Most financial data providers now offer MCP servers so Claude can connect to them directly. But the connection alone does not solve the real problem. If the MCP server still calls a rate-limited API behind the scenes, an AI model screening 10,000 stocks still needs 10,000 requests. The bottleneck is the architecture, not the protocol.
Shibui Finance is a free MCP server that connects Claude to 64 years of US stock market data: 31M+ daily prices, quarterly financials, daily valuations, and 56 technical indicators for nearly 10,000 companies. No API key, no rate limits, no per-request costs. Claude writes SQL queries against the pre-loaded database and returns results in seconds.
What people mean when they search for a financial data API
The search usually starts with a specific need: get historical stock prices into a model, pull quarterly earnings for a set of companies, or feed valuation data into an analysis pipeline. The expected solution is a REST API with JSON responses, authentication via API key, and some rate limit on a free tier.
That model works well for applications that need one data point at a time: a dashboard showing a single stock's price chart, a portfolio tracker refreshing positions, or a notification system watching a watchlist. These are request-response workflows. The application knows which ticker it wants before it asks.
AI workflows are different. When you ask Claude "find US stocks with P/E under 15 that grew revenue every year for five years," Claude does not know which tickers match until it has checked all of them. That is a cross-market scan, not a single-ticker lookup. Traditional APIs were not built for this.
Financial data API providers compared
These are the providers most commonly recommended for stock market data. Many now offer MCP servers alongside their REST APIs, so Claude can connect to them directly. But the MCP server matters less than what sits behind it: most wrap the same rate-limited API, while one pre-loads the data into a database.
| Provider | Free tier | Rate limit | Coverage | MCP server | Architecture |
|---|---|---|---|---|---|
| Alpha Vantage | 25 req/day | 25/day | Global stocks, forex, crypto | Yes | API-forwarding |
| Financial Modeling Prep | 250 req/day | 250/day | US + global, fundamentals | Yes | API-forwarding |
| EODHD | 20 req/day | 20/day | 60+ exchanges, 150K tickers | Yes | API-forwarding |
| Polygon.io | 5 req/min | 5/min, delayed | US stocks, options, crypto | Yes | API-forwarding |
| Twelve Data | 800 req/day | 8/min | Global stocks, forex, crypto | Yes | API-forwarding |
| Finnhub | 60 req/min | 60/min | US + global, alt data | Yes | API-forwarding |
| Tiingo | 1,000 req/hr | 1,000/hr | US + some intl, fundamentals | Yes | API-forwarding |
| Shibui Finance | Unlimited | None | ~10,000 US equities, 64 years | Yes | Pre-loaded database |
The "MCP server" column matters because having one does not mean the rate limits go away. An API-forwarding MCP server translates the model's request into the same REST API calls you would make yourself. The rate limit, the API key requirement, and the per-request cost all carry through. It is a convenience wrapper, not a different data model.
Why APIs break down for AI workflows
Consider a simple question: "Which US stocks with a market cap above $5 billion have grown revenue every quarter for the last two years?" Whether Claude calls a REST API directly or uses an API-forwarding MCP server, it still needs to:
- Fetch a list of all US stocks (1 request)
- Filter by market cap (1 request per stock, or a screening endpoint if available)
- Fetch 8 quarterly revenue figures per remaining stock (8 requests each)
- Compare quarters sequentially to check for continuous growth
Even with a generous API, that is thousands of requests. At 25 per day (Alpha Vantage free tier), the query would take months. At 60 per minute (Finnhub), it would still take over an hour and exhaust daily quotas. Most AI conversations expect answers in seconds.
"Which US stocks above $5B market cap grew revenue every quarter for the last 2 years?"
Claude writes a single SQL query that joins the valuation and financials tables, filters by market cap, checks 8 consecutive quarters of revenue growth, and returns the results. One query, a few hundred milliseconds, no rate limit. The difference is not optimization. It is a different data model.
API-forwarding vs pre-loaded: the architecture that matters
Most financial data providers now offer MCP servers, which is the right direction. But having an MCP server does not solve the rate-limit problem if the server still makes one API call per data point behind the scenes. There are two architectures:
API-forwarding MCP (Alpha Vantage, FMP, EODHD, Polygon, Twelve Data, Finnhub): the MCP server receives the model's request, translates it into REST API calls, and returns the results. The rate limits, API keys, and per-request costs all carry through. This is convenient (Claude connects directly instead of you writing code), but the screening bottleneck remains. A cross-market scan still hits the API thousands of times.
Pre-loaded database MCP (Shibui Finance): the data is already in a database. Claude writes SQL queries against it directly. No API call per request, no rate limit, no per-query cost. A screening query that would require 10,000 API calls runs as a single database query in milliseconds.
