GitHub Repository

Offline desktop finance tracker for macOS with local AI categorization (Tauri + React + SQLite + llama.cpp)

2 starsTypeScript

Ledgr – Offline finance tracker with local LLM categorization

Name: Ledgr – Offline finance tracker with local LLM categorization
Availability: InStock
Author: humanji

by humanji·Feb 20, 2026·1 point·0 comments

Visit Project View on HN

AI Analysis

●●SolidNiche GemShip It

Chase CSV → local SQLite → llama.cpp categorization, no Plaid, shows reasoning per transaction.

Strengths

•Zero network calls and explainability (confidence scores per categorization rule) rare in finance tools
•Learned rules improve over manual corrections, reducing friction vs. manual recategorization
•Tauri + Rust + React stack modern and performant; MIT licensed, fully open

Weaknesses

•Chase CSV only (no Plaid, other banks) severely limits adoption vs. YNAB/Monarch Money
•macOS ARM only for AI; Intel Macs lack categorization—Windows/Linux completely unsupported

Post Description

I built Ledgr because I was tired of giving Plaid access to my bank account and paying $99/year for the privilege.

It’s a desktop app for macOS (Apple Silicon only for now). You import Chase CSV files, everything goes into local SQLite, and transactions get categorized automatically. There are 50+ built-in rules, learned rules that improve as you correct them, and an optional local LLM via llama.cpp. You bring your own GGUF model. No API calls, no cloud, no telemetry.

Stack: Tauri 2.0 (Rust + React), SQLite, llama.cpp bindings. One thing I haven’t seen in other tools: every transaction shows why it was categorized (rule name, confidence score, or “manual”). You can set budgets, filter/search, and export to CSV.

Free and MIT licensed. Limited right now: only Chase CSV, only macOS. I wanted to ship and see if anyone else cares about this approach.

GitHub: https://github.com/Humanji7/ledgr Download (macOS ARM): https://github.com/Humanji7/ledgr/releases/latest