How to actually analyze stocks with Claude (not just ask about them)

#ai #finance #mcp #claude

You can now ask Claude "What's the technical outlook on NVDA?" and get a real answer backed by live data — not a disclaimer.

Here's how, and what the output actually looks like.

The setup

Install FinanceKit MCP (one command):

claude mcp add financekit -- uvx --from financekit-mcp financekit

That's it. Claude now has access to 17 financial tools running locally.

What you can ask

Single stock analysis:

Analyze AAPL technically. Is now a good entry point?

Claude calls technical_analysis("AAPL") and returns something like:

AAPL Technical Analysis — April 2026

Signal: MODERATELY BULLISH (2.0/3.0 signals)

RSI: 61.4 — healthy momentum, approaching but not overbought
MACD: Crossed above signal line 3 days ago (bullish crossover)
Bollinger Bands: Price in upper half, moderate volatility
ADX: 24.1 — trend present but not strong
ATR: 2.8 — normal volatility range

Assessment: Momentum is positive. The MACD crossover is confirmed
but ADX suggests the trend isn't strongly established yet. RSI has
room before overbought territory. Reasonable entry, but watch for
a pullback to the 20-day MA if the broader market weakens.

That's the output you actually get — not "past performance doesn't guarantee future results."

Portfolio risk check:

I have 40% AAPL, 30% NVDA, 20% MSFT, 10% GOOGL. What's my real risk?

Returns: correlation matrix, VaR (how much you could lose on a bad day), Beta vs S&P 500, Sharpe ratio, Max Drawdown over the past year.

Quick crypto check:

What's Bitcoin doing? And give me the top 5 trending coins right now.

Returns real-time BTC price + 24h change, plus live trending list from CoinGecko.

Compare two stocks:

Compare AAPL vs MSFT on risk-adjusted returns over the past 90 days

Returns side-by-side: Sharpe ratio, volatility, Beta, price performance.

What the tools actually compute

For those who want to know what's under the hood:

Tool	Data source	Computation
`technical_analysis`	yfinance (daily OHLCV)	RSI (14d), MACD (12/26/9), Bollinger (20d, 2σ), ADX (14d), ATR (14d)
`portfolio_analysis`	yfinance	Rolling covariance matrix, VaR via historical simulation, Sharpe/Sortino/Beta
`crypto_price`	CoinGecko	Real-time price, 24h change, market cap
`options_chain`	yfinance	Live options data from exchanges

All free data sources. Zero API keys required.

Important limitations

This is analysis, not advice. The numbers are real — computed from actual market data — but:

yfinance data can occasionally lag by 15-20 minutes
Technical indicators work better on liquid stocks (not penny stocks)
Portfolio risk metrics assume normal distribution of returns (they don't account for fat tails / black swans)
Past volatility doesn't predict future volatility

Use it to augment your research, not replace it.

Installation

# Claude Code / Terminal
claude mcp add financekit -- uvx --from financekit-mcp financekit

# Claude Desktop (claude_desktop_config.json)
{
  "mcpServers": {
    "financekit": {
      "command": "uvx",
      "args": ["--from", "financekit-mcp", "financekit"]
    }
  }
}