The predictions are the headline; the alternative data is what is genuinely hard to find in one place — congressional trades with the disclosure lag preserved, lobbying filings, patents by assignee, contract awards, Reddit mentions and app-store ratings. The paired design pays off in use: every per-ticker dataset has a screener twin, so moving between "what about this company" and "who else looks like this" does not mean changing tools. The CIK carried on the regulatory datasets is the small detail that makes any of it joinable to EDGAR.
A local server over the FinBrain v2 API that exposes its datasets in pairs: one tool per ticker and a screener twin covering the same signal across a market or region. Coverage runs from machine-learning price predictions to alternative data that rarely reaches an assistant — congressional trading, corporate lobbying, USPTO patents and US government contract awards.
- predictions_by_ticker returns the forecast as a time series of mid, low and high alongside short, mid and long expectations, and predictions_by_market screens predictions across a market or region
- News and sentiment: news_by_ticker for articles, news_sentiment_by_ticker as a dated score series, recent_news across every tracked stock, and screener versions of both
- Institutional and insider activity — analyst ratings with rating type, institution, signal and target-price movement; insider transactions from SEC Form 4 with relationship and transaction type; corporate lobbying filings with registrant, income, expenses and issue codes
- US Congress trades from both chambers, carrying the trade date and the public disclosure date so the reporting lag is measurable, the beneficial owner of the account, and amounts normalised to the statutory STOCK Act brackets with the original filing preserved
- Alternative signals: LinkedIn employee and follower counts, App Store and Play Store ratings, options put/call ratio with call and put counts, Reddit mention counts per subreddit, USPTO granted patents with CPC classification, and government contract awards with agency and amount
- A screener for each of those, so the same signal can be scanned across a market or region rather than ticker by ticker
- Rows in the insider, government-contract, lobbying and patent datasets carry cik — the SEC Central Index Key as a zero-padded ten-digit string — so they join to EDGAR filings or a security master
- available_markets, available_tickers and available_regions map the coverage before a query is written, health resolves the API key and reports versions, and every series is paged, JSON by default with CSV on request
Python 3.10 or newer, published on PyPI as finbrain-mcp and launched over stdio. It runs locally against your own FinBrain API key in FINBRAIN_API_KEY, with nothing proxied through a third party, and reaches FinBrain through the official finbrain-python SDK on the v2 API.
One command plus a key — pip install --upgrade finbrain-mcp, then supply credentials