A production-ready autonomous trading system powered by multiple AI models executing perpetual futures trades on Aster DEX. Six independent AI agents (OpenAI GPT-5, Claude, DeepSeek, Gemini, Grok, Qwen) analyze markets using 60+ technical indicators and make autonomous trading decisions 24/7. Powered by BNB Smart Chain & Aster DEX.

Status: Production Ready | Version: 1.0 | Last Updated: January 2025

Overview
Key Features
System Architecture
AI Agents
Technical Indicators
Performance Metrics
Trading Strategy
Technology Stack
Project Structure
Installation & Setup
Configuration
Usage
Monitoring
Documentation
Safety & Risk Management
Production Deployment
Contributing
License

Overview

This system represents a sophisticated approach to algorithmic trading, leveraging the analytical capabilities of six different AI models to make independent trading decisions. Each agent operates with its own wallet, risk parameters, and trading style, creating a diversified portfolio of strategies.

What It Does

Analyzes Markets: Processes real-time data from 20+ cryptocurrency pairs across 5 timeframes
Makes Decisions: Each AI agent independently decides when to open/close positions
Executes Trades: Automatically places orders on Aster DEX with leverage (1-50x)
Manages Risk: Sets stop-loss and take-profit orders for every position
Tracks Performance: Monitors win rate, PnL, Sharpe ratio, and other metrics
Syncs Continuously: Real-time position synchronization with exchange (every 5 seconds)

Core Statistics

| Metric | Value | |--------|-------| | AI Models | 6 (OpenAI, Claude, DeepSeek, Gemini, Grok, Qwen) | | Trading Pairs | 20+ cryptocurrencies | | Technical Indicators | 60+ across 5 timeframes | | Analysis Frequency | Every 30 seconds per agent | | Position Sync | Every 5 seconds | | Leverage Range | 1x - 50x (configurable) | | Database | Supabase PostgreSQL | | Exchange | Aster DEX Perpetual Futures |

Key Features

Multi-Agent Intelligence

Six independent AI agents analyze markets simultaneously, each with:

Unique Perspective: Different AI architectures provide diverse market views
Independent Decisions: No coordination between agents (reduces correlation risk)
Separate Wallets: Each agent manages its own capital
Custom Parameters: Configurable risk tolerance, leverage, and position sizing

Advanced Technical Analysis

The system calculates 60+ indicators across multiple timeframes:

Trend Indicators: EMA (20, 50, 200), MACD, SuperTrend, ADX Momentum: RSI (7, 14), Stochastic RSI Volatility: Bollinger Bands, ATR, True Range Volume: OBV, VWAP, Volume SMA Derivatives: Funding Rate, Mark-Index Spread, Open Interest

Robust Position Management

Fill Verification: Orders verified as filled before database update (prevents sync issues)
Position Tracking: Real-time PnL updates every 5 seconds
TP/SL Management: Automatic take-profit and stop-loss orders
Orphan Detection: Identifies and reconciles positions missing from database
Race Condition Prevention: Prevents duplicate closure attempts

Production-Grade Architecture

Error Handling: Comprehensive try-catch blocks and graceful degradation
Logging: Detailed trade logs for audit and debugging
API Optimization: Batch requests, caching, and efficient polling
Data Integrity: Foreign keys, constraints, and validation checks
No Data Loss: All decisions and positions permanently logged

System Architecture

graph TB
    subgraph "Market Data Pipeline"
        Timer1[Timer: 30s] --> AsterAPI[Aster DEX APIs]
        AsterAPI --> Klines[Klines Data]
        AsterAPI --> Premium[Premium Index]
        AsterAPI --> OI[Open Interest]

        Klines --> MarketEngine[aster.js<br/>Market Engine]
        Premium --> MarketEngine
        OI --> MarketEngine

        MarketEngine --> Indicators[60+ Indicators<br/>5 Timeframes]
    end

    subgraph "AI Analysis Layer"
        Indicators --> OpenAI[OpenAI GPT-5]
        Indicators --> Claude[Claude Sonnet]
        Indicators --> DeepSeek[DeepSeek]
        Indicators --> Gemini[Gemini Pro]
        Indicators --> Grok[Grok 2]
        Indicators --> Qwen[Qwen]

        OpenAI --> Decisions[Trading Decisions]
        Claude --> Decisions
        DeepSeek --> Decisions
        Gemini --> Decisions
        Grok --> Decisions
        Qwen --> Decisions
    end

    subgraph "Execution Layer"
        Decisions --> AgentMgr[Agent Manager<br/>agents.js]
        AgentMgr --> PosMgr[Position Manager<br/>position-manager.js]
        PosMgr --> Trader[AsterDex Trader<br/>makeTrade.js]
        Trader --> Exchange[Aster DEX<br/>Exchange]
    end

    subgraph "Monitoring & Sync"
        Timer2[Timer: 5s] --> Poller[Account Poller<br/>account-poller.js]
        Poller --> Exchange
        Exchange --> Poller
        Poller --> DB[(Supabase<br/>PostgreSQL)]
        PosMgr --> DB
    end

