Your money, intelligent and private.
On device AI personal finance for iPhone and Mac. Track accounts, budgets, subscriptions, and goals, with an AI advisor that runs entirely on your device.
The AI
Intelligence that lives on your device
Most finance apps send your data to a cloud LLM. Centmond offers three Gemma models: one tuned for iPhone, two for Mac. You install the app, pick a model, and download it once. After that, every receipt, prediction, and chat happens locally. Private by design, not by promise.
AI Predictions
Centmond forecasts where your spending is heading and shows you how to bend the curve, before you breach a budget, not after.
Drag, swipe, click side, or use arrows · 1 / 4
On device, always
Three Gemma models are available: a compact E2B for iPhone, plus E4B Q5 and Q6 variants for Mac. After installing the app you pick one and download it once. From there the model runs locally and your data never leaves.
Trained for money
Categorization, anomaly detection, duplicate finding, and budget rescue. Tuned specifically for personal finance, not generic chat.
Private by architecture
There is no cloud inference endpoint. Even if we wanted your data, the system literally cannot send it. Privacy is structural, not promised.
Instant answers
On device inference means no round trips. Receipts categorize in milliseconds. Chat responses stream as fast as you can read.
The Models
Three Gemma models, picked for your device
Centmond offers a different Gemma variant for iPhone and Mac, and on Mac you choose the trade off between speed and quality. After you install the app, you pick a model and download it once inside Centmond. From that point on, every prediction, chat, and receipt scan runs locally. Nothing streams to a server.
Gemma E2B
Q4_K_M · GGUF
Everyday questions, receipts, quick categorization.
A compact 2 billion parameter model tuned for iPhone hardware. Loads fast, runs cool, and handles the day to day intelligence (categorization, receipt parsing, simple chat) without ever leaving your device.
Gemma E4B
Q5_K_M · GGUF
The default Mac experience, balanced for most users.
The 4 billion parameter Gemma at 5 bit quantization. Nearly as accurate as the larger Q6_K variant, but noticeably faster on Apple Silicon. The default Mac model, and what most people will want.
Gemma E4B
Q6_K · GGUF
Deep analysis, complex predictions, the sharpest answers.
The same 4B model at 6 bit quantization. Maximum on device quality. Heavier on memory and a touch slower, but the right call for deep financial analysis, intricate scenario planning, and the most precise AI responses.
Models are not bundled with the app. After installing Centmond, you open the AI settings, pick the variant that fits your device, and download it once. From there it lives on your device and the AI works fully offline with zero ongoing data transfer.
Source: huggingface.co · downloaded by you · runs locally
Budgeting
A budget you'll actually stick to
Most budgets fail because they only tell you what already happened. Centmond watches your pace in real time and tells you what to do next, before you overspend.
Budget Overview
Set caps per category, watch real time progress, and get warned the moment you start tracking toward a breach. Not after.
Budgets that match how you actually spend
Pick from suggested category groups or build your own envelopes. Monthly, weekly, or rolling. Centmond adapts to your rhythm.
Alerts before you breach
Centmond projects the rest of the month from your current pace and warns you days before a category goes red. No surprises on the 30th.
AI powered budget rescue
Over budget? The on device AI suggests concrete cuts (which subscriptions to pause, which categories to trim) based on your actual habits.
Roll overs and shared budgets
Carry leftover budget into next month, share envelopes with household members, and track joint spending without merging accounts.
Subscriptions
Find every recurring charge. Cut what you do not use.
Centmond watches your transactions for recurring patterns and surfaces every subscription you are paying for. The AI explains why each one looks recurring, scores its confidence, and waits for your call before adding anything.
Subscriptions
Auto detected recurring charges with confidence scores. Monthly, annual, and next 30 days totals at the top. Active, Trials, Paused, and Cancelled tracked separately.
Auto detection from your transactions
Centmond scans your spending for recurring patterns and proposes subscription candidates. Streaming services, gym memberships, software trials, that newsletter you forgot about: each gets surfaced with the cadence and the next expected charge.
Confidence and reasoning
Every detected subscription is scored. You see how confident the AI is that something is a real recurring charge, and you can review or reject before anything is created. No silent guesses.
Lifecycle and trial alerts
Active, trial, paused, cancelled. Each state is tracked separately so a free trial does not quietly turn into a paid subscription. Get notified before renewals so you never get billed by accident.
Monthly and annual cost truth
Centmond rolls everything up: monthly total, annual total, and what is coming in the next 30 days. The 9.99 a month subscription you stopped using is shown for what it actually is: 120 a year.
