Crypto Indicators MCP Server for AI Trading Agents

Overview

Crypto Indicators MCP Server provides an MCP-compatible process that exposes 50+ cryptocurrency technical indicators and a set of corresponding trading strategies suitable for integration with AI trading agents and quantitative workflows. The server computes trend, momentum, volatility, and volume indicators on historical OHLCV data (defaulting to Binance via ccxt) and returns numerical results and strategy signals that downstream agents can use for decision-making, backtesting, or feature generation.

Because it runs as an external MCP server, your language model or agent can call indicators as tools (via the MCP client integration) rather than embedding indicator logic in prompts or code. This separation makes experiments reproducible, reduces model prompt complexity, and centralizes market-data handling and indicator computation.

Features

• 50+ technical indicators spanning trend, momentum, volatility, and volume.
• Trading strategies that output discrete signals: -1 (SELL), 0 (HOLD), 1 (BUY).
• Default data source: Binance (via ccxt). Exchange can be changed to any ccxt-supported provider.
• Modular structure: indicators and strategies are organized for maintainability and extension.
• Node.js implementation (tested on Node 18+), MIT licensed.
• Designed for easy MCP client registration (examples provided).

Installation / Configuration

Prerequisites:

• Node.js v18.x or later
• npm v8.x or later

Clone and install:

git clone https://github.com/kukapay/crypto-indicators-mcp.git
cd crypto-indicators-mcp
npm install

Run the server directly:

# defaults to EXCHANGE_NAME=binance if not set
EXCHANGE_NAME=binance node index.js

Example MCP client configuration (add to your MCP client config, e.g., Claude Desktop):

{
  "mcpServers": {
    "crypto-indicators-mcp": {
      "command": "node",
      "args": ["path/to/crypto-indicators-mcp/index.js"],
      "env": {
        "EXCHANGE_NAME": "binance"
      }
    }
  }
}

Environment variables:

• EXCHANGE_NAME: ccxt exchange id to use for fetching OHLCV (default: binance)
• (Optional) Provide API keys via ccxt-standard environment variables if private endpoints are needed.

Available Tools

The server exposes functions (tools) for many indicators. Each tool accepts parameters such as symbol (e.g., “BTC/USDT”), timeframe (e.g., “1h”), period/window sizes, and a limit for historical bars. Most indicator tools return arrays of computed values aligned to input candles; strategy tools typically return a current integer signal (-1/0/1).

Summary of categories and representative tools:

Full list of available tool names is included in the repository; tools are grouped in directories for trend, momentum, volatility, and volume so you can quickly find and extend implementations.

Usage Patterns & Example Calls

Typical MCP tool invocation (JSON payload passed by MCP client when calling a tool):

{
  "tool": "calculate_relative_strength_index",
  "args": {
    "symbol": "BTC/USDT",
    "timeframe": "1h",
    "period": 14,
    "limit": 200
  }
}

Typical response shape:

{
  "indicator": "rsi",
  "period": 14,
  "values": [45.2, 47.8, 50.1, ...],
  "last_value": 50.1
}

Strategy example (returns signal):

{
  "tool": "ema_crossover_strategy",
  "args": { "symbol": "ETH/USDT", "timeframe": "4h", "fast": 12, "slow": 26, "limit": 200 }
}

Response:

{ "strategy": "ema_crossover", "signal": 1 }  // 1 = BUY, 0 = HOLD, -1 = SELL

Note: exact tool names and argument names are available in the repository; confirm those when integrating.

Use Cases

• Feature generation for agent-based trading: compute indicators on live or historical data, pass numeric arrays or last values to your ML model or LLM as structured context.
• Strategy prototyping: call strategy tools to get immediate buy/hold/sell signals and combine them with your risk-management logic.
• Backtesting pipelines: run indicators across historical OHLCV to create signal traces and evaluate performance metrics offline.
• Research and signal aggregation: aggregate multiple indicator outputs to build composite signals without reimplementing indicator math in multiple agents.

Concrete example

• An AI agent evaluates market entries on BTC/USDT 1-hour candles. It requests calculate_relative_strength_index and calculate_bollinger_bands from the MCP server, combines the last RSI value and Bollinger band position, and then calls an EMA crossover strategy tool to finalize the signal. The agent receives a single integer signal to act on.

Resources

• Repository: https://github.com/kukapay/crypto-indicators-mcp
• Related MCP servers: https://github.com/kukapay/kukapay-mcp-servers

License: MIT

For developers new to MCP, register the process in your client configuration and treat each indicator as a callable tool that returns deterministic numeric outputs that your agent can reason about or consume as features.