    DB --> AgentMgr

Component Responsibilities

| Component | File | Responsibility | |-----------|------|----------------| | Market Engine | aster.js | Fetches and processes market data, calculates indicators | | Agent Manager | agents.js | Orchestrates AI agents, manages decisions | | Position Manager | position-manager.js | Handles position lifecycle (open, monitor, close) | | Wallet Manager | agents.js | Manages per-agent wallets and capital allocation | | Aster DEX Trader | makeTrade.js | Direct API wrapper for Aster DEX | | Account Poller | account-poller.js | Real-time sync with exchange (every 5s) | | Price Updater | price-updater.js | Fast price cache updates |

AI Agents

Each AI agent receives identical market data but produces unique trading decisions based on its architecture and training.

Agent Configuration

{
  "openai": {
    "model": "gpt-5",
    "initialBalance": 10000,
    "maxLeverage": 20,
    "riskTolerance": "medium",
    "preferredStyle": "swing"
  },
  "claude": {
    "model": "claude-4.5-sonnet",
    "initialBalance": 10000,
    "maxLeverage": 15,
    "riskTolerance": "low",
    "preferredStyle": "mixed"
  }
  // ... 4 more agents
}

AI Prompt Structure

Each agent receives:

Session Context: Runtime, invocation count, current timestamp
Account State: Balance, available capital, unrealized PnL, active positions
Market Overview: BTC trend, market sentiment, volatility
Per-Asset Data:
- Trend indicators (EMA, MACD) across all timeframes
- Momentum (RSI, Stochastic RSI)
- Volatility (Bollinger Bands, ATR)
- Volume (OBV, VWAP)
- Derivatives data (funding rate, open interest)
- Time series arrays (price history, indicator history)
Active Positions: Current positions with PnL
Recent Trades: Last 5 closed positions with outcomes

Decision Output Format

AI agents return structured JSON:

{
  "symbol": "BTCUSDT",
  "action": "long",
  "confidence": 0.85,
  "entry_price": 42150.00,
  "take_profit": 44000.00,
  "stop_loss": 41000.00,
  "leverage": 10,
  "sizeUSD": 500,
  "riskUSD": 50,
  "reasoning": "Strong bullish divergence on 1H RSI...",
  "invalidation_condition": "4h_close_below_41500",
  "exit_strategy": "Take profit at resistance, trail stop if breaks above"
}

Technical Indicators

Indicator Categories

1. Trend Analysis

EMA 20, 50, 200 (Exponential Moving Averages)
MACD (Moving Average Convergence Divergence)
MACD State (bullish/bearish/neutral)
SuperTrend (trend following indicator)
ADX (Average Directional Index)

2. Momentum

RSI 7 (fast)
RSI 14 (standard)
Stochastic RSI K/D
Rate of Change

3. Volatility

Bollinger Bands (upper, lower, middle)
ATR (Average True Range)
True Range
Bollinger Band Position

4. Volume Analysis

OBV (On-Balance Volume)
OBV State (accumulation/distribution)
Volume SMA
Relative Volume
VWAP (Volume Weighted Average Price)

5. Support/Resistance

Key Support Levels
Key Resistance Levels
Fibonacci Retracements

6. Derivatives (Perpetual Futures Specific)

Funding Rate (long/short bias)
Annualized Funding Rate
Mark Price vs Index Price
Open Interest
Next Funding Time

Multi-Timeframe Analysis

All indicators calculated across 5 timeframes:

1 minute: Ultra-short-term movements
5 minutes: Short-term trends
15 minutes: Intraday analysis
1 hour: Medium-term trends
4 hours: Long-term context

Time Series Data

Recent history arrays for pattern recognition:

Last 10 mid-prices
Last 10 EMA values
Last 10 MACD values
Last 10 RSI values

This enables AI agents to identify patterns like:

Divergences (price vs indicator)
Trend reversals
Momentum shifts

Performance Metrics

Agent Performance Tracking

Each agent tracks comprehensive metrics:

| Metric | Description | |--------|-------------| | Balance | Current account equity (wallet + unrealized PnL) | | Total PnL | Cumulative profit/loss across all trades | | ROI | Return on investment percentage | | Win Rate | Percentage of profitable trades | | Trade Count | Total trades executed | | Win/Loss Count

MoneySharks

Install / Use

README

Table of Contents