ACCOUNTS AGGREGATION & INTELLIGENCE
Every bank, credit card and cash account statement merged, classified and de-duplicated — to track spending, and to calculate tax liabilities.
Bank statement
Indian Bank
••••4821 · 412 rows
Credit card
ICICI Mastercard
••••7390 · 286 rows
Cash book
Offline entries
cash-book.xlsx · 114 rows
Duplicates
0 double-counted
matched on content
Merge all your account statements — bank, credit card, offline accounts — into a single classified statement, de-duplicated by content so that any overlapping uploads do not double up.
Business, Personal, Transfer or Cash in one column and a sub-head like Rent, Fuel or Travel in the other. You can edit any expense to classify it as personal or business.
Salary, Interest or Dividend, grouped the way a return is. Computed from your categories, so totals don’t drift.
Correct an entry once and every past and future statement follows it.
EMIs, SIPs, rent and salary detected with cadence and next expected date.
React Native, same engine as the web app, so corrections carry across devices.
The classification pass
The last row reads “GST”, but it is the bank’s own fee — not a tax remittance. Telling those apart, across every payee, is most of the work.
Who it serves
The platform is organised around who you log in for. Each person or business gets its own books, accounts and learned rules, with complete confidentiality.
Professionals
Salary, professional income, rent, interest and dividends; and as many bank and credit card statements as it takes, reconciled.
CA practices
Each client’s books under your single login, with their separate accounts, classification rules and output.
NRI filers
Indian accounts, Indian rent and interest income, and a filing deadline in a timezone you left behind.
Platform
Your bank and credit card statements. PDFs are decoded in the browser, so the file never sits on a server. Excel date entries are handled, along with header detection, day-first dates and split Dr/Cr columns.
Every row carries a head — Business, Personal or Transfer — and a sub-head such as Rent or Fuel, so the two columns stay linked but independent.
Correct an entry once and it stays corrected, across every statement already loaded and each that arrives later. Edit an entry and the narration gibberish is gone for good.
EMIs, SIPs, rent and subscriptions outbound; salary, rent received and interest inbound. Each with cadence, monthly value and the date it is next expected.
The output ledger grouped as a return: Salary, Business & Profession, Other Sources. Derived from your categories, so the totals don’t drift.
An ATM withdrawal is money moved, not money spent. It becomes cash in hand to account for. Expenses can be uploaded from an Excel sheet.
How it works
There is no chart of accounts to build and no mapping to configure before the first upload. The system arrives opinionated and you correct it where it is wrong — once per payee, not once per row.
The cycle
One file per account, any format the bank offers. Duplicate rows are identified by content, so re-uploading an overlapping period never doubles anything.
Payees extracted from the narration, heads and sub-heads assigned, recurring commitments detected. No setup step in between.
Anything left on a default is filtered into a single list. Set it, choose to remember, and that payee is settled for every statement you upload from then on.
Read the summaries on screen or export the full classified ledger to Excel. Next period, upload the new statement and start at step one.
Built to assist the Indian ITR
Classify by what a transaction was for, and the filing view follows. Because the tax grouping is computed from your categories, your totals tally with your ledger.
Transfers — money moved between your own accounts, loan receipts, redemptions, ATM cash — are held out of income and expense entirely. It is money changing pockets, and counting it is how spreadsheets end up reporting twice the income that was actually earned.
Also included
Security & data handling
Accounts are isolated at the database, and the sensitive parts of a statement are encrypted before they are ever written down.
Narrations, payees, account labels, balances and your own notes are encrypted with AES-256-GCM under a key held outside the database.
Row-level security enforces that only you can read your records. You can enrol an authenticator app and every sign-in will then require MFA; idle sessions sign out after five minutes.
Download everything held about you as a single file at any time, or permanently delete all of it — statements, rules and uploads — in one action.
AI features are optional and idle unless invoked. When invoked, only the transactions needed for that query are sent to Anthropic’s API, and they are not used to train models. Parsing, categorisation, recurring detection and the ITR lens are ordinary code that runs with no AI involvement.
Mobile In development
Android and iOS apps are being built in React Native against the same classification engine as the web app — so a payee you correct on your laptop is already correct on your phone.
No store links yet — and no ratings to quote until there are. The web app is finished and works in a phone browser today.
The unclassified list, in your hand.
Statements arrive by email — deal with them there.
Pricing
One fee for the year’s filing, not a subscription that outlives the deadline.
