Aggregators stop at the dashboard.
They show you balances. They don't tell you whether to refinance, when to harvest losses, or what the second-order effects of a job change are on your benefits.
sav.ink
A private financial brain that ingests, reasons, and plans — so the money decisions you keep putting off finally get the attention they deserve.
the problem
You have a dozen accounts, a couple of credit cards, a brokerage, a retirement plan, maybe a side business, a mortgage, recurring subscriptions you've forgotten about, and tax events you'll discover in April. The information exists. The patience to sit down and reason about it does not.
They show you balances. They don't tell you whether to refinance, when to harvest losses, or what the second-order effects of a job change are on your benefits.
A good CFP is excellent. They also cost $3–10k/year and you see them twice. The other 363 days, your finances run on guesswork.
Every personal finance system that requires you to manually categorize transactions has the same failure mode. You stop. Within 60 days. Every time.
how it works
ingest
Connect banks, brokerages, credit cards, retirement, crypto, real estate, and private holdings via Plaid + first-party APIs. Receipts and statements parse via OCR. Manual entry exists for what doesn't have an API.
reason
Ask 'should I pay down the mortgage or max out the 401(k)?' and get a numbered answer with assumptions, sensitivities, and the tradeoffs spelled out. The model has read your full position.
plan
Recurring monthly review: tax-loss harvesting opportunities, subscription audit, cash-flow forecast, retirement trajectory, and 3–5 concrete moves to make this month. With reasoning.
remember
Your goals, your risk tolerance, your tax bracket, your dependents, your timelines — all persisted. You don't re-explain your life every time you open the app.
privacy
Sav.ink runs as much as possible on-device. Financial data is encrypted at rest with a key derived from your device passphrase — we cannot read it. Reasoning that requires a frontier model runs against a privacy-preserving inference path: requests are not logged, not used for training, and not joined to any identity beyond a rotating session token. We do not sell data. We do not run ads. There is no advertising model that wouldn't compromise the product, so we don't have one.
platforms
Web
Private beta
iOS
Q1 2026
Android
Q3 2026
macOS
Evaluating
waitlist
We're letting in a small cohort weekly. Tell us a little about your setup so we can prioritize who'll get the most out of it early.
faq