How to actually use it

Install and first run

Everything runs in Docker, so Node is not needed on your machine. Five minutes from nothing to a budget you can click around.

1. Get the source. git clone https://github.com/flaxvin/Pathayam.git, then cd Pathayam. Everything below runs from there.
2. Make a config file. cp .env.example .env. The defaults are fine for a first look.
3. Seed a demo household. docker compose --profile seed up seed — two members, four accounts and a month of realistic behaviour including an overspend, so the screens have something true in them.
4. Start it. docker compose --profile dev up, then open localhost:8080.

Going live

You need a domain, a Google OAuth client for sign-in, and somewhere to run one container. Set BASE_URL to your exact public origin — sign-in redirects and cookie scoping both depend on it being right — then docker compose up -d. Start from an empty budget rather than the seed.

The first person to sign in becomes a member. Everyone else has to be invited from Settings, and the allow-list is checked on every request rather than only at sign-in.

Add your accounts

Accounts come in three kinds, and picking the right one matters more than anything else you will do on this screen.

• Budget accounts — current, savings, cash. Money here funds envelopes. The total across them is what you are allowed to budget.
• Credit accounts — credit cards. Spending creates a debt and reserves the cash to clear it. See the cards guide.
• Tracking accounts — a PPF, an EPF, anything you want on the net-worth figure but never want to budget from.

Give each one its balance as of today. That balance becomes an opening transaction, so the arithmetic starts from a real number rather than zero.

Closing, not deleting

Accounts are closed rather than deleted, always. The history attached to them is real and other figures depend on it. A closed account disappears from the places you pick accounts and stays in the places you read history.

Set up categories

Categories are the envelopes. Groups hold categories. Start with fewer than you think — five or six — and add more when reality demands it rather than in advance.

A useful starting set: Rent, Groceries, Eating out, Transport, Bills, and one honest category called something like Stuff I forget.

Order is information

The arrows beside each category and group move them. This is not decoration: fund every target spends your money down the budget in exactly this order, so whatever sits at the top gets funded first and any shortfall lands at the bottom where you will notice it. Put the things that must happen above the things that would be nice.

Envelopes the app manages

Some categories are created and retired by the app itself — one per credit card for payments, and one per savings goal. They carry no manual controls and are marked managed by the app, so it is obvious why you cannot rename them.

Hiding and deleting

Hide a category you have stopped using and its history stays intact. Delete one only if it never had anything in it.

Build your first budget

The rule is the whole method: you can only budget money you already have. Not money you expect, not next week's salary. What is in your budget accounts right now.

That number appears at the top as Ready to Assign. Your job is to get it to zero — not by spending, but by giving every rupee a job.

1. Fund what will definitely happen. Rent, the loan instalment, the electricity bill.
2. Fund what will probably happen. Groceries, transport, the things you spend on weekly.
3. Fund what you would like to happen. Savings, the trip, the emergency fund. Whatever is left goes here.

Reading the grid

Three columns per category. Assigned is what you gave it this month. Activity is what happened — spending is negative. Available is what is left, and it is the only number that matters when you are standing in a shop.

Assign, move and hold

Three different actions, and confusing them is the most common early mistake.

Assign

Moves money from Ready to Assign into a category. This is how the month gets budgeted. Ready to Assign falls.

Move

Moves money between two categories. Ready to Assign does not change — the total assigned is the same, it is just sitting somewhere else. This is how you cover an overspend, and it is not cheating. Eating out came out of the holiday fund, and now you know that.

Hold

Takes money out of this month's Ready to Assign and places it at the top of next month's. This is how you get a month ahead. See its own guide.

Targets and auto-assign

A target says what a category needs, so the app can tell you whether it has it. Rent needs ₹45,000 a month. An annual insurance premium needs ₹8,000 set aside each month so that in March it is simply there.

Each category shows its progress against the target as a chip — funded, partly funded, or not funded — so a glance down the grid tells you where the month stands.

Fund every target

Once targets are set, one tap does the monthly sit-down. You get a preview of exactly what would be assigned before anything commits. Money is spent down the budget in your order, so a thin month funds the top of the list and leaves the shortfall visible at the bottom.

