English | 繁體中文

FinLab AI

Your AI's shortcut to mass-produce alpha-generating quant strategies.

Quick Install

For Cursor Users (One-Click!)

For Cursor / Antigravity IDE Users (Manual)

Add this to your MCP config (~/.cursor/mcp.json):

{
  "mcpServers": {
    "finlab": {
      "url": "https://finlab-ai.koreal6803.workers.dev/mcp"
    }
  }
}

No installation needed - the MCP server is hosted remotely!

For Claude Code / Other AI CLI Users

Simply tell your AI assistant:

"Please look at https://github.com/koreal6803/finlab-ai and install the FinLab skill for me"

This works with Claude Code, ChatGPT Codex CLI, Gemini CLI, and other AI coding assistants.

Features

• Comprehensive Data Access - 900+ data columns: prices, financials, revenue, valuations, institutional trading
• Strategy Development - Factor-based strategies using FinLabDataFrame methods
• Backtesting Engine - Risk management, stop-loss, take-profit, position sizing
• Factor Analysis - IC calculation, Shapley values, centrality analysis
• Machine Learning - Feature engineering and label generation for ML models

Examples

Fetch Institutional Trading Data

Prompt:

"List following TW stocks 近 5 天外資買賣超: 2330, 2317, 2454, 2881, 2308, 2382, 2882, 2412, 2303, 2344"

Result:

Build & Backtest a Strategy

Prompt:

"Build a monthly rebalancing strategy for Taiwan stocks: select stocks with positive revenue YoY growth, P/E ratio below 10, P/B ratio below 1.5 (deep value), and price above 60-day moving average (uptrend). Pick top 20 stocks monthly and backtest."

Result:

Manual Installation

Option 1: Claude Code

# Add marketplace
/install marketplace add koreal6803/finlab-ai

# Install
/install finlab@finlab

Option 2: ChatGPT Codex CLI

git clone https://github.com/koreal6803/finlab-ai.git
cd finlab-ai

Then tell Codex:

請幫我安裝此 finlab 裡的 skills

Option 3: Gemini CLI

# Install skill-porter
npm install -g skill-porter

# Clone and convert
git clone https://github.com/koreal6803/finlab-ai.git
cd finlab-ai
skill-porter convert ./finlab/skills/finlab --to gemini --output ./finlab-gemini-extension

Then install the generated finlab-gemini-extension following Gemini CLI docs.

Option 4: Cursor IDE (MCP Server)

Add to ~/.cursor/mcp.json:

{
  "mcpServers": {
    "finlab": {
      "url": "https://finlab-ai.koreal6803.workers.dev/mcp"
    }
  }
}

Restart Cursor and start using FinLab documentation in your Agent.

Option 5: Antigravity IDE (MCP Server)

1. Open Agent session → "..." → MCP Servers → Manage → View raw config
2. Add to mcp_config.json:

{
  "mcpServers": {
    "finlab": {
      "url": "https://finlab-ai.koreal6803.workers.dev/mcp"
    }
  }
}

Option 6: Local MCP Server (Any MCP Client)

# Clone and install
git clone https://github.com/koreal6803/finlab-ai.git
cd finlab-ai
pip install -e .

# Run the MCP server
python -m mcp_server

Configure your MCP client to use stdio transport with command python -m mcp_server.

Prerequisites

Get your FinLab API token: https://ai.finlab.tw/api_token/

export FINLAB_API_TOKEN="your_token_here"

Documentation

Comprehensive reference docs included:

Document	Content
Data Reference	900+ columns across 80+ tables
Backtesting Reference	sim() API, resampling, metrics
Factor Examples	60+ complete strategy examples
Best Practices	Patterns, anti-patterns, tips
ML Reference	Feature engineering, labels

License

MIT

Author

FinLab Community