 0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb ethereum,polygon

Agent: ✅ Portfolio Registered!
       Wallet: 0x742d35Cc...5f0bEb
       Monitoring 2 chain(s): Ethereum, Polygon
       💬 Ask me: "What's my current risk?"

User: What's my current risk?

Agent: I'm analyzing your portfolio with MeTTa AI...
       
       Current Risk: 🟡 MEDIUM (45%)
       
       Your portfolio shows moderate concentration risk...

User: status

Agent: 📊 Portfolio Status
       🟡 Risk Level: MEDIUM
       Risk Score: 45%
       Updated: Oct 16, 2025 10:35 AM

Commands: register, status, history, portfolio, chains, help

Programmatic Usage

from uagents import Agent, Context, Model
from datetime import datetime, timezone

class Portfolio(Model):
    user_id: str
    wallets: list[str]
    chains: list[str]
    timestamp: str

client = Agent(name="my_client", mailbox=True)

@client.on_event("startup")
async def register_portfolio(ctx: Context):
    portfolio = Portfolio(
        user_id="my_portfolio",
        wallets=["0x742d35Cc6634C0532925a3b844Bc9e7595f0bEb"],
        chains=["ethereum", "polygon"],
        timestamp=datetime.now(timezone.utc).isoformat()
    )
    await ctx.send(
        "agent1qv3pywlds6n86hr55p7lpvncwtd22d25yfe82zjg5tgx325cg9dnqylzy6f",
        portfolio
    )

if __name__ == "__main__":
    client.run()

---

📊 Risk Scoring Methodology

Formula

Risk Score = (Concentration × 0.35) + (Volatility × 0.45) + (Chain Diversity × 0.20)

Risk Levels

Level	Score	Emoji	Action
Low	0-30%	🟢	Monitor
Medium	30-50%	🟡	Review weekly
High	50-70%	🟠	Rebalance within 24h
Critical	70-100%	🔴	Review immediately

MeTTa Knowledge Base

All thresholds defined declaratively:

; Asset classifications
(has-risk bitcoin low)
(has-risk ethereum low)
(has-risk-pattern leverage critical)

; Thresholds
(concentration-threshold critical 0.70)
(volatility-threshold extreme 50)

Benefits: Explainable, extensible, maintainable by domain experts

---

🛠️ Technologies Used

Component	Technology
Agent Framework	Fetch.ai uAgents v0.22.10
AI Reasoning	SingularityNET MeTTa
Chat Interface	ASI:One Chat Protocol, ASI-1
Deployment	Agentverse Cloud
Blockchain	Web3.py v7.13
APIs	CoinGecko, GoPlus, Honeypot.is
Language	Python 3.12

---

📚 Documentation

• README.md - Project overview
• DEPLOYMENT.md - Agentverse deployment
• docs/METTA_INTEGRATION.md - MeTTa AI details

Agent Docs:

• Alert Agent
• Portfolio Monitor
• Risk Analysis
• Market Data
• Fraud Detection

Knowledge Base:

• metta/risk_knowledge.metta

---

🧪 Testing

# Terminal 1: Start agents
python local.py

# Terminal 2: Run tests
python tests/test_system.py

# Verify MeTTa
python verify_metta.py

---

🔐 Security

• ✅ No private keys (read-only monitoring)
• ✅ ERC-55 checksum validation
• ✅ Zero/burn address protection
• ✅ Secure agent messaging (Fetch.ai)
• ✅ Open source & auditable

---

🤝 Contributing

1. Fork repo
2. Create feature branch (git checkout -b feature/AmazingFeature)
3. Commit changes (git commit -m 'Add AmazingFeature')
4. Push (git push origin feature/AmazingFeature)
5. Open Pull Request

Areas: Blockchain integrations, MeTTa rules, risk metrics, UI/UX, docs, bug fixes

---

🙏 Acknowledgments

• ASI Alliance - Hackathon & infrastructure
• Fetch.ai - uAgents & Agentverse
• SingularityNET - MeTTa AI technology
• CoinGecko - Market data API
• GoPlus Security - Fraud detection API

---

📞 Contact

• GitHub: https://github.com/DhanteyUD/DeFiGuard
• Demo Video: YouTube Link
• ASI:One: Search DeFiGuard Alert Agent
• Issues: GitHub Issues

---

📄 License

MIT License - Open Source

---

Built with ❤️ by DhanteyUD

Securing DeFi, one portfolio at a time 🛡️

⭐ Star this repo if you find it useful!

---

Powered by ASI Alliance | Fetch.ai uAgents | SingularityNET MeTTa | ASI:One