Three steps to get Spendif.ai running on your computer: download, install, first launch. All offline, no account required.
sudo pacman -Syusudo pacman -U ./spendifai-*.pkg.tar.zstLooking for a specific version? See all releases · Source code: spendifai/spendif-ai.
1. Double-click the downloaded .dmg.
2. Drag SpendifAi.app into the Applications folder.
3. Eject the DMG (Cmd+E) and launch the app from Launchpad.
The DMG is signed with a Developer ID certificate and notarised by Apple, so macOS opens it with no warnings.
Alternatively, with Homebrew:
brew tap spendifai/spendifai
brew trust --cask spendifai/spendifai/spendifai
brew install --cask spendifai
From then on you update with brew update && brew upgrade --cask spendifai.
assets/screenshots/macos-dmg-dragdrop.png
1. Double-click the downloaded .msix.
2. The built-in App Installer opens. Click Install.
3. The app appears in the Start Menu: launch from there.
The first few times, Windows may show "Windows protected your PC": click "More info", then "Run anyway". The warning fades as the certificate builds reputation. Details in the guide.
assets/screenshots/windows-msix-install.png
From the terminal:
sudo apt install ./spendifai_*.deb (Debian/Ubuntu)
sudo dnf install ./spendifai-*.rpm (Fedora/RHEL)
The postinst only installs uv. The venv, the model download and .env happen at first launch, so the first start takes a few minutes. Full guide: installation docs.
assets/screenshots/linux-deb-install.png
On first launch Spendif.ai sets up its environment: it's the longest step, but happens only once.
Double-clicking opens a native window with a splash screen. Spendif.ai detects your RAM/VRAM and kicks off a background download of the LLM best suited for your hardware (Qwen 2.5, Gemma 3: sizes from 1 to 8 GB).
The model lives at ~/.spendifai/models/ (macOS/Linux) or %USERPROFILE%\.spendifai\models\ (Windows) and survives reinstalls. The configuration wizard appears within seconds, without waiting for the download to finish.
assets/screenshots/splash-download-progress.png
While the model continues downloading in the background, the configuration wizard guides you through 4 steps:
1. Language: taxonomy, date format, number separators.
2. Holders: your name and the variants your bank uses.
3. Accounts: current account, cards, cash.
4. Summary + Start: everything is written only here, no partial persistence.
Every step shows a live download banner at the top (e.g. "📚 78% · ~3 min"). On the final summary, the Start button stays disabled until the model is 100%: so the Import page works immediately when you reach the app. Everything is changeable from Settings later.
assets/screenshots/onboarding-step-1-language.png
On confirm the wizard creates the SQLite database and lands you on the Import page. From here you load your first bank CSV/XLSX files.
From then on the flow is: double-click → splash (1-2 seconds) → app. No more downloads, no more wizard.
assets/screenshots/app-ready-import-page.png
Nothing stays on your computer against your will. The command depends on how you installed:
Not sure how you installed? If you used the Terminal it was Homebrew, otherwise it was the DMG.
macOS, Homebrew: brew uninstall --cask spendifai
macOS, DMG: drag SpendifAi.app to the Trash
Debian / Ubuntu: sudo apt remove spendifai
Fedora / RHEL: sudo dnf remove spendifai
Windows: Settings > Apps > Spendif.ai > Uninstall
Your data stays, on purpose. ~/.spendifai (%USERPROFILE%\.spendifai on Windows) holds your transaction database, your settings and the AI models, several GB in all. None of the commands above touch it.
Export what you want to keep before deleting it, from the Ledger page: once removed it cannot be recovered.
To remove the data too: with Homebrew, brew uninstall --zap --cask spendifai. Otherwise use the uninstaller for your system (macOS, Linux, Windows), which asks separately about the database and the models, or delete the folder yourself.
Full documentation: install guide · user guide · macOS deep-dive · Windows deep-dive.
Bugs, questions, requests: open a GitHub issue.
Prefer email? Write to support@spendif.ai.
Spendif.ai is in alpha. The features we add next are decided together with the people using the app today: no speculative work, no roadmap dictated from above.
Features under evaluation based on alpha-tester feedback:
Would one of these help you? Tell us in GitHub Discussions. When enough requests stack up on the same feature, it moves to the top of the queue.