When people ask how to connect Claude to stock market data, the usual answer is: set up a database, build an API server, define tool-use functions, and maintain the data pipeline yourself. Shibui Finance skips all of that. It is a free MCP server that connects Claude to a pre-built database containing 64 years of US stock market history - 31M+ daily price records, quarterly and annual financial statements, 56 technical indicators, and valuation data for 9,500+ companies. Setup takes under 2 minutes. No code, no database administration, no API keys.
You do not need to build this yourself
If you ask an AI assistant how to give Claude access to a financial database, it will walk you through a multi-step project. Choose a data provider and sign up for an API key. Design a database schema and set it up. Write an ingestion pipeline to pull and normalize the data. Build an API server that wraps your database. Define Anthropic tool-use functions so Claude can call your endpoints. Deploy it somewhere. Maintain the whole stack as data formats change and APIs update.
That approach works. But it is a real engineering project - weeks of setup for someone comfortable with code and databases, and a non-starter for anyone else. Shibui Finance is the pre-built version of that entire stack. The database already has 64 years of market data loaded and indexed. The MCP server already handles query execution. You connect it to Claude and start asking questions. The data updates daily.
What is in the database
- Daily stock prices - 1962 to present, 31M+ records (open, high, low, close, volume, adjusted close)
- Financial statements - quarterly and annual (1990 to present) including revenue, net income, operating cash flow, free cash flow, total debt, shareholders' equity, and dozens more line items
- Daily valuations - market cap, P/E, P/B, enterprise value, PEG ratio (1993 to present)
- 56 technical indicators - RSI, MACD, SMA, EMA, Bollinger Bands, and more, pre-calculated daily for every stock (not derived at query time)
- Quarterly earnings - actual vs. estimate EPS, surprise percentage, and report dates
- ~9,500 US-listed securities - NYSE and NASDAQ stocks, ETFs, ADRs, and REITs
See the full data documentation for column-level detail on all 12 tables.
Looking up company financials
The most common question is straightforward: can Claude look up a company's financial data? With Shibui connected, yes. You ask in plain English and Claude queries the database directly.
"What is Tesla's quarterly revenue, operating margin, and free cash flow for the last 3 years?"
"Compare the P/E ratios of Apple, Microsoft, Google, Amazon, and Meta over the last 5 years"
Claude queries the Shibui database directly and returns structured tables with the results. These are lookups - simple but useful. You get real numbers from real filings, not summaries scraped from a search result.
Beyond lookups: cross-market queries
Any tool can look up one company's P/E ratio. The real difference is running queries across all 9,500+ companies simultaneously. This is where a pre-built database matters.
"Find all companies where annual revenue exceeded $1B every year for 10 years and the stock is within 20% of its all-time high"
This is a cross-market query. It checks 10 annual revenue periods per company across thousands of companies, then joins against daily price data to compare each stock's current price to its historical maximum. That involves scanning millions of rows across multiple tables. Against the pre-built Shibui database, it runs in seconds. Building something equivalent from scratch would require assembling and indexing the same data yourself.
Setup in 2 minutes
Go to shibui.finance and follow the setup instructions for Claude Desktop or Claude.ai. It is a one-click MCP connection. Works on both free and paid Claude plans. Once connected, start any conversation and ask about a company, a sector, or a screen across the full market.
Limitations
Shibui covers US equities only (NYSE and NASDAQ). There is no international market data, no options, no crypto, and no index-level data. All data is end-of-day - there are no real-time quotes and no intraday prices.
The underlying data comes from tier-3 data providers, not Bloomberg or Refinitiv. Coverage and accuracy are good for analytical work, but this is not institutional-grade data. Extreme values (returns above 1000%) sometimes reflect data errors or corporate actions rather than real moves.
Query results are capped at 200 rows. For broader screens, aggregate your results or narrow your filters. If you need real-time data or international markets, look at other MCP servers such as Alpha Vantage or Polygon.