← Quay lại danh sách bài viết
May 12, 2026
5 phút đọc

TradingAgents: Multi-Agent AI Framework That Models a Hedge Fund

TradingAgents: Multi-Agent AI Framework That Models a Hedge Fund
#AI
#multi-agent systems
#LangGraph
#LLM
#trading
#automation
#risk management

TradingAgents multi-agent system

Most AI trading projects are a single LLM fed data and asked to "make a decision." TradingAgents (GitHub, arXiv: 2412.20138) takes a different path: instead of one agent — an entire trading firm staff, where each role is a separate LLM agent with its own tools, data sources, and prompt.

Built on LangGraph, the framework creates a system that doesn't just "look at data" — it debates with itself before making a decision.

Architecture: From Data to Decision

Decision pipeline

The full pipeline is a directed acyclic graph (DAG) of 12 nodes:

Analyst Team → Research Debate → Trader → Risk Debate → Portfolio Manager → BUY/HOLD/SELL

Analyst Team: Four Specializations

Agent Data Sources Focus
Fundamentals Analyst Balance sheet, cashflow, income statement Intrinsic value vs market price
Sentiment Analyst Yahoo Finance News, StockTwits, Reddit Multi-source sentiment aggregation
News Analyst Ticker news, macro headlines, insider transactions Event-driven signals
Technical Analyst OHLCV, MACD, RSI, Bollinger Bands Pattern detection and momentum

Research Debate: Bull vs Bear

Agent debates

After analysts produce reports, adversarial debate begins:

  1. Bull Researcher builds the bullish thesis
  2. Bear Researcher builds the bearish counter-thesis
  3. Multi-round debate (configurable via max_debate_rounds)
  4. Research Manager (deep thinking LLM) synthesizes both positions

Risk Management: Triple Filter

The trader's proposal passes through three risk managers debating each other:

Agent Profile
Aggressive Analyst High risk tolerance, upside focus
Neutral Analyst Balanced risk/reward
Conservative Analyst Low risk tolerance, downside protection

Portfolio Manager: Final Decision

Receives trader proposal + risk debate results + reflective memory from past decisions. Approves, rejects, or adjusts the trade.

Reflective Memory

The system maintains a decision log. On each subsequent run for the same ticker, it fetches realized returns (raw + alpha vs SPY), generates reflections, and injects history into the Portfolio Manager prompt — creating learning from mistakes without fine-tuning.

Tech Stack

Component Technology
Orchestration LangGraph (StateGraph, checkpoints)
LLM Providers OpenAI, Google, Anthropic, xAI, DeepSeek, Qwen, Ollama, Azure
Market Data yFinance, Alpha Vantage
Social Data StockTwits API, Reddit API

Links

Tuyên bố miễn trừ trách nhiệm: Thông tin được cung cấp trong bài viết này chỉ nhằm mục đích giáo dục và thông tin, không cấu thành lời khuyên về tài chính, đầu tư hoặc giao dịch. Giao dịch tiền mã hóa tiềm ẩn rủi ro thua lỗ đáng kể.

Tác Giả

Eugen Soloviov
Eugen Soloviov

Trading-systems engineer

Trading-systems engineer building bots since 2017: cross-exchange arbitrage (connected up to 30 venues), cointegration-based pairs arbitrage across spot and futures, scalping, news and sentiment-driven strategies, trend algorithms, and portfolio management and balancing algorithms. Also builds sub-millisecond order execution, big-data warehouses, backtesting engines, AI agents, and trading interfaces (incl. open-source profitmaker.cc). Stack: JS/TS, Python, Rust/Zig/Go, DevOps, backend, frontend, architecture.

Newsletter

Đi Trước Thị Trường

Đăng ký nhận bản tin của chúng tôi để có những thông tin chuyên sâu độc quyền về AI trading, phân tích thị trường và các cập nhật nền tảng.

Chúng tôi tôn trọng quyền riêng tư của bạn. Hủy đăng ký bất kỳ lúc nào.