Free ₹0 no card, no expiry |
Retail ₹2,000/year approx. US$24 for the year |
Pro ₹10,000/year approx. US$120 for the year — ten clients included |
Business Custom quoted per engagement — talk to sales |
|
| Sets of books included | 1 | 1 | 10 | Agreed with you |
|---|---|---|---|---|
| Statements held at a time | 1 — delete to swap | Unlimited | Unlimited | Unlimited |
| Bank accounts per set of books | 1 | 10 | 10/entity | Agreed with you |
| CSV, Excel and PDF ingest | ✓ | ✓ | ✓ | ✓ |
| Head / sub-head classification | ✓ | ✓ | ✓ | ✓ |
| AI Assistant (auto-categorize & insights) | — | 50 runs/month | 500 runs/month | 500 runs/month |
| Recurring and receipts detection | ✓ | ✓ | ✓ | ✓ |
| ITR reporting lens | ✓ | ✓ | ✓ | ✓ |
| Cash book and manual entries | ✓ | ✓ | ✓ | ✓ |
| Excel export, data export and delete | ✓ | ✓ | ✓ | ✓ |
| Authenticator-app 2FA | ✓ | ✓ | ✓ | ✓ |
| Per-client rule isolation | Single set of books | Single set of books | ✓ | ✓ |
| Business FY audit workspace | — | — | — | ✓ |
| Add more sets of books | Upgrade first | Separate account | Nothing further to buy | Part of the quote |
| Start free | Get started | Start with your clients | Talk to sales |
USD figures are indicative for reference only; billing is in Indian rupees. Start on the free plan with no card — it holds ten statements at a time, enough for ten bank accounts in one set of books, with classification, reporting and export unlocked — the AI Assistant is the one feature the paid plans add. Upgrading is a one-off payment by card, UPI or net banking, and your plan switches as soon as it succeeds; if you would rather be invoiced, request the plan from inside the app and pay by transfer instead. Extra set of accounts are bought or requested the same way.
Business is for companies auditing a full financial year rather than filing one return: a separate double-entry workspace over as many entities and years as the engagement covers. Import the day book you already keep, and it becomes a trial balance that foots, a profit and loss account, a balance sheet and one workbook your auditor can open. It is scoped and quoted with you rather than bought from this page — tell us what you need and we will come back with a quote.
Questions
Any Indian bank export carrying a date, a narration and an amount — which is all of them. Header names, day-first dates, split Dr/Cr columns, single amount-plus-type columns and Excel serial dates are all recognised, and the parser was built against real HDFC, ICICI and Indian Bank statements rather than idealised samples.
If a layout does not parse, send it to us and it becomes a supported layout.
Not yet, and we would rather say so plainly. LekhaLens is India-first in two ways that matter: dates are read day-first, so a month-first US statement would be misread, and the reporting lens maps to ITR heads rather than anything the IRS uses.
What US-based customers use it for today is their Indian accounts — NRI and resident accounts, Indian rent and interest income, capital gains — all of which still have to be filed in India. US bank ingest and a US-side reporting view are on the roadmap, not in the product.
Because the work is not the same size twice. Retail and Pro cover filing a return from a year’s statements, which is a known job at a known price. A business audit is not: how many entities, how many financial years, which ledgers have to be reconciled and what the auditor needs at the end all change the answer. So we scope it with you and quote it, rather than printing a number that would be wrong for most people who read it.
The workspace itself is a separate module, and it works today: you import the day book you already keep in Excel, and it comes back as a trial balance that foots, a profit and loss account, a balance sheet, and one workbook your auditor can open, sort and tie back. Rows your sheet never balanced are held in a suspense account rather than guessed at, so what is still unreconciled is a number you can see rather than an error hidden in a total. Debtor and creditor ageing, cost-centre reporting and the tamper-evident edit log are the next things being built. Tell us what your year looks like and we will tell you plainly whether we can do it now.
No, and there is nowhere to enter one. You export a statement from your bank yourself and upload the file. LekhaLens never connects to your bank, so it cannot move money and cannot see anything you did not hand it.
Most rows land correctly on the first pass, and the remainder is front-loaded: the payees you transact with repeat, so fixing one settles every past and future row for that payee. Anything still sitting on a default is collected into a single filtered list, so you are never hunting through 800 rows for the twelve that need attention.
No. AI features are optional and do nothing unless you invoke them. When you do, only the transactions needed for that answer go to Anthropic’s API, and they are not used for training. Everything else — parsing, categorisation, recurring detection, the ITR lens — is ordinary code that runs with no AI at all.
They are in development, in React Native for both Android and iOS, and there is no store link to give you yet. The web app is finished and works in a phone browser today, so nothing is waiting on them. Sign in and we will let you know.
Both, from the Profile tab, without asking anyone. Export gives you a single file containing everything held about you. Delete permanently removes your transactions, uploads, books and learned rules; your login remains, so you can start again from empty.
Reviews
It saved me a lot of time filling out my ITR for FY 2025-2026.
Thanks to Lekhalens.
A few lines about what it saved you — or where it fell short — is what the next person reads before they try it. No account needed.
Write a reviewNothing to configure first. Export a statement from your bank, drop it in, and read.