Record what you spend

Add is in the header on every screen, and a opens it from anywhere. On a household with more than one person the form also asks who spent it, which follows the account — a card in somebody's name is attributed to them — and that name appears as a column in the register and in Query.

Add sits in the middle of the navigation on purpose. Recording a transaction should take about five seconds.

• Amounts do arithmetic. Type 450+120 and it works it out — useful when you are splitting a bill at the table.
• Money out must name a category. “I will sort it later” is how you end up with a queue of hundreds. Money in is exempt, because its job is to sit in Ready to Assign until you give it one.
• Splits handle one receipt covering several envelopes — the supermarket run that was half groceries and half a birthday present.
• Transfers between your own accounts are never spending and are never counted as such.
• Owner records who spent it, which matters on a shared budget and on add-on cards.
• Cleared means the bank has confirmed it. Leave it unticked until you have actually seen it on their record.

Receipts

Attach a photo or a PDF to any transaction. It is stored in the database rather than on a disk somewhere, and served so that no device caches a copy.

When you go over

You will go over. The question a budgeting app has to answer honestly is what happens next, and there are two defensible answers. Choose one in Settings.

• Reduce next month (the default) — the overage comes off next month's Ready to Assign and the category starts fresh at zero. Last month is closed and paid for.
• Carry the negative — the category itself starts next month in the red and has to be dug out. Harsher, and some people want exactly that.

Whichever you pick, today's move is the same: cover it. Move money from another envelope. The app ranks suggestions by where there is most slack, so it takes one tap.

Converting a purchase to EMI

When the bank offers to split a big charge into instalments, open the transaction and say yes there. The converted amount comes off the card's balance and becomes a loan with its own schedule and its own envelope — so you are not asked to clear it on the card and pay it monthly, which is the mistake worth avoiding. The processing fee and its GST are charged to the card like any purchase, so they need an envelope, and the loan screen shows what the plan actually costs in interest and fees.

Two details make the plans livable. Name it — the conversion form offers the payee as a suffix, so you get “HDFC EMI — Croma TV” rather than a column of identically-titled plans you have to open one by one. And whatever the card's payment envelope was already holding against that charge moves across to the new plan's envelope, because the money you set aside to clear the purchase is the money that now pays its instalments. Nothing is stranded and nothing is asked for twice.

On a household commitment it means something else

The envelope you use to commit money to the household can go red too, and it is not an overspend: it means more of the household's spending was paid out of your money than you put aside for it. The button says Put it down to me rather than Cover, because covering it forgives nobody anything — it decides your share this month was larger. Squaring up a month covers the other two ways it can end.

Overspending on a card is different

Cash overspending means you spent money you had not budgeted. Card overspending means you created a debt with nothing set aside to clear it, which is why the card shows an unfunded amount until you deal with it.

Close the month

Ten minutes, once a month. Nothing is locked and a closed month can be reopened, so this is a ritual rather than a gate.

It shows you four things:

• What the month did — earned, spent, saved, the savings rate, and the biggest categories.
• How your net worth moved, split into money saved, market movement, currency movement and debt repaid. Four different reasons a number changed, separated.
• A dated snapshot, so the history is something recorded rather than recalculated later from whatever the prices are now.
• Whether next month is funded.

Should you keep it separate?

Most households pool everything. If that is you, skip this part entirely — there is nothing to switch on, no second grid, and the app never mentions any of it.

This is for the other arrangement: we keep our own accounts and split the shared bills. Couples who came to it late, one salary and one irregular income, somebody supporting a parent out of their own money, or simply a preference. It is a legitimate way to run a household and most budgeting apps cannot do it at all.

What you get is a budget of your own — your accounts, your envelopes, your Ready to Assign — plus an agreed amount you put toward the shared bills each month. What you give up is nothing: the household budget works exactly as it did.

What the other person can see

How much you committed, and how much of it has gone out. That is all. Not your balances, not your accounts, not your other envelopes — not through any total, and not by subtracting one number from another.

