💰 Financial AI (agentic systems)

Agent-based architectures for trading, research, and financial decision-making.

---

🧠 Core frameworks (multi-agent trading)

• TradingAgents — https://github.com/TauricResearch/TradingAgents
  → Multi-agent trading system (bull / bear / risk roles)
  → Clear coordination between agents
  → Melhor ponto de partida hoje

• FinRobot — https://github.com/AI4Finance-Foundation/FinRobot
  → Plataforma de agentes para finanças
  → Equity research + reasoning
  → Mais “produção” que paper

---

🧠 Memory / context (stateful reasoning)

Financial-specific (concept-driven)

• FinMem (paper-based, ver seção abaixo)

General memory systems (usáveis hoje)

• MemGPT — https://github.com/cpacker/MemGPT
• Letta — https://github.com/letta-ai/letta

👉 essenciais para:

• memória de longo prazo
• decisões financeiras sequenciais

---

🤖 Agent frameworks (builders)

• AutoGen — https://github.com/microsoft/autogen
• CrewAI — https://github.com/joaomdmoura/crewai

👉 usados para:

• trading agents
• research agents
• risk agents

---

📊 Trading / quant infrastructure

• Lean (QuantConnect) — https://github.com/QuantConnect/Lean
  → Backtesting + execução

• Backtrader — https://github.com/mementum/backtrader
  → Estratégias e simulação

• Freqtrade — https://github.com/freqtrade/freqtrade
  → Bot de trading (crypto)

---

📡 Data / signals

• yfinance — https://github.com/ranaroussi/yfinance
• pandas-datareader — https://github.com/pydata/pandas-datareader

---

⚙️ Execution / integration

• ccxt — https://github.com/ccxt/ccxt
  → integração com exchanges

---

📄 Research papers (financial agents)

• FinAgent — Multimodal Foundation Agent for Financial Trading
  https://arxiv.org/abs/2402.18485
  → multimodal (price + texto)
  → tool use + memory + reasoning

• https://arxiv.org/abs/2405.14767
  → sistemas de agentes financeiros com LLM

• https://arxiv.org/abs/2311.13743
  → early-stage LLM trading agents

• TradingAgents — Multi-Agent LLM Financial Trading Framework
  https://arxiv.org/abs/2412.20138
  → bull / bear / risk agents
  → melhora Sharpe e drawdown

• QuantAgents — Multi-agent Financial System via Simulated Trading
  https://arxiv.org/abs/2509.09995
  → simulação + aprendizado contínuo
  → múltiplos agentes coordenados

---

📌 Architecture pattern

Financial agent systems typically include:

• Signal agents (price, news, indicators)
• Strategy agents (bull / bear theses)
• Risk agents (exposure, drawdown control)
• Execution agents (orders, timing)
• Memory layer (market + decision history)

---

🔥 Key insight

Financial AI is not about:

• predicting prices

It is about:

• managing decisions over time
• coordinating agents
• handling risk under uncertainty