Trading Analysis MCP

An MCP (Model Context Protocol) server that gives Claude the ability to analyze trading setups across IDX (Indonesian stocks), Forex, and Crypto markets — without executing any trades.

Claude acts as your analyst: fetching live market data, calculating technical indicators, detecting market structure, and scoring setups by confluence. You make the final call.

What Claude can do with this MCP

Each analysis includes: EMA stack, RSI, MACD, Bollinger Bands, ATR, Support/Resistance, BOS/CHoCH, Order Blocks, Fair Value Gaps, and a confluence score (0–8).

Related MCP server: KaiaFun MCP

Requirements

• Python 3.10+

• Claude Desktop (for local MCP)

• Internet connection (for live market data)

Quick Start

1. Clone the repo

git clone https://github.com/YOUR_USERNAME/trading-mcp.git
cd trading-mcp

2. Install dependencies

pip install -r requirements.txt

3. Configure Claude Desktop

Find your Claude Desktop config file:

Add this entry (replace the path with your actual clone location):

Windows:

{
  "mcpServers": {
    "trading-analysis": {
      "command": "python",
      "args": ["C:/path/to/trading-mcp/server.py"]
    }
  }
}

macOS / Linux:

{
  "mcpServers": {
    "trading-analysis": {
      "command": "python3",
      "args": ["/path/to/trading-mcp/server.py"]
    }
  }
}

4. Restart Claude Desktop

The tools are now available. Try asking Claude:

"Scan EURUSD, GBPUSD, USDJPY for the best setup today"

Available MCP Tools

analyze_symbol

Full technical + structure analysis for a single symbol.

Parameters:
  symbol     — e.g. BBCA.JK, EURUSD, BTC/USDT, XAUUSD
  timeframe  — 1m / 5m / 15m / 30m / 1h / 4h / 1d / 1w  (default: 1d)
  limit      — number of bars to load (default: 250)

Output includes: market structure, EMA stack, RSI, MACD, Bollinger Bands, ATR, volume ratio, key S/R levels, active Order Blocks, open Fair Value Gaps, and confluence score.

get_multi_timeframe_analysis

Top-down analysis across multiple timeframes to confirm bias before entry.

Parameters:
  symbol     — e.g. EURUSD
  timeframes — comma-separated, large to small (default: "1d,4h,1h")

Output includes per-timeframe summary + HTF vs LTF synthesis.

scan_setups

Scan a watchlist and rank symbols by confluence score.

Parameters:
  symbols    — comma-separated, e.g. "BBCA.JK,TLKM.JK,BMRI.JK"
  timeframe  — timeframe to scan (default: 1d)

Output: ranked table (score, structure, bias, RSI) + detailed breakdown of top 3.

calculate_rr

Risk/Reward calculator with trade direction validation.

Parameters:
  entry        — entry price
  stop_loss    — stop loss price
  take_profit  — take profit price
  risk_amount  — (optional) capital at risk in USD/IDR for profit projection

get_market_session

Current trading session info and which instruments to focus on.

Output: active sessions (Asian/London/NY/IDX), overlap status, recommended pairs.

Supported Symbols

Auto-detection: just type the symbol in any format — the server detects the market automatically.

Data Sources

No API keys required for basic usage. All data is read-only.

Analysis Concepts

This server implements SMC (Smart Money Concepts) / ICT principles algorithmically:

• Swing High/Low — pivot detection with configurable lookback

• BOS (Break of Structure) — trend continuation confirmation

• CHoCH (Change of Character) — trend reversal signal

• Order Blocks — institutional supply/demand zones

• Fair Value Gaps — price imbalances that tend to get filled

Combined with classic technical indicators (EMA stack, RSI, MACD, Bollinger Bands, ATR) to produce a confluence score (0–8).

Architecture

Claude Desktop
    │
    └─ MCP (stdio)
         │
         └─ Trading Analysis MCP Server
               ├── DataFetcher
               │     ├── yfinance  → IDX + Forex (+ crypto fallback)
               │     └── ccxt      → Crypto (Bybit, OKX, KuCoin, Gate.io)
               │
               ├── TechnicalEngine  (pandas-ta)
               │     └── EMA 8/21/50/200 · RSI · MACD · BB · ATR · Volume
               │
               └── StructureEngine  (SMC/ICT)
                     └── Swings · S/R · BOS/CHoCH · Order Blocks · FVG

Option B — Remote MCP (Hosted on Railway)

Deploy once, share a URL. Others connect without installing Python.

1. Deploy to Railway

1. Go to railway.app and create a new project

2. Click Deploy from GitHub repo → select trading-mcp

3. Railway auto-detects railway.toml and starts server_remote.py

2. Set environment variables

In your Railway project → Variables tab, add:

API_KEYS=your_secret_key_here

Generate a secure key:

python -c "import secrets; print(secrets.token_urlsafe(32))"

You can add multiple keys separated by commas: key1,key2,key3

Optional:

RATE_LIMIT_PER_MIN=20

3. Get your URL

After deploy, Railway gives you a URL like:

https://trading-mcp-production.up.railway.app

Health check: GET /health — should return {"status": "ok", ...}

4. Connect from Claude Desktop

{
  "mcpServers": {
    "trading-analysis": {
      "type": "http",
      "url": "https://your-app.up.railway.app/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Share the URL and API key with anyone — no Python installation needed on their end.

Running Tests

python -m tests.test_fetcher
python -m tests.test_technical
python -m tests.test_structure
python -m tests.test_mcp_tools

License

MIT — free to use, modify, and share.