Agentic Edge: A Multi-Agent Quantitative Investment Platform Simulating the Full Workflow of a Hedge Fund

Agentic Edge is a quantitative research platform based on a multi-agent architecture, simulating the full workflow of a hedge fund. Its core converts investment themes into ranked candidate targets through Bull/Bear adversarial reasoning, options flow analysis, and fundamental evaluation, with the goal of achieving 3x index excess returns. The platform aims to free humans from repetitive work, allowing them to focus on target judgment and strategy decisions. It also emphasizes interpretability—no black-box scoring, and the reasoning process is transparent and visible.

Traditional discretionary research teams spend a lot of time on repetitive work every week: building stock pools, integrating data providers, pre-screening theme stocks, summarizing capital flows, etc. Agentic Edge migrates this cycle to a collaborative agent network, allowing humans to focus on valuable judgments (such as evaluation of system-screened targets, position/holding/exit strategies). The core philosophy is interpretability: clicking any agent allows you to view its conclusions and reasoning process for a stock code, and the reasoning process itself is the deliverable.

Agentic Edge runs the full set of hedge fund workflows: analyst coverage, bull-bear debates, risk assessment, position management, rebalancing, execution—scoring investment themes through institutional options flow, fundamental data, and dialectical AI to output candidate targets with the highest winning potential. The frontend is built on Next.js15, including core modules: Digital Trading Hall homepage (real-time agent network graph animation), Theme Management (add investment themes like AI infrastructure), Run History (record agent timeline and results), Scorecard (rankings integrating bull-bear debates, etc.), Performance Tracking (simulated account profit/loss and positions), Emergency Brake (one-click stop of automated cycles).

1. Theme Priority: Users input investment arguments, and the system finds matching targets to ensure research is logic-driven; 2. Adversarial Reasoning: Bull and bear researchers debate, and the research manager synthesizes and eliminates low-confidence targets; 3. Options-Aware Analysis: Treats abnormal flows, gamma exposure, and maximum pain as core signals; 4. Integrated Dual Strategies: Pure long strategy (clean setups), covered LEAPS strategy (maximizing capital efficiency); 5. Simulated Account Priority: Real-world broker connections require manual code changes, reflecting a prudent approach to risk control.

Three-Stage Scoring: 1. Fundamental Quality (6 dimensions: profitability, revenue growth, capital efficiency, free cash flow, balance sheet health, future outlook—with higher weight on the latter); 2. Hot Cycle Weighting (adjust scores based on current theme bottlenecks, e.g., Micron may outperform NVIDIA in the storage theme); 3. Entry Timing (entry points determined by signals like options flow, GEX positioning, trends). Agent Network: Market Analyst (price trends), Fundamental Analyst (financial report data), Options Flow Analyst (abnormal flows), Bull/Bear Researchers (build pro/con cases), Research Manager (synthesize views), Trader (generate final recommendations).

Risk Control: Stated as research decision support software, not investment advice; real-world execution is disabled by default; five-stage gating and emergency brake ensure human control. Tech Stack: Frontend Next.js15, custom multi-agent framework, institutional-level data sources, IB API (simulated accounts only). Summary: Agentic Edge represents a mature approach to AI financial applications—freeing humans from repetitive work to focus on decision-making; providing open-source references through adversarial reasoning, multi-dimensional integration, and interpretability mechanisms; theme priority and simulation priority reflect a deep understanding of investment essence and risk.