Open a budget of your own

On the budget screen, + My own budget. It appears immediately with a few envelopes to start from and one more you did not create: the household envelope, which is how you put money toward the shared bills.

Then move your accounts into it. Open an account, Edit, and set Which budget this account funds to your own. That is the setting that changes the figures: the balance leaves the household's Ready to Assign and joins yours, and spending from the account lands in your envelopes.

Making an account private

Once an account is in your own budget you can set Who can see it to private, and nobody else sees the account or its balance at all.

Switching between them

The sidebar gains a Looking at section. Every screen that shows one budget's money — the grid, Overview, Cards, Schedules, Categories — follows it. Reports and Query are different: they ask whose money as a filter, because "what did we spend on groceries" and "what did I spend" are different questions.

Commit to the household

In your own budget there is an envelope called → Household. Assign to it and that money is committed: it shows up in the household's Ready to Assign for the two of you to assign to rent, groceries, whatever you have agreed.

No money moves. It is still in your account. That is the whole point, and it is what lets a private account fund shared spending without publishing its balance.

Set it once

On The household's money, set a monthly plan — ₹40,000 a month, say. It is an ordinary target, so the budget screen measures you against it and auto-assign fills it. Agreeing the figure once beats remembering it every month.

Paying for shared things

Pay the rent from your own account and file it to the household's Rent envelope. Two things happen at once, and no transfer is involved:

• The household's Rent envelope falls, because the rent is paid.
• Your → Household envelope falls by the same amount, because that is the commitment being spent.

When you pick an envelope from another budget, the Add screen says so under the picker. It is the ordinary way to pay for shared things, not an error.

Shared cards and add-on cards

• A shared card is one whose account sits in the household budget. Anybody may spend on it. The bill is paid from the card's own payment envelope, in the household's budget, however the spending was filed — so a shared card's payment is never split between you.
• An add-on card — your card on somebody else's account — keeps the debt with the primary holder, because that is who the bank chases. Filing an add-on charge to your own envelope means you have used their credit, so the balance between you moves by that much.
• Filing your money to the other person's envelope, with nothing shared between you, is refused. It would invent a balance neither of you arranged.

One receipt, two budgets

Split it. Half to the household's Groceries, half to your own Personal. Each line lands in its own budget and draws on the right money.

Squaring up a month

Some months one of you puts in more. The app does not call that a debt — you are not a creditor and your partner is not in default on the electricity. It says the commitment is underfunded: more has gone to the household than was put aside for it.

On The household's money, the row reads left to right and adds up: brought forward, plus what went in, less what went to the household, is where it stands. There are three ways to settle it, and they are genuinely different.

• Put it down to me. Fund it from your own Ready to Assign. Your share this month was simply larger. Nothing is forgiven and nobody else is asked for anything.
• I'll pick it up. The other of you commits the amount on top of what they already had, so it is funded from their budget instead.
• Call it even. Nobody funds it. This is the only one that lets something go.

Closing the month

Each budget closes on its own. The household's month can close while yours is still open, and the other way round — one person's procrastination should not block the other's ritual. The household's close reports what each of you put in and how much went out.

When somebody leaves

Removing a member never deletes anything. Every transaction they entered keeps their name, the history stays exactly as it was, and adding the same email address back restores them — they pick up where they left off. Settings lists anybody removed with an Add back button, so it is one click either way.

What is not automatic is the money. If they kept a budget of their own there is usually a balance outstanding, and the app will not remove them until you say how it ends. It offers the options and never picks one.

• They have money set aside and unspent. It is theirs — they promised it, and the promise ends with the arrangement. Giving it back is the only ending offered, because recording a debt here would invent one.
• The household has had the use of their money. Either record it as money owed — which becomes ordinary family lending, with them as the counterparty, because they are outside the household now and that is exactly what family lending is for — or they call it even and let it go.

Import statements and CSVs

Typing everything in by hand is how budgeting apps get abandoned. Three ways in, and all three land in the same queue rather than posting themselves.

Statement PDFs

Drop the file on the Import screen. Eleven banks are recognised from their own layouts — HDFC, ICICI, Axis, SBI, Union, Canara, YES, IndusInd, Kotak, RBL and HSBC — plus broker contract notes.

