CoinGlass Server

MCP Server yang menghubungkan Claude AI dengan CoinGlass API untuk analisis crypto derivatives real-time — dioptimasi untuk Ricoz Scalping Framework.

Arsitektur

Claude AI (claude.ai) ←── Streamable HTTP ──→ CoinGlass MCP Server (VPS)
                                                      │
                                                      ↓
                                              CoinGlass API V4

18 Tools

Critical (Ricoz Scalping Core)

Tool	Endpoint	Fungsi
coinglass_spot_cvd	Spot Aggregated CVD	VETO SIGNAL — SpotCVD negatif = jangan long
coinglass_futures_cvd	Futures Aggregated CVD	Entry filter — konfirmasi arah
coinglass_funding_rate	FR semua exchange	Sentiment — extreme = contrarian
coinglass_open_interest	OI aggregated history	Positioning — OI + price direction
coinglass_liquidation_map	Liquidation clusters	Target — magnetic zones
coinglass_orderbook	Orderbook delta history	Pressure — bid/ask imbalance

Important (Supporting Analytics)

Tool	Fungsi
coinglass_price_ohlc	Price OHLC history
coinglass_liquidation_history	Liquidation volume over time
coinglass_long_short_ratio	L/S ratio (contrarian indicator)
coinglass_taker_buysell	Taker aggressor side
coinglass_fr_arbitrage	Extreme FR across all coins
coinglass_coins_markets	Market overview semua coin

Nice-to-Have

Tool	Fungsi
coinglass_whale_alert	Hyperliquid whale positions
coinglass_fear_greed	Fear & Greed index
coinglass_footprint	Footprint chart data (90d)
coinglass_spot_netflow	Exchange net flow
coinglass_orderbook_heatmap	Orderbook depth visual

Composite

Tool	Fungsi
coinglass_full_scan	ALL-IN-ONE — 9 metrics dalam 1 call

Quick Start

1. Clone & Install

git clone https://github.com/rcz87/McPCG.git
cd McPCG
python3 -m venv .venv
source .venv/bin/activate
pip install -e ".[dev]"

2. Setup API Key

cp .env.example .env
nano .env  # masukkan COINGLASS_API_KEY

3. Run Server

# HTTP mode (untuk claude.ai remote)
python -m coinglass_mcp.server

# Stdio mode (untuk Claude Desktop lokal)
MCP_TRANSPORT=stdio python -m coinglass_mcp.server

Server berjalan di http://0.0.0.0:8787/mcp

4. Connect ke Claude

claude.ai (remote): Settings → Integrations → Add MCP Server → URL: http://YOUR_VPS_IP:8787/mcp

Claude Desktop (lokal):

{
  "mcpServers": {
    "coinglass": {
      "command": "python",
      "args": ["-m", "coinglass_mcp.server"],
      "cwd": "/path/to/McPCG",
      "env": {
        "COINGLASS_API_KEY": "your-key",
        "COINGLASS_PLAN": "standard",
        "MCP_TRANSPORT": "stdio"
      }
    }
  }
}

Deploy ke VPS

One-Command Setup

# Di VPS:
git clone https://github.com/rcz87/McPCG.git ~/coinglass-mcp
cd ~/coinglass-mcp
bash deploy/setup-vps.sh

Management Commands

bash deploy/manage.sh start     # Start server
bash deploy/manage.sh stop      # Stop server
bash deploy/manage.sh restart   # Restart
bash deploy/manage.sh status    # PM2 status
bash deploy/manage.sh logs      # View logs
bash deploy/manage.sh update    # Pull + restart
bash deploy/manage.sh health    # Health check
bash deploy/manage.sh tools     # List tools

HTTPS (Optional tapi Recommended)

bash deploy/setup-nginx.sh yourdomain.com

Tech Stack

Component	Version
Python	>= 3.10
FastMCP	3.1.1
httpx	>= 0.25.0
tenacity	>= 8.2.0
CoinGlass API	V4

Project Structure

McPCG/
├── src/coinglass_mcp/
│   ├── __init__.py     # Package
│   ├── config.py       # Plan tiers, intervals, rate limits
│   ├── client.py       # Async HTTP client + caching + retry
│   └── server.py       # FastMCP server + 18 tools
├── tests/
│   ├── test_config.py  # Config tests
│   └── test_client.py  # Client tests
├── deploy/
│   ├── setup-vps.sh    # VPS setup script
│   ├── setup-nginx.sh  # Nginx + SSL setup
│   └── manage.sh       # Management commands
├── docs/
│   └── RICOZ-TRADING-CHEATSHEET.md
├── BLUEPRINT-MCP-SERVER.md
├── pyproject.toml
├── .env.example
└── .gitignore

Lisensi

MIT