Agentic FinSearch

Accurate retrieval of real-time financial data is a core infrastructure for building FinAgents. Our Agentic FinSearch is considered as a cornerstone towards democratizing financial intelligence to the general public. Also, it serves as a reference architecture for safe AI adoption in the financial services industry.

• Air-gapped deployment — runs fully on-premise in production, so user data never leaves your environment.
• Numerical accuracy — purpose-built for finance, where general-purpose LLMs routinely get the numbers wrong.
• No hallucination or misinformation — a hard requirement for safe AI adoption in high-stakes domains like finance.
• Source transparency — every response cites verifiable sources, so you can check exactly where each fact came from.

No trading suggestions. For research and informational purposes only.

Screenshots

1. The search agent overlay: drag, resize, and minimize. Provides context-aware information on the user's current page.

   

2. Source verification — check exactly where each piece of information came from.

   

Installation

If you just want to test it out, it is recommended to use the link above to install the official Agentic FinSearch's Chrome Extension.

For local installation, see below.

Prerequisites

• Python 3.12.x (uv will download it automatically)
• Node.js 18+
• Google Chrome browser

🐳 Quick Start with Docker (Recommended)

Build and run everything with a single command.

# Clone the repository
git clone https://github.com/Open-Finance-Lab/Agentic-FinSearch.git
cd Agentic-FinSearch

# Copy backend environment template and add your API keys
cp Main/backend/.env.example Main/backend/.env
# Edit Main/backend/.env and add at least one API key (OPENAI_API_KEY, ANTHROPIC_API_KEY, or DEEPSEEK_API_KEY)

# Start everything with one command (first run builds the image)
docker compose up

The backend API will be available at http://localhost:8000

Manual Backend Setup with uv (Optional)

Use this path if you want to run Django without Docker.

# Clone the repository
git clone https://github.com/Open-Finance-Lab/Agentic-FinSearch.git
cd Agentic-FinSearch

# Install backend dependencies with uv (Python 3.12)
cd Main/backend
uv sync --python 3.12 --frozen

# Install Playwright's Chromium once per machine
uv run playwright install chromium

# Start Django (no database migrations required)
uv run python manage.py runserver

For the browser extension, run once per environment:

cd Main/frontend
bun install
bun run build:full

Post-Installation

1. Configure API Keys (Required)

   Edit Main/backend/.env (copied from Main/backend/.env.example) and add at least one:

   OPENAI_API_KEY=your-actual-openai-key
   ANTHROPIC_API_KEY=your-actual-anthropic-key
   DEEPSEEK_API_KEY=your-actual-deepseek-key
   

   Note: The server will refuse to start without at least one valid API key configured.

2. Load Browser Extension

   • Open Chrome and navigate to Extensions page → chrome://extensions
   • Enable Developer mode
   • Click "Load unpacked"
   • Select Main/frontend/dist folder

Troubleshooting

• "No API keys configured!": The server won't start without valid API keys in Main/backend/.env
• Running locally: Use uv sync --python 3.12 --frozen inside Main/backend to bootstrap dependencies.
• Port 8000 in use: Close other servers or continue anyway.
• Non-English systems: UTF-8 encoding is automatically handled.
• Playwright browser errors: If you see browser-related errors, install Chromium:

  cd Main/backend
  uv run playwright install chromium

  See Main/backend/PLAYWRIGHT_INTEGRATION.md for detailed troubleshooting.

Usage and Documentation

For detailed usage instructions and more information, see: https://fingpt-search-agent-docs.readthedocs.io/en/latest/introduction.html

Citing

This project was originally published as FinGPT Search Agents and has since been renamed to Agentic FinSearch.

@inproceedings{tian2024customized,
  title={Customized fingpt search agents using foundation models},
  author={Tian, Felix and Byadgi, Ajay and Kim, Daniel S and Zha, Daochen and White, Matt and Xiao, Kairong and Liu, Xiao-Yang},
  booktitle={Proceedings of the 5th ACM International Conference on AI in Finance},
  pages={469--477},
  year={2024}
}

Disclaimer: We are sharing codes for academic purposes under the MIT education license. Nothing herein is financial advice, and NOT a recommendation to trade real money. Please use common sense and always first consult a professional before trading or investing.