Most Indian statements are password-protected, and each bank has its own rule. The screen shows you that bank's hint. If you save your statement identity in Settings, the app derives the password itself and you never type one.

CSV

Paste it in. Headers are detected, Indian number formats are understood, and once you map a bank's columns the mapping is saved and offered next time.

Undoing an import

Any import can be reversed in one step from the import log, taking every row it created with it. Worth knowing before your first attempt, because it makes experimenting safe.

The review queue

Nothing posts itself. Everything imported waits here with its raw text beside it, so you can always see exactly what the bank said rather than only what the app made of it.

• Uncategorised spending — file it. If you have filed that payee before, the category is offered as a single tap rather than a dropdown to hunt through.
• Possible duplicates — the same purchase arriving twice, once from an alert and once from the statement. Five separate tiers of matching decide how confident the app is; you merge or keep both.
• Overspent categories and unfunded cards — the two things most worth acting on, so they are here rather than only on the budget screen.
• Proposed rules — see the next guide.
• Money you are owed — anything marked reimbursable stays listed until it comes back.

Rules and automatic filing

File the same payee twice and the app proposes a rule, stating exactly what it inferred and from which transactions. It never applies one on its own.

Confirm a rule and you are offered the chance to apply it to your existing history, with a preview of every transaction that would change before anything does.

Writing one yourself

Rules match on the payee, the narration, the amount, the account, or a combination, and can set a category, rename the payee, add tags or flag for review. There is a tester: paste in a narration and see what would fire.

Why it learns from filing rather than importing

Filing counts wherever it happens — from the review queue, from the transaction screen, from anywhere. That matters more than it sounds: an app that only learned from imports would watch you categorise the same delivery service a hundred and thirty times and never once offer to help.

Bank alerts by email

Optional, read-only, and off until you turn it on. Once connected, a bank alert becomes a queued row within seconds of arriving, which is as close to live as a budget gets without a bank integration.

• It reads only the banks' own sending addresses, never the rest of your mail.
• Access is read-only. Nothing is sent, labelled, or deleted.
• You connect your own Google project, so the credentials and the consent are yours rather than a third party's.
• Disconnecting revokes the token and stops the fetching immediately.

Statement emails too

When a bank sends the monthly statement as an attachment, it is fetched, decrypted using your saved identity, parsed and queued — the whole chain, without you opening the mail.

Credit cards

Set the statement day and the due day when you add the card, or from Edit afterwards. They do different jobs: the due day projects the payment onto the cashflow calendar, and the statement day tags every charge on the register with the cycle it bills in — a cycle runs from the day after one statement up to and including the next, so a charge on the 19th with a statement day of 18 bills the following month. Both clamp to the last day in a month too short to have theirs.

This is where most budgeting apps get vague, so it is worth two minutes. Spending on a card is still spending.

Buy groceries on a card and two things happen at once: the Groceries envelope goes down by the amount, and the same amount moves into that card's payment envelope. The cash to clear the bill is reserved the moment the charge lands, so the statement is never a surprise.

The Cards screen

Every credit account in the order it falls due, with a summary strip across the top: total owed, total not funded, and how many fall due within a week.

1. Look at what is unfunded. That is balance with no envelope behind it — usually an overspend, or a charge nobody has filed yet.
2. Fund it from another envelope, using the same cover flow as an overspend.
3. Record the payment as a transfer from a bank account to the card. The payment envelope empties, the balance clears, and no category is involved — the categories were charged when you spent.

Add-on cards

A second card on someone else's account shares one limit, one statement and one payment. Add it as a card on that account rather than as an account of its own, and every transaction still records which card it was and whose.

Loans

Add a loan with its principal, rate and tenure. The one thing to get right is the interest model.

Instalments

Record each instalment with the lender's own principal and interest split rather than the app's projection. Where the two disagree, that is drift, and it is surfaced rather than silently absorbed — the app projects, the lender reconciles, and when they differ you want to know.

