# QuantToGo MCP Servers

MCP (Model Context Protocol) servers for **QuantToGo** - a quantitative trading platform providing live trading signals, market data, and portfolio management through AI-native interfaces.

## Overview

This repository contains three MCP servers that expose QuantToGo's quantitative trading capabilities:

| Server | Description | Tools |
|--------|-------------|-------|
| **quanttogo-signals** | Real-time trading signals from live quant strategies | 3 tools |
| **quanttogo-market-data** | Product catalog, NAV history, backtest reports | 5 tools |
| **quanttogo-portfolio** | Portfolio positions, dual-track performance, trade history | 6 tools |

Plus an **HTTP server** for remote access (MCP Connector mode).

## Quick Start

### Installation

```bash
npm install quanttogo-mcp-servers
```

Or clone and build:

```bash
git clone https://github.com/michaeljiangmingfeng-debug/quanttogo-mcp-servers.git
cd quanttogo-mcp-servers
npm install
npm run build
```

### Configuration

Set environment variables:

```bash
export QUANTTOGO_API_BASE="https://www.quanttogo.com"  # API endpoint
export QUANTTOGO_API_KEY="your-api-key"                 # API authentication key
export QUANTTOGO_USER_ID="your-user-id"                 # Your user ID
```

### Usage with Claude Desktop

Add to your `claude_desktop_config.json`:

```json
{
  "mcpServers": {
    "quanttogo-signals": {
      "command": "npx",
      "args": ["tsx", "src/signals-server.ts"],
      "cwd": "/path/to/quanttogo-mcp-servers",
      "env": {
        "QUANTTOGO_API_KEY": "your-api-key",
        "QUANTTOGO_USER_ID": "your-user-id"
      }
    },
    "quanttogo-market-data": {
      "command": "npx",
      "args": ["tsx", "src/market-data-server.ts"],
      "cwd": "/path/to/quanttogo-mcp-servers",
      "env": {
        "QUANTTOGO_API_KEY": "your-api-key"
      }
    },
    "quanttogo-portfolio": {
      "command": "npx",
      "args": ["tsx", "src/portfolio-server.ts"],
      "cwd": "/path/to/quanttogo-mcp-servers",
      "env": {
        "QUANTTOGO_API_KEY": "your-api-key",
        "QUANTTOGO_USER_ID": "your-user-id"
      }
    }
  }
}
```

### Remote HTTP Mode (Connector)

For remote access via Streamable HTTP:

```bash
npm run serve
# Server starts at http://localhost:3000/mcp
```

## Available Tools

### Signals Server (`quanttogo-signals`)

| Tool | Description |
|------|-------------|
| `get_trading_signals` | Get latest BUY/SELL signals with strategy, symbol, quantity, price |
| `confirm_signal` | Execute or skip a pending signal |
| `get_signal_stats` | Signal performance metrics and win rate |

### Market Data Server (`quanttogo-market-data`)

| Tool | Description |
|------|-------------|
| `get_products` | List all quantitative trading products |
| `get_product_detail` | Detailed product info with parameters |
| `get_nav_history` | Historical NAV data with daily returns |
| `get_backtest_report` | Strategy backtest with Sharpe ratio, drawdown |
| `search_products` | Search by risk level, currency, strategy type |

### Portfolio Server (`quanttogo-portfolio`)

| Tool | Description |
|------|-------------|
| `get_portfolio` | Complete portfolio overview |
| `get_dual_track_comparison` | AUTO vs MANUAL track performance comparison |
| `get_positions` | Current open positions across strategies |
| `get_trade_history` | Historical trade records |
| `get_subscriptions` | User subscription status |
| `get_performance_metrics` | Return, drawdown, Sharpe ratio, win rate |

## Architecture

```
quanttogo-mcp-servers/
├── src/
│   ├── common/
│   │   ├── client.ts          # QuantToGo API client
│   │   └── types.ts           # Shared TypeScript types
│   ├── signals-server.ts      # Signals MCP Server (stdio)
│   ├── market-data-server.ts  # Market Data MCP Server (stdio)
│   ├── portfolio-server.ts    # Portfolio MCP Server (stdio)
│   └── http-server.ts         # Combined HTTP server (Connector mode)
├── package.json
├── tsconfig.json
├── glama.json
├── LICENSE
└── README.md
```

## Key Concepts

### Dual-Track System

QuantToGo uses a unique dual-track system:
- **AUTO track**: All signals are automatically executed (virtual trading)
- **MANUAL track**: Only user-confirmed signals are executed

This allows users to compare their decision-making against the algorithm's full execution.

### Signal Sources

Signals come from two quantitative platforms:
- **QC** (QuantConnect): US market strategies
- **JQ** (JoinQuant): China market strategies

## Development

```bash
# Run individual servers in dev mode
npm run dev:signals
npm run dev:market-data
npm run dev:portfolio

# Run HTTP server
npm run serve

# Build for production
npm run build
```

## License

MIT