Ashwin

TransactionsChatAI: An AI Analyst for Your PayPal Data

PayPal's dashboard tells you what happened. This tells you what it means — and lets you just ask.

If you run a PayPal-connected business, you know the feeling: you want to answer a normal question about it — who are my top customers this quarter, is my fee rate creeping up, what did last month actually net after refunds — and the honest answer is "export a CSV and go find out." PayPal's own reporting is built for statements and compliance, not for conversation. There's no way to just ask it something.

So I built TransactionsChatAI: connect a PayPal account, and get an AI-native way into your own transaction history — chat with it in plain English, run reports without writing a query, and get a written business-health summary on a schedule instead of only when you remember to go looking.

What Problems It Actually Solves

  • Data analysis without spreadsheets. A ChatGPT-like chat interface sits directly on top of your transaction data — ask a question, get an answer backed by a real query, instead of building a pivot table by hand. Example: "How much did I make last month?"
  • Insights in plain English — or whatever language you think in. No SQL, no query builder, no export step. Ask the way you'd ask a person, in the language you'd ask it in, and get a direct answer back. Example: "Who are my top 5 customers this quarter?"
  • Readymade reports for the questions every business actually has. Monthly cashflow, top customers, fee totals, refund rate — the report is already there, not something you rebuild from scratch every month. Example: open "Monthly Cashflow" and see money in vs. money out, already charted.
  • AI Analysis catches what you'd miss by not looking. A scheduled, structured business-health report that flags problems and surfaces opportunities in your own numbers — the "revenue is flat but fees crept up" that's easy to miss when nobody's actively checking. Example: "Revenue is flat, but fees rose 12% this month — worth a look."

The Product, Feature by Feature

  • Chat. Ask questions about your transactions in plain English — "how much did I make in March," "who are my biggest customers this quarter," "show me refunds over $50" — and get answers backed by real tool calls against your actual data, not a canned summary. Built on Claude's tool-use loop.

    Chat answering a question about top customers with a real data table

  • Prebuilt Reports. 27 one-click reports across six categories — Cash Flow, Customers, Transactions, Fees, Time Patterns, and Transaction Types. Each one renders a chart plus a table, and has its own "ask about this report" chat widget so you can follow up on what you're looking at without leaving the page.

    Monthly Cashflow prebuilt report showing a bar chart and data table

  • Custom Report Builder. When the 27 presets don't cover it: pick your own group-by dimensions (month, week, day, counterparty, status, currency, item), metrics (count, sum, average, fees), filters, and sort order, and get a report on demand — same chart-plus-table output as the prebuilt ones.

    Custom report builder with group-by, metric, and filter controls

  • AI Analysis. A periodic, structured business-health report — revenue trends, concerns, highlights — computed first from real numbers and then handed to Claude to turn into a written narrative. This runs on a schedule (and on demand), so insight shows up whether or not you remembered to go looking for it.

    AI Analysis report listing highlights and concerns with a written summary

  • Transaction Browser. The ground truth underneath every report: a full, paginated, filterable list of every transaction, with parent/child rows handled correctly (PayPal records internal accounting entries alongside real payments — this app knows the difference), expandable line-item detail, raw JSON for the truly curious, and a built-in T-code legend for PayPal's cryptic event codes.

    Transaction list with an expanded row showing line-item detail

  • Multi-Account Credentials. Connect more than one PayPal API credential — sandbox and production side by side, each fully scoped so reports, chat, and analysis for one account never leak into another. Each credential can also carry a short business-context note ("I sell handmade candles online"), which AI Analysis uses to make its narrative more relevant instead of generic.

    PayPal credential card with the business-context fields open

  • Demo Mode. Try the whole product instantly against a sample bookstore account — no signup, no PayPal credentials to enter, just click and go.

    Landing page's Try the Demo card

  • Built for mobile, properly. The app started as a desktop-first layout with nested sidebars (main nav, plus a per-page list like conversations or the report menu) — fine on a laptop, cramped on a phone. I recently reworked the whole shell so both sidebars collapse into off-canvas drawers on small screens, with the inner one opening automatically to whatever list you need until you pick something.

    Main navigation drawer open on a narrow mobile viewport

Try It

The live demo is up at TransactionsChatAI — click Try Demo on the landing page to drop straight into a sample bookstore account, no signup required.

Full source in a private repo for now. If you want the technical deep-dives on how specific pieces work — PayPal's Reporting API quirks, the Claude tool-use design, the streaming chat interface, the prebuilt-reports architecture, or the AI Analysis pipeline — those are their own posts.