I ask Claude about my investments way too often. “How does this ETF compare to that one?” “What’s the analyst take on this stock?” The problem? The answers come back stale or completely made up.
AI models have a knowledge cutoff. They don’t know what happened in the market today, last week, or last month. For investing decisions, that’s basically useless.
FinMCP is a financial data MCP server for ChatGPT Work, Codex, Claude, and Cursor. It gives AI workflows access to Yahoo Finance market, company, options, crypto, and forex data.
The gap
I started looking into Finance MCP servers — tools that give AI assistants live access to real data. Brilliant idea on paper. Reality check: all the free finance MCPs only work for US markets. Popular non-US markets (ASX, LSE, and most others) were out of reach.
What I built
I found Yahoo Finance has a free, undocumented API that covers everything — US stocks, ASX, crypto, forex, ETFs. No API key required. So I took an existing MCP wrapper, rebuilt it properly, and shipped FinMCP.
It’s a production-grade financial data server. Connect Claude, ChatGPT Work, Codex, or Cursor to current market data across any market Yahoo Finance touches. It can run as a local server, or be deployed to Railway or Docker and shared with mates over HTTPS.
What you get from this finance MCP server
15 tools covering what investors actually need:
- Real-time quotes across any market (
BHP.AX,AAPL,BTC-USD,EURUSD=X— up to 100 at once) - Crypto and forex quotes
- Financial statements (annual or quarterly)
- Options chains with Greeks (delta, gamma, theta, vega)
- Analyst consensus and price targets
- Institutional holdings data
- News with relevance scoring
- Market screener (filter by sector, P/E, yield, beta, etc.)
Real example: I asked Claude to compare free cash flow trends across three ASX mining companies over four years. It pulled the numbers and gave me a clear breakdown. Normally that takes 20 minutes of spreadsheet work.
Use FinMCP with your AI client
FinMCP works with ChatGPT Work, Codex, Claude, and Cursor, so you can use the same financial data MCP in the AI workflow you already use. Run it locally for personal workflows or deploy it with Docker or Railway when you need a hosted endpoint.
How to run FinMCP
Locally — Node.js 22+, clone the repo, point Claude Desktop at it. Five minutes.
In the cloud — Deploy to Railway with one click, get a public HTTPS URL, share with friends. Free tier is plenty for personal use.
Why Yahoo Finance
It’s free, global, and needs no API key. Alpha Vantage caps you at 25 requests per day. Polygon costs money. Everyone else is either US-only or paid.
Yahoo covers what matters — US, ASX, LSE, crypto, forex, ETFs.
The catch: it’s unofficial. But the circuit breaker, rate limiter, retry logic, and caching handle hiccups gracefully.
How FinMCP keeps data workflows reliable
FinMCP uses caching, rate limiting, retry logic, and a circuit breaker around the Yahoo Finance API. That keeps a temporary upstream issue from breaking an entire market-data workflow.
Automation ideas
- Pre-market brief — scheduled prompt every morning with your watchlist moves and news
- Portfolio health check — monthly scan for deteriorating cash flow or rising debt across holdings
- Options monitoring — track put/call ratios and unusual activity
- Earnings tracker — pull analyst estimates before results
- Dividend calendar — track ex-dividend dates for ASX income plays
- Sector rotation check — spot rotations by combining screener data with analyst sentiment
Get it
GitHub: github.com/mustafa-ramax/finmcp
MIT licensed and free to use. Open an issue if something’s broken.
If you’re working with current market data and tired of AI tools that don’t know what happened in the market today, this one solves that problem.
For personal market awareness rather than AI workflows, I built MarketPulseBar, a Mac menu bar market tracker.
FinMCP provides market data and research tooling, not investment advice. Verify important data before making financial decisions.