"Compare Tesla and Ford: revenue growth, margins, P/E, and debt-to-equity for the last 5 years, side by side."
With an API-forwarding MCP server, that comparison still requires separate API calls for each metric, each company, each year. With Shibui, Claude writes one query joining the fundamentals and valuation tables. The model controls the SQL, so it can adapt the comparison to whatever you ask without hitting fixed endpoint structures. See the full MCP server comparison for a deeper breakdown.
What Shibui Finance covers
Shibui connects Claude to a single, pre-built database covering US equities on NYSE and NASDAQ:
- Daily prices: 31M+ rows back to 1962, OHLCV for ~10,000 symbols
- Quarterly and yearly financials: income statements, balance sheets, cash flow
- Daily valuations: P/E, P/B, market cap, enterprise value, PEG ratio (daily history back to 1993)
- Derived metrics: EV/EBITDA, P/S, FCF yield, dividend yield, earnings yield
- Technical indicators: 56 indicators including RSI, MACD, Bollinger Bands, moving averages, ADX, candlestick patterns
- SEC filings: 6.4M filing records back to 1993, every form type, updated daily
Data refreshes daily after market close. Setup takes about two minutes with the step-by-step guide, and requires no API key, no account, and no payment. For full schema details, see the data sources page.
"Screen for US stocks above $10B market cap with RSI below 30, positive free cash flow, and P/E under 20."
That query touches four tables (stock_quotes, technical_indicators, fundamentals, valuation) and returns in under a second. With a traditional API, you would need separate calls for each data type for each of ~10,000 stocks.
When you still need a traditional API
Shibui is not a replacement for every financial data API. It serves a specific use case well (AI-driven analysis of US equities) and does not pretend to cover everything:
- Real-time or intraday data: Shibui is end-of-day only. If you need tick-level or streaming data, Polygon.io or Twelve Data are built for that.
- International markets: Shibui covers US equities (NYSE, NASDAQ). For global coverage across 60+ exchanges, EODHD or Finnhub are stronger options.
- Options, crypto, forex: Not in Shibui's database. Alpha Vantage and Polygon cover these.
- Programmatic scheduled access: If you need to call an endpoint from a cron job or pipeline (not an AI assistant), a REST API is the right tool. Shibui is designed for interactive AI sessions.
- Custom application backends: Shibui has no REST endpoint. If you are building a web application that needs to serve financial data to a frontend, use a traditional API.
The honest positioning: Shibui is built for one thing - AI assistants analyzing US equity data across companies and time periods. It does that well. It is not a general-purpose financial data API.
Frequently asked questions
Is there a free financial data API with no rate limits?
Shibui Finance provides free access to 64 years of US equity data with no rate limits. Unlike traditional APIs, it connects Claude to a pre-loaded database via MCP, so there are no per-request costs and no throttling. Alpha Vantage offers a free tier limited to 25 requests per day. Financial Modeling Prep and EODHD also have restricted free tiers.
What is the best free stock data API for AI?
For AI assistants like Claude, an MCP server outperforms a traditional API. Shibui Finance connects Claude to a pre-loaded financial database: no API keys, no rate limits, and the model writes SQL directly instead of making thousands of individual HTTP requests. For programmatic access outside of AI, Alpha Vantage and Financial Modeling Prep both offer free tiers with broad coverage.
Can Claude access financial data without an API key?
Yes. Shibui Finance requires no API key, no account registration, and no configuration beyond a one-click connection. It gives Claude read access to nearly 10,000 US equities with 64 years of daily prices, quarterly financials, daily valuations, and 56 technical indicators. Setup takes about two minutes.
What is MCP and how does it compare to a REST API?
MCP (Model Context Protocol) is a standard for connecting AI models to external data sources. Where a REST API returns one response per HTTP request, MCP gives the model direct access to query a database. For financial analysis, this means Claude can screen 10,000 stocks in a single query instead of making 10,000 API calls. MCP is to AI what REST was to web applications: the native protocol for the platform.
Does Shibui Finance have a REST API?
No. Shibui Finance is an MCP server, not a REST API. It connects directly to Claude and other AI assistants that support MCP. The model writes SQL queries against the database and gets structured results back. If you need a traditional REST API for a dashboard or application, Alpha Vantage, Financial Modeling Prep, or Polygon are better options.
What financial data providers work with Claude?
Several providers now offer MCP servers for Claude: Shibui Finance (pre-loaded US equity database, free), Alpha Vantage (API-forwarding MCP, free tier with limits), Financial Modeling Prep (API-forwarding MCP, free tier with limits), and EODHD (API-forwarding MCP, 20 calls/day free). Shibui is the only one that pre-loads the data, meaning queries run against the database directly with no external API call per request. See the full comparison.