forked from TauricResearch/TradingAgents

TradingAgents
Silicon Valley fork

A multi-agent LLM framework extended with a custom analyst that simulates how 8 legendary investors debate a stock — then synthesises their conflicting views into a single recommendation.

Python 3.11+ LangGraph Claude / OpenAI / Ollama yfinance Silicon Valley panel — custom

⚠ Not financial advice. Research and educational use only.

01 — WHY THIS FORKWhy it produces better analysis

🧠

8 distinct mental models, not one

The original framework runs standard technical + fundamental analysis. This fork adds 8 legendary investor personas — each with a radically different framework for evaluating a stock.

Adversarial debate within the panel

Buffett's moat thinking directly challenges Musk's first-principles skepticism. Jim Simons ignores both and looks at pure patterns. That tension surfaces blind spots a single analyst misses.

🎯

Nuanced, not just BUY / SELL

Because the personas disagree, the final output includes price targets, re-evaluation triggers, and position sizing guidance — not just a binary call.

🔬

Domain expertise built into the prompt

Jensen Huang's lens is particularly sharp on AI infrastructure stocks. Simons' quant frame works for technical divergences. The right persona is strong where others are weak.

02 — ARCHITECTUREHow the agents work together

Analyst team — parallel data gathering

Market analyst

RSI, MACD, Bollinger bands, price action

News analyst

Headlines, macro, insider transactions

Fundamentals

Balance sheet, cash flow, P/E, margins

Silicon Valley panel ✦

8 investor personas, debate each other

Silicon Valley panel — custom extension NEW IN THIS FORK
Warren Buffett

Durable moat, 20-year hold test

Jeff Bezos

Day 1 vs Day 2, compounding

Elon Musk

First principles, physics of biz

Tim Cook

Margins, supply chain, services

Sundar Pichai

AI integration, platform moat

Bill Gurley

Market size, rake, competition

Jensen Huang

Real AI vs hype, compute demand

Jim Simons

Pure quant — patterns only

personas debate each other → adversarial tension produces more nuanced output than any single analyst
Research team — structured debate

Bull researcher

Builds the upside case, challenges bear

Bear researcher

Identifies risks, challenges bull thesis

Trader agent

Trader

Synthesises all analyst and researcher reports into an initial investment proposal with position sizing guidance

Risk management — three-way review

Aggressive

Argues for higher risk tolerance

Neutral

Balances risk and return

Conservative

Capital preservation focus

Final output
BUY
HOLD
SELL
with price targets, re-evaluation triggers, and position sizing

03 — QUICKSTARTRun it in 4 steps

01

Clone this fork

git clone https://github.com/sanaships/TradingAgents
cd TradingAgents
02

Create a virtual environment (Python 3.11 required)

python3.11 -m venv venv
source venv/bin/activate
pip install -r requirements.txt
03

Add your API key

cp .env.example .env
# Add ANTHROPIC_API_KEY to .env
# Get one free at console.anthropic.com
04

Run an analysis

python3 run.py

Edit run.py to change the ticker or model.