ASI:BlockAgent
A real-time whale detection and trading intelligence system for Hyperliquid
Demo
About
A cutting-edge AI-powered system that detects and analyzes large cryptocurrency deposits ("whale activity") on Hyperliquid in real-time, providing traders with instant intelligence about market-moving events. ## 🎯 What It Does - Real-Time Whale Detection : Monitors Ethereum mainnet for large deposits (>$100k) to the Hyperliquid bridge using Blockscout MCP integration - Instant Position Analysis : Automatically fetches whale trading positions, leverage, and liquidation prices from Hyperliquid APIs - AI-Powered Chat Interface : Natural language queries like "Show me recent whales" or "What positions do large traders have?" - ASI:ONE Integration : Connects to Agentverse agents for advanced AI capabilities with fallback to local processing
How it's made
## 🔧 Technical Architecture & Implementation ### Core Technologies Stack - Frontend : Next.js 14 with TypeScript, Tailwind CSS, Lucide React icons - Backend : Python with uAgents framework for multi-agent orchestration - Blockchain Integration : Blockscout MCP (Model Context Protocol) for Ethereum monitoring - AI Integration : ASI:ONE protocol connecting to Agentverse agents - APIs : Direct integration with Hyperliquid's trading APIs ### Multi-Agent System Design 1. Local Mailbox Agent ( whale_agent_mailbox.py ) - Built with uAgents framework running on port 8001 - Implements two core tools: - get_recent_whales() : Queries Blockscout MCP for large bridge deposits - enrich_wallet() : Fetches position data from Hyperliquid APIs - Handles natural language processing for whale detection queries 2. HTTP Server Agent ( whale_agent_http_server.py ) - FastAPI server on port 8002 providing REST endpoints - Bridges between the frontend and the mailbox agent - Enables stateless operation without database persistence 3. ASI:ONE Integration ( asi-one.ts ) - Custom TypeScript client for ASI:ONE protocol - Implements agent discovery and message routing - Fallback mechanism: tries ASI:ONE first, then local agents ### Blockchain Monitoring Innovation Blockscout MCP Integration : ``` # Custom MCP client for real-time whale detection async def get_recent_whales(threshold_usd: float = 100000, lookback_minutes: int = 60): # Query Hyperliquid bridge contract: 0x2Df1c51E09aECF9cacB7bc98cB1742757f163dF7 # Decode deposit events using verified ABI # Filter transactions above threshold ``` Particularly Hacky & Notable : 1. Stateless Architecture : No database! Everything processes live data from APIs 2. Hybrid Agent Routing : Frontend intelligently routes between local and cloud agents based on preferLocal setting 3. Real-time Bridge Monitoring : Direct integration with Blockscout's MCP to catch whale deposits as they happen ### Frontend Architecture Chat Interface ( chat-interface.tsx ): - React hooks for real-time message streaming - Dual-mode operation: local vs ASI:ONE agents - Connection status monitoring with automatic fallbacks - Quick action buttons for common queries