For first-time buyers — understand if a business is worth buying before you commit
Upload files or paste text — or both. The more you share, the better the assessment.
Upload financial documents or fill in what you know. Skip anything you don't have yet.
Understanding day-to-day operations is crucial before buying.
Helps tailor the assessment to your situation as the buyer.
Scoring all the key areas and preparing your report...
I wanted to build something for someone in the exact spot a friend of mine was in: a first-time buyer looking at a small business for sale, with no idea what questions to ask or what to be worried about. Most people in this position either pay an accountant a few hundred dollars just to get a gut-check, or they skip that step entirely and find out about problems after they've signed.
Who it's for: first-time buyers looking at a small business (cafes, retail shops, trades, online businesses) before they make an offer. What you get: a structured score across financials, operations, market, and risk, a plain-English explanation of the verdict, the exact questions to put to the seller, and — if you choose to add your own Claude API key — a second, deeper pass where Claude actually reads your uploaded documents instead of just your form answers.
The tool is a 5-step form. As you answer, nothing is sent anywhere — it all stays in your browser. When you hit "Generate My Assessment," a scoring engine I wrote in plain JavaScript adds and subtracts points across four areas (financial health, operations, market/customers, risk) based on rules like "owner refused to share financials → big red flag" or "long lease + repeat customers → good sign." That gives you an instant, free, no-AI-needed result.
If you want to go further, there's an optional "Run AI Assessment" step. That's where it gets more interesting: it takes everything you typed or uploaded (PDFs and Word docs are read directly in your browser using PDF.js and Mammoth.js — no file ever leaves your computer except as plain text sent to Claude) and sends it to Claude's API, asking it to actually read the listing, the job ads, and the financials, and come back with its own assessment in a strict format my code knows how to display. You need to supply your own Anthropic API key to use this part, since this is a static portfolio site with no server of its own to pay for API calls on your behalf — your key is used only for that one request and is never saved.
This is the most advanced thing I've built so far. My earlier projects (the financial health calculator, the artist roadmap) were written with Claude's help but only run logic in the browser — no live AI calls. This one is the first time I've actually wired a live call to Claude's API into something I built, the first time I've handled real file uploads (PDF/Word), and the first time I've had to think properly about handling someone else's API key responsibly.
Getting Claude to reliably reply in a strict JSON format took a few tries — the prompt has to ask for it explicitly and my code still has a fallback to search for a JSON object in the reply in case Claude wraps it in extra text. I also learned that reading PDFs and Word docs entirely in the browser is genuinely possible without a backend, which changed how I think about what a "simple static website" can actually do. And building the free rule-based version first, before adding the AI layer, taught me to separate "does the logic make sense" from "does the AI call work" — debugging one at a time was much easier than building both together.