HN 표시: API 청구서가 없는 TradingAgents – Claude Code에서 다중 에이전트 실행
hackernews
|
|
🏗️ 프레임워크
#ai 모델
#anthropic
#claude
#openai
원문 출처: hackernews · Genesis Park에서 요약 및 분석
요약
Claude Code 플러그인을 통해 외부 LLM API 비용 없이 Claude 구독만으로 멀티 에이전트 트레이딩 분석이 가능하며, yfinance를 통해 무료로 시장 데이터를 가져옵니다. BEAR 분석에서는 높은 베타 수치와 붐비는 롱 포지션을 지적했으나, VERDICT은 펀더멘털 약세 없이 불리하다고 판단했습니다. 리스크 분석가는 군집 투자와 베타 하락 시나리오를 고려해 초기 포지션 규모를 2~3%로 줄이고 단계적으로 진입할 것을 제안했습니다.
본문
Multi-agent trading analysis as a Claude Code plugin — powered entirely by your Claude subscription, no external LLM API cost. /plugin marketplace add lucemia/trading-agents-plugin /plugin install trading-agents-plugin Then install the Python data dependency (fetches market data via yfinance): git clone https://github.com/lucemia/trading-agents-plugin cd trading-agents-plugin uv sync # or: pip install yfinance pandas /trading-analysis NVDA /trading-analysis GOOG /trading-analysis SOXL Phase 1 (parallel): Technical + News + Fundamentals + Macro analyst Phase 2 (sequential): Bull → Bear rebuts Bull directly → Risk Analyst stress-tests both Phase 3: Research Manager synthesizes all three voices Phase 4: Trader sets entry / stop / size anchored to technicals Phase 5: Portfolio Manager → final BUY / SELL / HOLD 7 subagents total. Each phase uses Claude Code's native Agent tool. Data is fetched for free via yfinance — no OpenAI, Anthropic, or any other LLM API key required. | Analyst | Coverage | |---|---| | Technical | EMA10/SMA50/SMA200 trend, RSI14, MACD, Bollinger Bands, ATR, key support/resistance | | News | Top headlines, sentiment, sector tailwinds/headwinds, earnings signals | | Fundamentals | P/E, forward P/E, PEG, P/B, revenue/earnings growth, margins, ROE, FCF, D/E ratio, current ratio, beta, short ratio, quarterly income/balance sheet statements, analyst consensus | | Macro | S&P 500, 10Y Treasury, gold, oil news — geopolitical/monetary policy signals | | Bull | Strongest buy case using all Phase 1 data | | Bear | Directly rebuts Bull's specific claims with data | | Risk | Stress-tests both sides — crowding risk, beta drawdown scenarios, tail risks neither Bull nor Bear addressed | TICKER: NVDA DATE: 2026-05-02 SIGNAL: BUY RATING: Overweight ENTRY: $197–$199 (tranche 1); $188–$190 (tranche 2 if macro deteriorates) STOP: $183 (below SMA200) SIZE: 2–3% of portfolio initial; reserve 2% for second tranche BULL: PEG 0.18, FCF $58.1B, net cash $51B, and 57 analysts pricing in 35% upside — NVDA's AI dominance is structurally mispriced. BEAR: Beta 2.33 in RISK-OFF macro with crowded longs means a guidance miss could trigger a 20%+ unwind against a 30.7x P/B backdrop. VERDICT: Bull wins on fundamentals — no operational weakness identified. Risk Analyst flagged crowding and ~23% beta drawdown scenario, reducing sizing to 2–3% initial with staged second entry near SMA50. - Claude Code with an active subscription - Python 3.10+, uv (or pip) If you prefer not to use the plugin system: git clone https://github.com/lucemia/trading-agents-plugin cd trading-agents-plugin uv sync mkdir -p ~/.claude/commands cp .claude/commands/trading-analysis.md ~/.claude/commands/ Then update the script path in ~/.claude/commands/trading-analysis.md to match your local clone path. Inspired by TradingAgents — a multi-agent LLM trading framework by Tauric Research. This project adapts their analyst pipeline design as a native Claude Code plugin, replacing the LangGraph/OpenAI stack with Claude's built-in Agent tool and yfinance for free market data. MIT
Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.
공유