FinanceMCP

Why we included this project

FinanceMCP exposes 19 Model Context Protocol tools that pull market data from Tushare, Binance, Qveris, and public news sources, so a language model can answer finance questions with real numbers instead of guessing. Teams building assistants in Claude, Cursor, or Codex can connect it over stdio or Streamable HTTP and query A-shares, Hong Kong and US equities, funds, bonds, futures, FX, macro indicators, and crypto without writing their own data integrations. The routing layer is the part worth noticing: it picks the best source per request, falls back when one is rate-limited or lacks coverage, and marks the actual origin on every response, which keeps an agent's answer traceable. There is also a technical-indicator engine (MACD, RSI, KDJ, BOLL, MA) and news tools, and public crypto and news data work with no API keys at all. For developers who want their model to give data-backed answers rather than plausible-sounding ones, this is a solid self-hosted option.