Portfolio Tracker & Analyzer

A local-first stock portfolio tracker with a built-in MCP server for Claude Desktop. Talk to Claude naturally — it pulls live prices, technicals, SEC filings, and news from your portfolio without you opening the app.

What is this?

Two tools in one repo:

No cloud. No subscription beyond your own API keys. Runs entirely on your machine.

MCP Server — Talk to Claude About Your Portfolio

Once connected, just ask Claude:

"Give me a full review of my portfolio this week."
"What's the technical setup for NVDA right now?"
"Any breaking news or SEC filings for my holdings today?"
"Compare GOOGL, META, and AMZN — which looks best technically?"
"What are the upcoming earnings dates for my holdings?"

Claude calls the right tools, fetches live data, and writes the analysis — all using your existing Claude subscription.

8 tools Claude can call

The technical analysis, news, SEC, and comparison tools work for any ticker — not just your portfolio holdings.

MCP Quick Setup

1. Install dependencies

pip install -r requirements.txt

2. Edit the Claude Desktop config

Mac: ~/Library/Application Support/Claude/claude_desktop_config.json Windows: %APPDATA%/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "portfolio-analyzer": {
      "command": "python",
      "args": ["/FULL/PATH/TO/portfolio-analyzer/mcp_server.py"]
    }
  }
}

3. Restart Claude Desktop — look for the 🔨 hammer icon in the chat bar.

See MCP_SETUP.md for the full guide including troubleshooting.

Streamlit App

Setup

# 1. Install dependencies (Python 3.9+)
pip install -r requirements.txt

# 2. Add your OpenRouter API key (for AI summaries)
cp .env.example .env
# Edit .env and add your key from https://openrouter.ai/settings/keys

# 3. Run
streamlit run app.py

Open http://localhost:8501 in your browser.

Pages

Dashboard — Total value, cost basis, P&L, allocation pie chart, holdings table

Manage Holdings — Add/remove positions with ticker validation

Stock Analysis — Candlestick chart with SMA/EMA/Bollinger overlays, RSI, MACD, volume, fundamentals, and an AI memo (Claude Sonnet 4.6)

Performance — Portfolio vs S&P 500 (normalized), volatility, Sharpe, drawdown, beta, returns distribution

Summary — Claude Sonnet 4.6 writes a personal investment memo (daily / weekly / monthly) covering winners/losers, SEC events, macro themes, and recommendations

Monitor — Real-time news alerts from SEC EDGAR, Business Wire, PR Newswire, Yahoo, and Google News — scored by market impact

Project Structure

portfolio-analyzer/
├── app.py               # Streamlit app (6 pages)
├── mcp_server.py        # MCP server for Claude Desktop
├── requirements.txt
├── .env.example         # API key template
├── data/
│   ├── portfolio.json   # Your holdings
│   └── summaries/       # Saved AI summaries
└── src/
    ├── analytics.py     # Portfolio metrics & risk
    ├── data_fetcher.py  # yfinance wrapper with caching
    ├── indicators.py    # SMA, EMA, RSI, MACD, Bollinger
    ├── llm.py           # Claude Sonnet 4.6 via OpenRouter
    ├── news_monitor.py  # Multi-source news fetcher + scorer
    ├── portfolio.py     # Load / save / add / remove holdings
    ├── sec_edgar.py     # SEC EDGAR API client
    ├── snaptrade_integration.py  # Auto-import holdings from your brokerage
    └── summary_pipeline.py  # Weekly/daily/monthly memo pipeline

Adding Holdings

There are three ways to get your positions into the tracker. Pick whichever you prefer.

Option 1 — Auto-connect your brokerage (SnapTrade) ⭐ recommended

Connect your broker once and import your real positions (shares + cost basis) automatically — no CSVs, no manual typing. Works with Robinhood, Fidelity, Schwab, Vanguard, E*TRADE, Webull, Wealthsimple, and many more.

1. Create a free SnapTrade account Sign up at dashboard.snaptrade.com and verify your email. The free tier gives real brokerage data for a personal account at no cost.

2. Generate an API key in the dashboard — you'll get a Client ID and a Consumer Key.

3. Add them to your .env:

SNAPTRADE_CLIENT_ID=your-snaptrade-client-id
SNAPTRADE_CONSUMER_KEY=your-snaptrade-consumer-key

4. Install the SDK (already included if you ran pip install -r requirements.txt):

pip install snaptrade-python-sdk

5. In the app: open the 🏦 Connect Brokerage tab on the Manage Holdings page → Connect Brokerage. A SnapTrade window opens; log into your broker there. Back in the app, click Check Connection, then 🔄 Sync Holdings to import.

Your broker credentials go directly to SnapTrade and never touch this app. Your SnapTrade user secret is stored locally in data/.snaptrade_user.json (gitignored). Note: SnapTrade provides shares + average cost basis, but not the original lot purchase date, so imported positions use a placeholder date. Your P&L is accurate.

Option 2 — Upload a CSV export

On the Manage Holdings page, use the 📥 Import CSV tab to upload a holdings export from your broker (Robinhood, Fidelity, Schwab, Vanguard, etc.). The app standardizes common brokerage CSV formats automatically.

If your broker's CSV isn't supported: You can easily generate the portfolio.json file using any AI assistant (like Claude or ChatGPT). Just attach your CSV and ask it to "Convert my brokerage CSV into this JSON format:".

Option 3 — Manual entry / edit the file

Use the 🟢 Buy / 🔴 Sell tabs on the Manage Holdings page, or edit data/portfolio.json directly:

{
  "holdings": [
    {
      "ticker": "AAPL",
      "shares": 10,
      "purchase_price": 150.0,
      "purchase_date": "2024-01-15"
    }
  ]
}

Indicator Cheat-Sheet

Notes

• Yahoo Finance data has a ~15 min delay on the free feed.
• API responses are cached for 5 min to avoid rate-limit issues.
• Performance history assumes you held current share counts for the full selected period — it's a "how would my current book have done?" view, not a true time-weighted return.
• AI features require an OpenRouter API key. The MCP server uses your Claude Desktop subscription directly (no separate key needed).
• Not investment advice.