FactorHub MCP Server

China A-share market data for AI agents. Query factor scores, market quotes, valuations, and run strategy backtests — all through the Model Context Protocol (MCP).

Features

Quick Start

1. Get an API Key

Sign up at factorhub.cn and generate your API key at the API Keys page. Free tier includes 10 API calls/day.

2. Install

pip install factorhub-mcp

Or install from source:

git clone https://github.com/michaelfeng/factorhub-mcp.git
cd factorhub-mcp
pip install -e .

3. Configure

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "factorhub": {
      "command": "factorhub-mcp",
      "env": {
        "FACTORHUB_API_KEY": "fh_your_api_key_here"
      }
    }
  }
}

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "factorhub": {
      "command": "factorhub-mcp",
      "env": {
        "FACTORHUB_API_KEY": "fh_your_api_key_here"
      }
    }
  }
}

OpenClaw

Copy openclaw/factorhub-skill.yaml to your OpenClaw skills directory, or add the MCP config directly:

providers:
  factorhub:
    type: mcp
    command: factorhub-mcp
    env:
      FACTORHUB_API_KEY: "fh_your_api_key_here"

Smithery

Install via Smithery for automatic setup.

Usage Examples

Once configured, ask your AI assistant:

• "查看 FactorHub 有哪些因子"
• "获取 ROE 因子的历史表现"
• "帮我查一下贵州茅台最近一年的行情"
• "用低估值策略回测沪深300成分股"
• "对比动量因子和价值因子的夏普比率"

Pricing

Upgrade at factorhub.cn/pricing.

Architecture

This MCP server is a thin client that calls the FactorHub public API. It does not access any database directly or contain proprietary logic.

AI Agent  →  MCP Protocol  →  factorhub-mcp  →  HTTPS  →  factorhub.cn/api/v1

Security

• API key is stored in environment variables, never hardcoded
• All communication uses HTTPS
• No user data is stored locally
• Rate limiting is enforced server-side

License

MIT