guangxiangdebizi/FinanceMCP

What it solves

FinanceMCP provides AI agents with a unified, routable, and traceable way to access financial market data across multiple markets. It eliminates the need for AI agents to integrate multiple disparate financial APIs individually by providing a standardized set of tools for stocks, funds, bonds, and macroeconomic data.

How it works

Built on the Model Context Protocol (MCP), the project implements a server that exposes 19 stable tools. It acts as a routing layer between AI clients (like Claude, Cursor, or Trae) and various data providers such as Tushare, Qveris, Binance, and Twingly. The system supports two transport modes: local stdio and Streamable HTTP. It features an intelligent routing system that automatically falls back to alternative data sources if the primary source is unavailable or doesn't cover the requested data.

Who it’s for

• AI Agent Developers: Those building financial analysis agents or investment research tools.
• MCP Client Users: Users of Claude Desktop, Cursor, or Codex who want to add real-time financial data capabilities to their LLM.
• Financial Researchers: Users of systems like FinNote who need structured financial data integrated into an AI-driven document workflow.

Highlights

• Multi-Market Coverage: Supports A-shares, Hong Kong stocks, US stocks, indices, funds, bonds, futures, forex, and crypto assets.
• Intelligent Routing & Fallback: Automatically switches between providers (Tushare, Qveris, Binance, etc.) based on priority and availability.
• Technical Indicator Engine: Automatically calculates MACD, RSI, KDJ, BOLL, and MA by pre-fetching necessary historical data.
• Source Transparency: Every response explicitly labels the actual data source used and the routing path taken.
• Dual Transport: Supports both local stdio for desktop apps and remote HTTP for cloud deployments.