Crypto Market MCP

Python-based Model Context Protocol (MCP) server for real-time and historical cryptocurrency market data using ccxt.

Features

• MCP tools for:
  • get_ticker(symbol) – latest ticker for a trading pair
  • get_ohlcv(symbol, timeframe, limit) – historical OHLCV candles
  • get_order_book(symbol, limit) – order book snapshot
  • stream_ticker(symbol, interval_seconds) – streaming-style ticker updates via polling
• Pluggable exchange backend (default: Binance via ccxt)
• Redis cache support for distributed caching across multiple server instances
• Redis Pub/Sub for delegating requests to worker pools
• Simple TTL cache fallback (in-memory)
• Structured error handling
• Small CLI chatbot client to exercise the MCP server.

Installation

python -m venv .venv
.venv\Scripts\activate
pip install -e .[dev]

Or with uv:

uv sync

Running the MCP server

The server is installed as a console script named crypto-market-mcp.

Standard mode (stdio)

crypto-market-mcp

It speaks MCP over stdio, so it can be wired into any MCP-compatible host.

Worker mode (Redis Pub/Sub)

Start one or more workers to process requests via Redis:

crypto-market-worker

Configure via environment variables:

$env:REDIS_HOST="localhost"
$env:USE_REDIS_CACHE="true"
crypto-market-worker

See REDIS_INTEGRATION.md for detailed Redis setup and usage.

Chatbot client

A small chatbot is provided to exercise the MCP server locally.

crypto-market-chatbot

Example conversation:

you> price BTC/USDT
mcp> {"symbol": "BTC/USDT", "price": 50000.0, ...}

you> ohlcv BTC/USDT
mcp> [ {"timestamp": ..., "open": ...}, ... ]

Running tests

pytest

This runs unit tests for the cache, exchange client wrapper, and MCP tools.