Summary

Open-Finance-Lab/AgenticTrading offers an experimental, open-source platform for developing and evaluating LLM-powered trading agents. It targets students, researchers, and developers, enabling prototyping, traceable backtests, paper-trading simulations, decision inspection, and performance benchmarking under realistic financial constraints.

How It Works

The platform uses a backtest -> API -> dashboard pipeline. A CLI backtest engine fetches market data, runs agent/baseline logic, and stores results in SQLite. A FastAPI REST API exposes endpoints for backtest results, paper trading (via Alpaca API), and dashboard data. The frontend visualizes agent performance. A key component is the integrated FinAgent Orchestration Framework, providing multi-agent capabilities, memory, and DAG planning.

Quick Start & Requirements

• Install/Run: Python dependencies via pip. Backtesting via scripts (e.g., scripts/backtest_hourly_agent.py). FastAPI backend runs locally.
• Prerequisites: Alpaca API credentials for paper trading/data access. SQLite for storage.
• Documentation: Local preview in docs/README.md. Full docs at finagent-orchestration.readthedocs.io.
• Setup: Requires Alpaca API key configuration. Resource footprint/time unspecified.

Highlighted Details

• Agent Customization: Build LLM agents with configurable models, prompts, assets, and logic.
• Interactive Simulation: Backtest on historical data; deploy to live-market paper trading accounts.
• Decision Transparency: Inspect detailed logs and reasoning traces for BUY/SELL/HOLD decisions.
• Performance Evaluation: Analyze key metrics (return, Sharpe ratio, drawdown) with benchmarks.
• FinAgent Framework: Utilizes a multi-agent framework for advanced orchestration, memory, and planning.

Maintenance & Community

The project incorporates the FinAgent Orchestration Framework by Jifeng Li et al. No specific community channels or active maintenance signals beyond future roadmap items are detailed.

Licensing & Compatibility

Licensed under OpenMDW-1.0 (Copyright Jifeng Li @ SecureFinAI Lab). Specific terms for commercial use or closed-source linking are not detailed.

Limitations & Caveats

Described as an "experimental playground" and "research platform," it may not be production-ready. The leaderboard uses mock data, pending integration of real multi-agent runs. LLM integration requires specific configuration, with only an example provided.