Pay it from the account the lender actually debits. A bank EMI leaves a bank account. A card EMI is a charge on the card, not a transfer out of savings — so recording one against a credit account posts a categorised charge the way the statement will show it, and the card balance rises by the instalment before the loan balance falls by the principal. Anything else would have the money leaving twice.

Its envelope funds itself

Every loan gets a payment envelope, and its monthly target is set to the EMI the moment the loan exists — you do not have to remember to set one, and a rate change or a fresh disbursement moves the target with it. The EMI is the least negotiable number in your month; it should not be the one you forgot to budget for.

Settling early

When the loan is paid off, the page says so and offers to close it. You can also settle before term: enter what you are paying and the foreclosure charge the lender wants for the privilege, along with the account it comes out of and the envelope it spends from. Most lenders charge to be repaid early — a percentage on a personal loan, a flat fee on a card plan — and a prepayment decision taken without it is taken against a saving larger than the one actually on offer.

Before you prepay

Use the calculator. It puts reduce the tenure against reduce the EMI side by side and tells you what each saves. Reducing tenure almost always saves more — on a worked ₹50 lakh example it is several lakh more — and the app shows the figure rather than asking you to take that on faith.

Pick the column, and that is what happens. Choosing to reduce the tenure holds your instalment where it is and shortens the loan; choosing to reduce the EMI keeps the closure date and lowers the monthly bill. Say which envelope the lump sum comes out of, too — a prepayment is the largest single thing most households will ever pay, and it has to leave a named envelope like everything else rather than being conjured.

When the rate resets

A repo-linked loan resets several times a year, and your lender has to offer you two ways to take it: keep the instalment and let the tenure move, or keep the tenure and let the instalment move. Enter the new rate, press work out what it does, and the app prices both against your actual balance — the new instalment, and how many months you would add or save. Then choose one, and it is applied: the loan's envelope is re-targeted either way, so next month's budget asks for the right figure without you rewriting it.

Whose loan it is

A loan belongs to a budget like everything else, and a personal one can be marked private — then only its holder sees it, on the list, in Everything you owe, and in net worth. A household loan is visible to everyone in the household, which is the point of it being household. The owner is shown on the list, not just buried in the detail page, so you can tell at a glance whose debt you are looking at.

Disbursements

A loan paid out in tranches records each one with its destination. A builder-paid tranche raises the liability without ever touching your budget; a personal loan landing in your savings arrives as money you can assign.

Lending to family

Money lent to or borrowed from a person rather than an institution, tracked without pretending it is a bank loan.

• Advances and repayments are recorded as they happen, and the balance is derived rather than typed.
• The other side is a name, not a member of your household — which is almost always the case.
• Write-off exists, because sometimes that is what happened, and an app that cannot record it makes you lie to it.

Outstanding amounts appear in the review queue so they do not quietly become gifts.

Overview and runway

The screen to open when you want the month in one glance rather than in detail.

• Runway — how many days of typical spending your money covers, measured against what you actually spend rather than a guess.
• Due soon — bills and card payments approaching, with whether they are funded.
• The month so far — in against out, and where it went.
• Insights — a category well above its own three-month average, a subscription about to renew, a spare month sitting unassigned.

Reports

Four reports, all rendered on the server as plain SVG. No charting library, and nothing about your figures leaves the machine to be drawn.

• Income against spending by month, with a net-saved line, so you can see whether the gap is widening or you just had one good month.
• Category trends over time — the report that catches groceries drifting up ₹500 a month for a year.
• Loan interest by financial year, April to March, which is the figure you need at tax time.
• Realised gains by financial year, split by how long each parcel was held. It reports the holding period and names no tax category, because which threshold applies depends on the asset and that year's rules — a judgement that belongs to whoever files.

Query and export

If anybody keeps a separate budget, Query and Reports both ask whose money, defaulting to everything. "What did we spend on groceries", "what did I spend" and "what did all of it come to" are three different questions, so the app offers all three rather than picking one. A transaction that crosses budgets — your money paying for something shared — counts under both, because which one somebody means depends on what they asked.

One filterable, groupable table that every other screen drills into. Filter by date, category, account, payee or tag; group by category, payee, month or account.

