Asking a language model to compute RSI from a list of closes is a good way to get a confident wrong number. Here the indicator, the trend classification and the support levels are calculated server-side and returned with their signals attached. The multi-symbol calls — comparison, correlation, volatility ranking — are the part the upstream project does not have.
A fork of snjyor's binance-mcp that keeps the original market-data surface — prices, order book, historical trades, K-line candles, 24hr statistics, exchange info, symbol tickers — and adds an analysis layer on top of it, so the model reads a computed indicator instead of trying to derive one from a candle array.
- Moving averages (SMA, EMA) with a period you choose, and momentum indicators RSI and MACD
- Volatility measures — Bollinger Bands and ATR — plus VWAP for volume-weighted levels
- Market analysis for a symbol and timeframe: bullish / bearish / neutral trend classification, support and resistance levels, and overbought / oversold detection
- Price statistics including returns, volatility and Sharpe ratio
- Multi-symbol work: performance comparison across up to 5 symbols, correlation analysis to find related pairs, and volatility ranking
- Responses carry the current values together with their signals, rather than numbers you have to interpret
The package is `@snjyor/binance-mcp-enhanced`, version 1.0.2-enhanced. The documented route is Docker — `docker build -t binance-mcp-enhanced .` then a client entry running `docker run -i --rm`; locally it is `npm install`, `npm run build`, `npm start`. Set `BINANCE_TESTNET=true` for paper trading against the testnet. The container runs as a non-root user.
One command — docker run -i --rm binance-mcp-enhanced