Build in Public ·

Using AI to Create an Impressive Stock Tools Site for Investors, Swing Traders, and Day Traders

KB Cafe is a build-in-public shop: we use AI to build real sites, then write about how it went. Our newest one is a big one, and it is live. Meet StockTools.AI, a free stock-research platform for investors, swing traders, and day traders, built almost entirely with AI on top of public SEC data. This is the tour: what it does, who it is for, and how AI helped put it together.

The one-line pitch

StockTools.AI calls itself “the research edge Wall Street pays for, free.” The idea is simple and a little contrarian: most finance sites are a data layer, endless quotes and charts. StockTools aims to be the decision layer, one connected workspace where SEC filings, insider trading, congressional trades, institutional ownership, executive pay, earnings, and AI research all live together and hand off to each other. Research any public company in one place, then act on evidence instead of vibes.

The part that hooks people: who is actually buying

The strongest section is the smart-money group, the answer to “forget the pundits, who is putting real money in?” It is all built from primary filings, not scraped estimates:

  • Congressional stock trades, straight from STOCK Act disclosures, with each trade’s return since it was filed versus the S&P 500, the filing delay, and late-filer flags.
  • Insider trades from SEC Form 4, the exact shares and prices executives and directors bought or sold, with cluster-buy signals, refreshed daily.
  • Whale 13F holdings, what the big institutions own (Buffett, Burry, Ackman, Dalio and the rest), plus a guru portfolios view of legendary investors’ latest buys.
  • A single money-flow feed that merges 13F changes, Form 4 insider activity, and congressional trades into one ranked stream of big buys and sells, every row linked back to its source filing.
  • A CEO pay tracker pulled from DEF 14A proxy statements: salary, bonus, and stock awards, the real numbers from the filing, no net-worth guesses.

AI research that cites its sources

This is where StockTools overlaps most with what we care about at KB Cafe: using AI honestly. The StockTools AI section is pitched as “not another chatbot.” You point it at a stock and it reads every underlying engine, filings, Congress, insiders, 13Fs, exec pay, valuation, red flags, then synthesizes a brief where every conclusion traces back to the exact filing it came from. Its motto is “conclusion left, evidence right.” A few flavors:

  • The Research Brief: a plain-English overview and auto-SWOT for any ticker, built only from SEC filings and market data. Every point is cited, and it makes no price predictions.
  • The Research Committee: a bull, a bear, a risk manager, and a portfolio manager debate a stock using only its filings. You get the arguments from every side, cited, with no buy or sell call at the end. It is a genuinely clever way to use AI: not to tell you what to think, but to stress-test what you already think.
  • The AI Trade Planner (paper, not advice): fuses the filings, the health checks, and the smart-money engines into a single conviction score, then lays out a sized paper plan with an honest backtest. It never places a trade; it is a thinking tool.

The whole thing runs on the Anthropic Claude API, and here is a nice engineering touch: each shared AI write-up is generated once and cached against the hash of its source filing, so thousands of readers share one generation. Good for cost, good for consistency.

Built for how you actually trade

The site has three front doors, one for each kind of trader:

Day traders

A small-cap research desk built for speed: live US trading halts with the reason and resume time, plus quick reads on dilution risk, float, cash runway (“do they need money?”), and a filing-based red-flag health score. Then the risk math you should do before every trade: position size, max loss, R multiples, and risk of ruin, all in the free calculator set. There is even a free trading journal with a P&L calendar so you can grade your own decisions over time.

Swing traders

Risk-reward and timing tools, an earnings calendar derived from each company’s own 8-K cadence (so it confirms on the filing, not on a rumor), and industry hubs that rank each sector by market cap and overlay the insider buying and Congress trades happening across it. A fast way to see where the smart money is leaning within a theme.

Long-term investors

The fundamentals kit: a DCF calculator, a Graham number, WACC and Sharpe, dividend and compounding math, and a side-by-side compare view that scores companies on the classic forensic checks (Piotroski F-Score, Altman Z-Score, Beneish M-Score). Every number is built from roughly fifteen years of EDGAR financials, not a data vendor’s estimate.

The rest of the menu

A few more that are worth a click. Options explained walks through calls, puts, premium, and the Greeks with a free dashboard that computes delta, gamma, theta, and vega live. The Quant Research Library decodes the newest arXiv quantitative-finance papers into plain English. The Lab shows every scoring formula it uses, no black boxes, and there is an IPO radar that reads fresh S-1 prospectuses. To learn the vocabulary, there is a worked-example glossary and study flashcards, the same learn-by-doing idea we lean on here.

How AI actually helped build it

The title of this post is honest: AI did a lot of the work. The site is a Next.js app on Vercel, backed by Postgres, that turns raw SEC EDGAR filings (Form 4, 13F, 8-K, S-1, DEF 14A) into clean datasets. AI wrote large amounts of that plumbing, designed and iterated the interface (there is a whole gallery of AI-generated design explorations in the repo), and powers the research features themselves. The build was documentation-driven, an architecture doc, a strategy doc, and a numbered log of design decisions, which is exactly the kind of setup that lets an AI coding agent stay coherent across a project this size. It is a real example of the thing we keep writing about: writing real code with AI instead of fighting it, and wiring AI into a product rather than bolting on a chatbot.

The honest part

We like StockTools for the same reason we like building here: it is upfront. The data comes from primary filings with the methodology shown, the AI cites what it read and refuses to predict prices or hand you a buy signal, and the footer says the quiet part out loud, these are educational tools, not financial advice, and AI can make mistakes, so check anything that matters. That is the right posture for money.

Go poke around

If you research stocks at all, StockTools.AI is worth ten minutes. Start on the money-flow feed to see who is buying right now, run a company through the bull-versus-bear committee, and grab a position-size calculator for your next trade. It is free, it is filing-backed, and it is a genuinely impressive example of what a small team plus AI can ship. We are proud to have it in the network.