Dear Community,
I'm sure everyone here has worked with AI at some point.
For portfolio analysis, equity research or simply as a sparring partner.
But how many of you have already implemented a complete project with Claude Code, Codex & Co.
I've been a bit quieter here in recent months. Instead of posting on GetQuin, I built my own portfolio tool with Claude Code in my spare time, completely prompt-driven, without any programming knowledge to speak of.
🚨 Short disclaimer first:
The whole thing was primarily a self-experiment. No GetQuin competitor, no advertising. The project has since been discontinued (more on this later). What follows is purely an experience report and perhaps a motivation for one or the other to try something new themselves.
━━━━━━━━━━━━━━━━━━━━━━
🔧 The project
Goal: A portfolio tracking & management tool with the features that I personally find interesting.
My framework conditions:
- Approx. one month project duration, a few hours per day
- I am not a developer, only rudimentary dev skills
- Everything should be prompt-controlled, minimal manual effort
The tech stack (Next.js, GitHub, automatic deployment) was built entirely through prompts in Claude Code.
From idea to implementation: prompt, test locally, commit, deploy.
━━━━━━━━━━━━━━━━━━━━━━
🎯 What was created in approx. 60 hours
Core features:
- Functioning web app (desktop & mobile)
- User management with signup & magic link/password login
- Multilingualism, light & dark mode
- Onboarding with demo mode, dashboard tour and tooltips
- Live stock market data with worldwide ticker support
Portfolio management:
- Complete transaction history
- CSV import from Trade Republic, Scalable Capital and Comdirect
- Automatic & manual sector grouping
- Performance vs. market benchmark
- Dividend calendar
- "What if I sell?" incl. tax calculation for individual stocks
- Social sentiment
Monte Carlo simulation:
- Various withdrawal strategies (fixed, SWR, dynamic, etc.) & rebalancing
- Savings & retirement phase with tax analysis
- Historical return/volatility analysis per sector (manually customizable)
- Simulation of target weightings
It's easy to get lost in the possibilities...
Here are a few screenshots from the tool.
Shown are demo data, so please don't be surprised if some numbers don't make sense.
━━━━━━━━━━━━━━━━━━━━━━
💡 My learnings
The planning phase is essential. Debugging and testing make up about 80% of the development time, less the initial build.
The CSV imports in particular (every broker has a slightly different format...) were a headache. I can imagine how time-consuming it is to maintain dozens of broker formats and make improvements when changes are made.
Maximizing the efficient use of API data through smart caching is essential.
Claude has built me an "architecture flow" for this, where you can see exactly where and when which API calls occur, including information on cache efficiency, latency, etc.
I didn't plan everything from the outset, but implemented new ideas on the fly.
When it comes to system-wide interventions in increasingly complex projects, this can sometimes lead to endless loops for Claude, which can often only be solved if you have an idea of what the problem might be.
━━━━━━━━━━━━━━━━━━━━━━
🤦♂️ The great disillusionment: stock market data
And here comes the catch.
Live stock market data is expensive.
For private use, it is still reasonably affordable, but even then the price is significantly higher than a GetQuin premium membership.
If you want to use global stock market data commercially, you have to pay at least €2,500/month per month.
That's without the growing infrastructure costs for thousands of active users.
Even for purely private use, I prefer to invest the money.
━━━━━━━━━━━━━━━━━━━━━━
👉 My conclusion
It's amazing what you can get up and running in such a short time with just a little know-how. Especially when your own platform feels like a professional product in the end.
The project is now on hold and I'm using GetQuin again.
The €9.99/month for Premium is more than justified. Anyone who knows the infrastructure and data costs behind such a tool knows why.
I don't see the 60 hours invested as wasted time, but as an enormous gain in experience.
If you have been afraid of implementing a web project yourself due to a lack of technical knowledge: Give it a try. The barrier to entry has never been so low.
