I'm Not a Developer. I Built 10 Apps With Claude AI.

The Problem

I run BNN Wealth, an asset and wealth management business. I trade derivatives every day. And every evening, I was manually analyzing order books, creating Excel sheets, reconciling P&L across 3 brokers — Zerodha, Dhan, Upstox. It took hours.

So I decided to build my own tools. Zero coding knowledge. Just Claude and a problem to solve.

1. Order Analyzer

Every evening after markets close, I download my order book CSV. Earlier, I'd spend 45 minutes manually matching buy-sell orders.

Now? I type /orders in my terminal. Done.

A Python script reads raw broker CSVs, groups orders into trade cycles (bought at 9:16, sold at 9:27, bought again at 9:51 — it separates these automatically), calculates weighted average prices, capital deployed, and cumulative P&L.

Output: a professional 2-sheet Excel workbook. Summary with hyperlinks on Sheet 1, complete order flow on Sheet 2. Green for profits, red for losses.

We then built 4 more supporting scripts — a daily workflow orchestrator, an HTML dashboard generator, a client dashboard combiner, and an interactive CLI updater. The entire end-of-day workflow runs with one command.

2. Client Performance Dashboard

When you manage someone's money, transparency isn't optional. I was sending Excel sheets over email. Not professional enough.

So I built a 13 MB self-contained HTML file. No server needed. Just open it.

Three tabs:

• Client Profile — Capital journey timeline, all account IDs documented
• Trading Performance — Day-by-day interactive carousel with every trade, 4 interactive charts (P&L trend, ROI%, win rate, cumulative P&L)
• Performance Comparisons — Broker-wise breakdown, gross profit minus charges = net profit, withdrawal tracking with invoice links, and returns vs Nifty vs Gold vs FD

Client agreement PDF embedded. Invoices embedded. BNN Wealth logo branded. Glassmorphism design.

My client thought I hired a design team.

3. Personal Trading Dashboard

Opens every morning when I start my Mac:

• Live Indian indices (Nifty, BankNifty, Sensex, Nifty Mid Select) via Dhan API
• Global markets (BTC, Gold, Silver, Dow Futures)
• Trading calendar with NSE/BSE holidays and remaining expiries
• Europe & US market timers with flash alerts
• Sector heatmap with live stock data
• Multiple modes including a dedicated "Trading Mode"
• 2000+ rotating motivational quotes

4. Trading Risk Manager — A Full Desktop App

I told Claude — "I'm bored with HTML pages. Build me a proper application."

We built a real desktop app using Tauri (Rust) + React + Python FastAPI. Installable. With its own icon.

7 tabs: Dashboard with three-tier circuit breakers (daily loss, weekly loss, notional cap). Live positions with real-time Greeks (Black-Scholes). Position sizing calculator that adjusts for expiry day gamma risk. Trade log that auto-captures every trade from Zerodha via Kite API. A scaling roadmap from Rs. 5 lakhs to Rs. 1 crore. Live margin tracking with 90-day history.

And when I book a profit? Confetti animation with celebration sound.

Full-stack app. Python backend. React frontend. Rust wrapper. Live API integration. Built in 2 days.

5. Trading Companion — 5 Years of Self-Analysis

This project changed how I see myself as a trader.

813 trading days of raw data (2018–2023) sitting in Zerodha CSVs. Never analyzed properly. Claude built a complete data pipeline — raw CSVs → 813 daily Excel files → JSON analysis → 5 interactive HTML reports.

The Fall (2019–2022): 697 days. Win rate 52.1%. Average 15.5 trades/day. Overtrading was the biggest killer.

The Recovery (2022–2023): 116 days. Win rate 69.2%. Trades cut to 8.2/day. Daily P&L jumped 12.8x. Same trader. Same market. Just discipline.

Claude identified patterns I couldn't see — averaging down, revenge trading, every green month followed by catastrophic losses from overconfidence.

11 Python scripts. A complete trading psychology audit — built by someone who doesn't know Python.

6. Bank Statement Analyzer

Multiple bank accounts (ICICI, Kotak, Saraswat), no idea where my money was going. And when audit season hits, sorting through months of transactions is a nightmare.

Built a Flask web app: upload any statement (CSV, Excel, PDF) → auto-detects the bank → auto-categorizes every transaction using 130+ keyword rules across 35 categories. Recognizes Swiggy, Zomato, Amazon, Uber, Netflix, IRCTC, and dozens more.

From raw bank statements to audit-ready categorized data. No more scrambling during tax filing or GST submissions.

My Claude Journey

Free → Pro ($20/mo) → Max ($100/mo). All within 3 weeks.

Why I upgraded each time:

• Free: Curiosity
• Pro: Hit rate limits on Day 2
• Max: Hit rate limits again building the Order Analyzer

The numbers:

• 23,747 messages sent
• Longest session: 32 hours straight (1,108 messages — that was the client dashboard)
• Peak day: 4,300 messages across 6 parallel sessions
• 1,100+ total conversations

From "let me try this free AI tool" to "this is my extended brain" in under 2 months.

What I Learned

You don't need to be a developer to build software. You need a problem and the willingness to iterate. I've probably said "still not good, give me something better" to Claude about 200 times. And every single time, it delivered.

If you're a non-technical founder, a trader, a business owner — stop thinking you need to hire developers for every small tool. Start building. The barrier to entry has never been lower.