The money app I built for myself

In my previous note I mentioned in passing that I’d built my first ever web app. Today I want to actually show it — along with a few updates I’ve made since.

Why I built it

None of the apps stuck

Managing money is a routine I’ve never been good at. Over several years I tried seven or ten different apps, and something was always off, so I never stuck with any of them. Sometimes it was the lack of automation — too much manual input. Sometimes the architecture didn’t fit the way I actually live. Sometimes the way the data was presented just didn’t make sense to me.

A spreadsheet that worked — for the company

A while ago I started using Google Sheets for my company’s financial planning, and that worked well. I could clearly see how much I was going to earn and spend over the coming months. So I tried the same thing for my personal finances — and failed again.

What works for company planning doesn’t work for personal. For the company I know quite precisely when, how much, and on what I’ll spend. The variable is income: I don’t know whether in a few months there’ll be one more project or a few more subscribers. So I always plan from confirmed income. With personal finances it’s the other way around — the income is clear every month, and the spending is what varies. On top of that I wanted categories. Once I’d added budgets, categories and everything else, the spreadsheet couldn’t keep up, and after three weeks I stopped entering my spending again — because it was all manual.

An AI agent and an Excel file

My next attempt was to have Claude, and later ChatGPT, create and then edit an Excel file locally on my machine. I built a workflow where I’d just tell the agent what I’d spent that day and it would add it to the file.

That still had problems. Some were technical — the Excel file would break somehow. The bigger one was representation: I couldn’t see my money. All I got were text replies from an agent. Not exactly convenient.

All of that added up to one feeling: I wanted a separate app for managing my money, built exactly the way I want it.

The app

So here it is. All the data in these screenshots was generated by AI for the demo — none of it has anything to do with my real money.

The main summary

The first thing I see when I open the app is the summary:

• how much money I have right now
• my income this month
• how much I’ve spent
• and what my profit for the month is going to be — whether I’m ending the month with more money than I started with, or less

Below that are two graphs.

The first one shows what I spend on each day of the month, broken down by category. Hovering over a day gives me the exact amount for each category.

The second shows my cumulative spending against my plan. There are two lines: the dashed one is how much I’m supposed to have spent by each day of the month, and the solid one is what I’ve actually spent. When the dot is red, I’m over plan — and the numbers above tell me exactly by how much.

Plan and actual for this month

Scrolling down, I see a table with this month’s budget: a breakdown by category, and for each one the plan, the actual, and how much is left. I can see that this month I planned to spend ¥18,000 on having fun and so far I’ve spent less than ¥2,000. Still ¥16,000 left to go out on!

This view makes it very easy to see how I’m doing against my plan in the current month.

Monthly history and forecast

Scrolling deeper, there’s a month-by-month history and forecast of what I spend, my profit, and my balance. This one is convenient in a different way: I can look into the future and see how much will be in my pocket. From there I can start thinking about adjusting my spending or my income to meet my financial goals.

A bit about budgets

The real problem with personal finances — the one I mentioned at the beginning — is that spending varies a lot. We’re human beings, we buy things spontaneously, and that’s fine. So we need a mechanism that helps us plan ahead anyway. In my app, that’s the budgets page.

Here I see my year as columns of months and rows of categories. Each cell is how much I planned to spend in one month in one category, and the months that have already passed also show what I actually spent.

So I can see that my restaurants plan was ¥40,000 from January through July, and that I spent more than that almost every single month. That’s a clear signal: either I rethink how I eat out, or I simply adjust the budget so it fits my actual behaviour. These budgets are set for each month separately, by the way. So if I decide to raise the restaurants budget in August, the earlier months keep their history and go on showing me that I was overspending there.

Transactions

One more page is called transactions, and it shows exactly how much I spend, when, and on what. Line by line — or, actually, through this very functional year calendar too. GitHub is what inspired me to build it.

The calendar shows how much I spent on each day of a year. Green is actual spending — the past. Blue is planned spending. The bigger that day’s transactions add up to, the darker the colour gets, and it’s all calculated from each month’s budget.

Now if I click on a single day, it works as a filter for the list below and shows me only the transactions that happened on that day.

The filter works with date ranges. I can set by hand which day I want to see my spending from and until. I can also click a month name or a week number on the calendar. And I can even just drag-select a range of days straight across the calendar cells.

Finally, I can also use the filter to show all the planned transactions for a particular range of dates.

Transaction vs operation

This one is big. It’s one of the reasons I couldn’t stick with the other apps I used.

Most money apps keep things simple: money comes into your account — plus to your income; money goes out — plus to your spending. But that’s not how life works. Sometimes I pay not only for myself but for other people together with me, and then get money transferred back.

Imagine going to a restaurant with friends, paying the whole bill with your card, and getting money back from each of them for what they owe you for the dinner. It happens all the time.

With all of those apps, if I first pay ¥12,000 for the whole bill, that’s immediately ¥12,000 added to my spending. Then they give me back ¥8,000 — plus to my income. Let that happen a few times and the app shows my spending and my income as far higher than they actually are.

In my app I did it right.

One transaction holds the information about what I spent money on and when it happened. The money spent is calculated from the sum of the operations attached to that transaction. So in this case it’s −¥12,600 from the bank card, +¥4,200 in cash from one friend, and +¥4,200 from another. In total the transaction shows ¥4,200 of spending — and my statistics don’t lie to me with fake monthly income or spending.

The secret sauce

The best part of this app is MCP. One of the reasons I gave up on money apps was the need for manual input, and that would have happened here as well — with this correct, but slightly complex, transaction-and-operation structure.

So my answer to it was connecting an AI agent to the app, so that the only thing I need to do is tell it how much I spent and on what. Or send a screenshot of my banking app’s history. Or set up an automation that reads my email and picks the bank notifications out of it. Plenty of ways to automate it. MCP can be used with Claude, ChatGPT, or any other agent. So convenient.

Final words

My app is still very young. I’m going to keep using it, adding features, doing fixes. Maybe someday I’ll open it up for other people to use. But for now it’s a closed version, for me only.

Stay tuned for more updates — and a possible release of an open version. Bye!