AI Berkshire

Active
GitHub Python MIT

Description

An AI-era value investing research framework based on Claude Code/Codex, integrating methodologies of four masters (Buffett, Munger, Duan Yongping, Li Lu) with multi-agent parallel research.

Key Features

  • Multi-agent parallel - simultaneously call multiple Claude Code/Codex instances to analyze different dimensions in parallel
  • Four-master methodology - built-in investment frameworks and decision checklists from Buffett, Munger, Duan Yongping, Li Lu
  • Deep financial analysis - auto-parse 10-K, 10-Q, annual reports, extract key financial metrics
  • Valuation model integration - built-in DCF, relative valuation, asset valuation three mainstream methods
  • Risk assessment - quantitative models evaluate moat, debt ratio, ROIC and other risk indicators
  • Decision records - complete investment decision logs for retrospective review

Use Cases

💡 Value investors use AI Berkshire for in-depth single-stock research, generating multi-dimensional investment reports.
💡 Investment researchers use it to batch-track multiple targets, producing weekly full-market scanning reports.
💡 Investment education scenarios as assistant tool, letting students follow multi-Agent to learn master thinking frameworks.

Quick Start

```bash
# Clone the repo
git clone https://github.com/xbtlin/ai-berkshire.git
cd ai-berkshire

# Load in Claude Code
claude --system-prompt "$(cat prompts/system.md)" \
  --tools "$(cat tools/financial.yaml)" \

# Start multi-agent research
./run.sh --stock AAPL --masters all
```

Related Projects