Search

Search runs across both the cleaned payee name and the raw imported narration. When you half-remember a transaction as “something with ZEPTO in it”, the raw string is what finds it.

Getting everything out

The full export is one file in an open format containing every account, transaction, assignment, lot, loan and event. Deliberately excluded, and asserted absent by tests: your statement identity and any connected mail token.

Schedules and cashflow

Recurring things — rent, EMIs, subscriptions, a transfer on the 25th — recorded once and then expected.

• Detection. The app notices something arriving monthly and offers to make it a schedule rather than waiting for you to think of it.
• The calendar projects your balance forward, so the question “will I make it to the 30th?” has an answer rather than a feeling.
• Dips below your floor are flagged in advance, which is the entire point of projecting at all.
• Subscriptions are listed with annualised cost, because ₹149 a month is ₹1,788 a year and the second number is the one that changes minds.
• Money coming in is a schedule too. A salary on the 1st, rent from a tenant, a quarterly dividend — put them in and the calendar stops asking you to survive to the 30th on the balance you have today.

Every outgoing schedule names an envelope

You cannot save a payment that spends from nowhere. A schedule is a promise about money leaving, and a promise with no envelope behind it is the thing zero-based budgeting exists to prevent — so the app asks which envelope before it will take it. Incoming schedules have no envelope, because money arriving is assigned when it arrives, not before.

Arrived, paid, skipped

When a scheduled thing happens, mark it. Paid and Arrived post the transaction for you — the right amount, the right account, the right envelope — and roll the schedule to its next date. Skip rolls it forward without posting anything, for the month the gym was closed or the tenant paid late. The projection redraws immediately either way.

Changing one

Schedules can be edited and deleted. The rent went up, the subscription moved to annual, the standing instruction was cancelled — open it, change it, and the calendar reflects it the moment you save. Deleting a schedule leaves every transaction it already posted alone; it only stops the app expecting more.

Savings goals

Each goal gets its own envelope, created and kept in step by the app, so its progress measures money put aside for that goal and nothing else. If anybody keeps a budget of their own, you also choose at creation whether the goal is shared or yours — and that cannot be changed afterwards, because a goal is measured by its envelope's balance and moving it between budgets would change what the months you have been watching meant.

Long-horizon saving kept deliberately off the monthly grid, so a three-year goal does not clutter the screen you look at every week.

Each goal has a target amount and optionally a date. The app works out what that means per month, shows progress as a ring, and tells you whether you are on track or what you would have to raise it to.

Portfolio

Holdings are tracked as units with FIFO lots, never as a rupee balance. That is what makes cost basis, the realised-versus-unrealised split and a real return figure possible at all.

• XIRR is the headline return, not a naive percentage — it accounts for money going in at different times, which is how anyone actually invests.
• Every price carries its date, so a stale figure is visible as stale rather than passing as current.
• Selling shows a preview naming the exact lots it will consume before you confirm.
• Splits and bonus issues are recorded as events, so the unit count changes without the history becoming wrong.

Getting holdings in

The monthly consolidated statement is the primary route: it already covers every fund folio and demat holding, and rows reconcile against lots you already have rather than duplicating them. Prices for funds and currencies refresh on their own.

Hand-valued assets

Gold, a retirement account, property, anything that does not arrive in a consolidated statement. These need you to say what they are worth, and the app makes that one sitting rather than a hunt.

The valuations screen lists every hand-valued pot at once with what it was last worth and when. Fill in the ones you know, leave the rest blank, save.

Anything not valued recently is flagged, so a gold holding last priced eight months ago cannot quietly hold up your net worth.

Allocation

Where your money actually sits, by asset class, by geography and by currency — three donut charts, drawn on the server.

Instruments are classified once and remembered. The point of the screen is the gap between what you believe your allocation is and what it is, which is usually larger than expected and always more equity than people think.

Net worth

The figure, with the date it is as of, and a history you can see the shape of.