Net Worth & Reports
The whole picture, in one place
Beyond budgets and transactions. Centmond tracks your real net worth and turns months of data into reports you'll actually use.
Net Worth
One number that tells you how you're really doing. Assets minus liabilities, tracked over time, broken down by mix.
Drag, swipe, click side, or use arrows · 1 / 2
Net worth at a glance
Watch your trend over 1M, 3M, 6M, 1Y, or All. With debt ratio, total assets, and total liabilities front and center.
Asset and liability mix
See where your wealth actually lives. Checking, savings, investments, and what you owe, side by side.
Heatmaps that show your year
Category performance month by month, with over budget months highlighted. Spot the patterns no spreadsheet can.
Export anywhere
PDF, CSV, Excel. Clean exports ready for tax season, your accountant, or your own archives. Privacy preserved.
By the numbers
Private by design, not by promise.
Inference happens on your device, every time.
Gemma E2B for iPhone. Two E4B options for Mac. You pick and download the right one inside the app.
Categorization, predictions, chat, insights, all local.
Receipts, chat, predictions, anomalies, rescue, and more.
Features
Everything money, nothing extra.
Tap any feature to see how it works in detail.
iPhone & Mac
Built native for both screens
Centmond is a real Mac app and a real iOS app. Not a web wrapper, not a port of one to the other. Every screen designed for the device it runs on.
macOS, top to bottom
Native Mac window chrome, full keyboard navigation, menu bar quick add, multi window, and Mac widgets that sync with the app.
iOS, the way it should feel
Built in SwiftUI from day one. Gestures, haptics, large titles, and Home Screen widgets. No web wrappers, no compromises.
Same model, same data
Your accounts, budgets, and AI memory follow you between devices, encrypted in transit and at rest.
Privacy
Your money is yours.
We made sure of it.
Most finance apps treat your data as a product. Centmond's architecture makes that impossible by design.
Theming
Three modes, all dialed in.
Light, Dark, and a true black OLED mode for nighttime use. Every surface tuned for AA contrast. Click any mode below to apply it across the whole site.
Built with
The best of the platform, honestly chosen.
No proprietary black boxes. Every part of Centmond is a tool we picked because it lets your data stay on your device and your AI run privately.
Try Centmond,
before everyone else.
Join the TestFlight beta on iPhone and Mac. Help shape the launch.
Why we built this
“Every finance app I tried wanted to know everything about my money, and then sell that knowledge back to me, or to advertisers, or to someone else. I wanted the opposite: a tool that understands my money completely, but tells no one. So we built Centmond, an AI that runs entirely on your device, sees everything, and shares nothing.”
Mani
Founder, Centmond
FAQ
Questions, answered
The honest details. If something isn't covered here, write to us and we'll add it.
Centmond also covers more ground in one app: accounts, budgets, subscriptions, recurring detection, goals, household splits, net worth, forecasting, and reports. You do not need three apps to manage one financial life.
This is structural, not policy. Even if Centmond wanted to read your finances, the AI subsystem has no network path to send them. You can verify this by putting your device in airplane mode after the model is installed: the AI keeps working, because it never needed the internet for inference in the first place.
iPhone · Gemma E2B (Q4_K_M) about 2.9 GB. Compact and fast, tuned for iPhone hardware.
Mac · Gemma E4B (Q5_K_M) about 5.5 GB. Recommended for most Mac users. Best balance of speed and quality.
Mac · Gemma E4B (Q6_K) about 7 GB. Highest on device quality. Slower, heavier on memory, best for deep analysis.
Models are sourced from Hugging Face. After download they live on your device and never need to be fetched again.
Adding a Gemma model unlocks the intelligent layer: smart categorization, receipt scanning, AI Chat, AI Predictions, smart insights, anomaly detection, and budget rescue. You can install or remove models at any time without losing your data.
Two things do not sync. First, the AI model itself stays on each device because each variant is hardware specific (E2B for iPhone, E4B for Mac) and far too large to sync efficiently. Second, AI inference is always local: the conversations and reasoning never leave your device.
We will give clear notice before any feature changes price, and TestFlight beta users will be considered for grandfathered pricing as a thank you for helping us shape the launch.
On Mac the AI runs best on Apple Silicon (M1 or later). The Q5_K_M model is the safe default and runs comfortably on 16 GB of RAM. The Q6_K model is sharper but heavier and is most at home on 24 GB and up. On iPhone, the E2B model runs on any device meeting the iOS 17 requirement. Newer chips deliver faster responses.
You will need to download the Gemma model again on the new device, since models live locally on each one and are too large to sync. The flow is identical to first install: open AI Settings, pick the model, download once.
Still have questions? Email mani.scs.gh@gmail.com.