They already have your data
Tax invoices, withholding slips, and bank movements reach Coretax before your return does. Your report becomes a comparison, not new information.
Your numbers are already with the tax office. What they check is no longer whether your report is complete, but whether it agrees with the data they already hold.
Tax invoices, withholding slips, and bank movements reach Coretax before your return does. Your report becomes a comparison, not new information.
A gap used to surface only during a field audit. Now it takes one comparison on a desk.
A gap found eleven months later means hunting for documents nobody remembers, in files that have moved warehouse.
You need books that know, every day, how far they differ from the tax office’s data — and can explain every rupiah of that difference with a document and a statute. That is what KLOPT does.
Every company really has two versions of profit: the accounting one and the tax one. They genuinely differ, and should. The only question is how that difference is managed.
One correctly recorded sales invoice immediately produces the journal entry, the e-invoice line, the tax matching line, the receivables ageing, and the audit working paper. Without a single figure retyped.
Pick the customer, enter the items, set the date. VAT status and payment terms already sit on the customer record, so your staff are never asked a tax question.
Debits and credits are assembled by the system from the transaction itself, guaranteed to balance, and posted to both the accounting and tax reports.
This month’s e-invoice line, output VAT, any withholding exposure, and whether you under- or over-paid this month. All calculated, none typed.
The same four checks a tax auditor uses run automatically. Each gap is matched to an explanation, and anything left over is clearly flagged.
When you prepare the annual return, or when a query letter arrives, the file is already complete. Every figure traces back to its original document.
The same four checks a tax auditor uses run automatically at each month end. Gaps are matched against explanations in your own data. Whatever is left is written down plainly as unexplained.
Rp6,500,000 has no explanation yet. The system does not guess and does not force the numbers to agree. It raises the question now, eleven months before someone else does.
Two marks run through the whole app. Ink for figures a person has approved. Pencil for anything from the AI: grey, dashed, and always erasable.
AI output never goes straight into the ledger. It appears as a suggestion awaiting approval, and if rejected leaves no trace at all.
A statutory reference is not a sentence the AI wrote; it is pulled from the rule list in the system. So the AI cannot cite an article that does not exist.
This ban lives in the code structure, not just in an internal policy. If anyone calls the AI from a tax calculation, the build fails.
“The Rp6,500,000 input-VAT gap looks like a pattern in three vendor invoices whose invoice date falls in a different month from their booking date. Likely: BILL-2026-0341, BILL-2026-0352, BILL-2026-0377.”
A rule guarded only by application code will eventually be broken. So KLOPT’s bookkeeping rules are guarded by the database. Wrong code is rejected, not forgiven.
Security & dataThere is no delete or edit. Corrections go through a reversing entry linked to the original.
If one figure is changed mid-year, the lock breaks and it shows on the next check.
Not by application code. One wrong line cannot leak another company’s data.
Including backup and disaster-recovery locations.
Ordinary accounting software is good at recording. But there is one question it cannot answer: why does the revenue on your annual return differ from the total tax base on your invoices for the year?
There, matching those figures stays a March spreadsheet job. In KLOPT it runs by itself every month, from the same data.
No. The AI is forbidden from touching tax calculations, and that ban is built into the code structure. If someone tries, the build fails.
The AI does what machines are good at: reading PDF invoices, suggesting accounts, flagging odd transactions. The result is always a suggestion. You can reject it, and rejecting it leaves nothing in the ledger.
You are, as always. That is exactly why it is built this way: every tax figure comes from a rule list with effective dates. A 15 March transaction uses the rule in force on 15 March, not today’s rule.
Every judgement is labelled honestly: certain, conditional, or grey. A system that claims everything is certain loses your trust the moment one case is wrong.
One company with a year of history usually takes two to three weeks. Week one: import customers, vendors, items, and opening balances. Week two: run alongside the old books. Then switch.
Do not switch mid-period. The cleanest point is the start of a month. If that date is only days away, wait for the next one.
In Indonesia. Separation between companies is enforced by the database, not by application code. A single mistyped line cannot leak another company’s data, because the database itself refuses.
The full explanation is on the security page.
Another question? Send it to us — answered by a person, not a bot.
Forty minutes with our team. We load sample data from your industry, run the last four months of checks, then show which gap is most likely to be questioned first.
No credit card. No commitment. We prepare the sample data.