The change over any period is split four ways, which is the part that makes it useful rather than merely satisfying:

• Money saved — what you actually put aside
• Market movement — what prices did
• Currency movement — what the rupee did, kept separate from the asset's own gain
• Debt repaid — the quiet one, and often the biggest

Activity and undo

Every change ever made, newest first, with the undo beside it. Thirty days.

The log is append-only and the undo is itself an entry — nothing is edited away or deleted, so the record of what happened stays true even after you reverse it.

• Where a later change touched the same record, the undo tells you what it would discard before it does anything.
• Where the record is load-bearing — a loan instalment, a portfolio lot, a reconciliation — it refuses outright and says what depends on it.
• Imports undo as a batch, taking every row they created.

Explain this number

Ready to Assign and every category balance carry an explain link listing the events that produced them, in order. When a figure is not what you expected, this is the fastest route to why.

Backups and health

Your data is one SQLite file. Back that up and you have backed up everything.

The Health screen

The one to open at 2am: backup status, last verified restore, price-feed health and remaining quota, feature flags, error counts, and whether the development bypass is present. Every check shows its state, when it last ran, and a plain-language reason.

Restoring

Use the restore command rather than copying a file over the database by hand. Copying leaves the database's write-ahead journal behind, and the first query then reports the database as malformed — turning a recoverable afternoon into a lost ledger at exactly the wrong moment. The command lists what is available, keeps what it replaced, and reads the restored copy back before claiming success.

Settings, members and tokens

Everything that configures the household rather than the month.

• Members — the allow-list. Invite by email address; they sign in with that Google account. Removing someone keeps their historical attributions.
• Overspend model — the choice described in the overspend guide. Changing it re-derives your history.
• Statement identity — name, PAN, date of birth, mobile, used to derive statement passwords. Never exported, never logged, and asserted absent from exports by tests.
• Notifications — each kind of digest item can be muted individually, per member.
• Theme — light, dark, or follow the system, stored server-side so the first paint is already correct.
• Devices — every signed-in session, revocable one at a time.
• API tokens — scoped read or read-write, shown once, and unable to sign in or mint more tokens. See the API reference.
• Feature flags — turn off loans, assets or multi-currency entirely and they vanish from navigation and reports rather than sitting greyed out.

Viewing as another member

Read-only by default, always logged, and shown as a banner on every screen while it is active. Writes require an explicit toggle within the session.

Groups and tidying the grid

A group is the heading envelopes sit under. On Categories you can rename one in place, reorder it, and delete it once it is empty.

Deleting needs it empty, and that is deliberate. Deleting the envelopes with it would destroy balances and history at one click; leaving them behind would put money in envelopes no screen renders. Move them out first — it is one change per envelope, and each is a decision worth making.

The ones the app keeps

Credit card payments, loan payments and Goals are marked managed by the app. They hold envelopes the app creates and retires alongside the card, loan or goal they belong to, so they carry no manual controls and cannot be deleted by hand — they go when the last of those does.

A budget of your own starts with a few

Opening a personal budget gives you two groups and four envelopes to start from, plus the household one. Rename or delete any of them; they are ordinary.

When something looks wrong

A number is not what I expected

Use the explain link beside it. It lists the events that produced the figure, in order, which resolves nearly all of these in under a minute.

A past month reads lower than it did at the time

Known, and written down rather than hidden. Opening July in December subtracts assignments made since. The budget is a live document rather than a series of monthly photographs, and this is the price of that.

My card shows an unfunded amount I do not understand

Something was spent on the card without an envelope behind it. Usually a charge still sitting unfiled in the review queue, or a genuine overspend. Filing it or covering it clears the warning.

An import brought in the wrong things

Undo the batch from the import log; every row it created goes with it. Then check the column mapping and try again.

The same purchase is in there twice

The review queue's duplicate tier should have caught it. If one was entered by hand and the other imported, merge them from the queue — the merged row keeps the raw imported text.

Prices are stale

Check Health for the feed's status and remaining quota. Free providers have daily limits, and the app deliberately stops asking rather than burning through them.

I am locked out

Sign-in depends on BASE_URL exactly matching your public origin, because the OAuth redirect and the cookie scope are both derived from it. A mismatch after moving domains is the usual cause.