Plan: financialdatasets.ai Integration

Context

financialdatasets.ai provides a REST API and Claude Code MCP server with access to 17,000+ stocks, financial statements (income statement, balance sheet, cash flow), and crypto prices. No Bloomberg Terminal required.

MCP setup (Claude Code dev sessions):

claude mcp add --transport http financial-datasets https://mcp.financialdatasets.ai

Then run /mcp in Claude Code to complete the OAuth flow. After setup, Claude can pull live fundamentals during development sessions (e.g. for ad-hoc research or the /buffett skill).

REST API docs: https://docs.financialdatasets.ai

Where it adds value

Area	Benefit
ML feature engineering	Add fundamental features: P/E, EPS growth, revenue growth, debt/equity — currently the model is price + sentiment only
Buffett scorecard	Replace proxy signals (RSI/BB as "safety") with real financial statement data
Crypto	Expand the watchlist beyond US/KR equities
Broader universe	17,000+ tickers vs. the current curated ~40

Integration approach (when ready)

For production scheduled jobs, use the REST API directly in Python — not the MCP server, which is only available in Claude Code sessions.

1. Add FINANCIALDATASETS_API_KEY to .env and config.py
2. Create scraper/financialdatasets.py — fetch quarterly fundamentals per ticker
3. Store in a new Fundamentals DB table (or extend StockPrice)
4. Add fundamental features to build_features() in predictor.py (remember to update FEATURE_ORDER and delete stale .pkl files)
5. Add a new scheduler job to refresh fundamentals (quarterly data — daily refresh is fine)

Status

Not started. No API key obtained yet.