Skip to content

CapyExpense · tool no. 5


.

A desktop expense dashboard that reads a spreadsheet you type into yourself. No account, no bank login, no cloud, nothing uploaded.

Updated · by Unfolding Dimensions

Coming soon · Windows & Linux

A year on one screen

£31k

Made-up numbers, real charts. This is the app’s own code, running in the page.

OctNovDecJanFebMarAprMayJunJulAugSep

Five charts, four numbers, and a range picker that actually works.

01 · What it is

A spreadsheet you own, read by something that draws.

You type rows into a workbook the app generates — twelve month sheets, a category list, nothing exotic. The dashboard reads that file and turns a year of it into five charts and four numbers. It never writes over your rows, and there is no import step, because the file was always yours.

01Your file

One workbook a year

The app generates a per-year .xlsx — twelve month sheets, a category list, a metadata sheet. You type into it with Excel, LibreOffice, anything. Reading never writes.

02The read

Five charts, four numbers

A burn curve, category bars, a weekday heatmap, a subscription ribbon, a month grid — plus a range picker. All of it drawn from the rows you typed, nothing inferred.

03The binary

No network code at all

The Tauri capability file grants the app no HTTP permission. Not a promise about what it chooses to send — there is no path for it to send anything.

02 · Why typing it out

The manual part is the feature.

Every mainstream expense app opens by asking for your bank login. Skipping that is usually framed as a limitation. It reads better as the whole design — and the literature is on that side of it.

  1. 01

    Recording your own expenses is one of the few money habits with meta-analytic backing: financial self-control strategies show a medium effect across 29 studies (d = 0.57). Writing an amount down rehearses it into memory in a way an automatic bank feed does not, and that rehearsal is what lowers later spending.

    PLOS ONE 2021 — meta-analysisSoman 2001 — J. Consumer Research

  2. 02

    People underestimate their upcoming spending by roughly half, and the expenses they miss are the discretionary ones. Being made to think category by category raises remembered expenses by 36–60%. The category dropdown is not decoration — unpacking spending into categories is itself the intervention.

    NBER w35430 — retrieval failuresSussman & Alter 2012 — exceptional expenses

  3. 03

    The apps that do connect to your bank route it through an aggregator first. Plaid settled a credential-harvesting class action for $58M; lawmakers asked the FTC to investigate Yodlee for selling transaction data. Nothing here connects to anything, so none of that has a place to happen.

    Reuters — Plaid $58M settlementInvestmentNews — Yodlee FTC letter

  4. 04

    The default categories are not invented. They map to the US Bureau of Labor Statistics' Consumer Expenditure Survey major groups and the UK ONS Family Spending classification, so your own numbers can be read against a published baseline.

    BLS Consumer Expenditure SurveyONS Family Spending in the UK

Thirty-six sources were checked before this tool was built; the ledger lives in the repository under docs/research/capyexpense/.

03 · Questions

The ones worth asking before you run a binary.

Do I have to connect a bank account?

No, and you cannot. There is no bank connection, no aggregator, no import. You type rows into a spreadsheet. That is the entire input.

Where does my data actually live?

In one .xlsx file, in a folder you chose, on your disk. The app reads it and draws. It writes only when you explicitly ask it to, and version history goes to a plainly visible CapyExpense History/ folder next to it — not a hidden database.

How do I know it is not sending anything?

The Tauri capability file that ships inside the binary grants no HTTP permission, so the app has no network access to use. That is checkable in the source rather than promised in a policy — which is the point of building it this way.

Will it overwrite the spreadsheet I have been keeping?

Reading never writes. Before any write the app takes a copy into the history folder first, so the destructive version of any action is already undoable before you take it.

Is there a Mac build?

No. Gatekeeper blocks unsigned apps outright rather than warning, so a Mac build needs a paid Apple developer account to be worth shipping at all. Windows and Linux warn and let you continue, so those are the two.

Do I need Excel?

Any tool that reads and writes .xlsx works — LibreOffice Calc, Numbers, Google Sheets via export. The file is an ordinary workbook with no macros in it, deliberately: Office has blocked macros in downloaded files by default since 2022.

What does it cost?

Nothing. It is Apache-2.0 licensed, like the rest of the suite. There is no account to make, so there is nothing to charge for.

Why is there no AI in it?

Because categorising your own spending is the part that does the work. A model guessing categories for you removes the step the research says is doing the lifting, and it would need somewhere to send your transactions.

04 · Status

Not out yet — this is what it will look like.

Everything above is the real app running on sample data. The Windows and Linux builds are still being packaged, so there is nothing to download today. When they land they will be unsigned — a certificate costs more per year than this tool earns, which is nothing — so Windows will warn you, and every release will ship with a SHA-256 you can check first.

A desktop tool — stored on your machine, never ours. Windows & Linux builds coming soon.