AI 에이전트에게 실제 A-Share Quant Power를 제공하는 MCP 서버를 구축했습니다. 방법은 다음과 같습니다.

I Built an MCP Server That Gives AI Agents Real A-Share Quant Power — Here’s How

Give Claude Code and Cursor access to professional-grade China A-stock analysis. Factor scoring, industry chains, and live trading signals — all through MCP.

tl;dr: npx opc-a-stock-mcp → 3 MCP tools → instant quant analysis for any A-stock. Pro edition with 30-factor model coming soon.

The Problem: AI Agents Are Blind to A-Shares

Ask Claude Code “should I buy this stock?” and it’ll give you a disclaimer. Cursor has no idea what a stock score is. These tools are incredible for coding, but they’re missing the entire Chinese equity market.

As a quant developer running a live A-share trading system, I wanted my AI agents to have the same analytical power I use daily. So I built an MCP server.

What It Does

// Ask: "Score 中际旭创"
{
  "symbol": "300308",
  "score": 72,
  "signal": "STRONG_BUY",
  "factor_breakdown": {
    "trend": 18, "momentum": 15, "volume": 12,
    "rsi": 8, "fundamentals": 14, "macro": 5
  }
}

Enter fullscreen mode Exit fullscreen mode

Three tools, zero fluff:

Tool What It Returns get_stock_score 0-100 composite score + signal + factor breakdown get_industry_chain Complete upstream→downstream mapping for 6 chains, 80+ stocks get_daily_signals Buy/sell list with confidence scores + market conditions

The Factor Engine

Community edition runs a 10-factor model. The Pro edition (launching soon) runs the full 30-factor V10 engine — the same one driving a live trading desk.

Try It

npx opc-a-stock-mcp

Enter fullscreen mode Exit fullscreen mode

Then add to your Claude Code/Cursor MCP config. Full docs: github.com/liu-coco/a-stock-mcp

Pro edition ($19/mo): 30 factors, live data, debate-engine signals. DM or check GitHub for details.
https://keithsaga.gumroad.com/l/ykjzsm