Less guesswork. More traceability.

Trust the rows.
Check the exceptions.

01Import02Inspect03Export

A small CSV review desk. Find broken fields, inspect possible duplicates, and leave with a clear record of what needs attention.

Nothing uploaded. Nothing silently removed.

Bring a CSV. Keep the original.

Header: date, description, amount · ISO dates · dot decimals
Up to 5 files / 1 MiB combined / 5,000 rows. UTF-8 only. Drop files here.

Records

—

Clean

—

Ready for export

Review

—

Possible duplicates

Invalid

—

Excluded from net

Valid net

—

Integer cents. No currency conversion.

The working ledger SYNTHETIC SAMPLE

Description / source	Date	Amount	State

Three clear outputs. Clean rows, a separate exception list, and a complete traceable JSON report.

01 / A count you can reconcile

Every row is clean, review, or invalid — never more than one. The valid net includes clean and review rows. Invalid rows are excluded, never treated as zero.

02 / A flag is not a deletion

Possible duplicates share a date, trimmed case-sensitive description and amount. Every candidate stays in the report; review rows are held out of the clean export.

03 / Your original stays yours

Processing uses only this tab’s memory. CSV text is protected against formula injection; exact original fields remain in JSON. No storage, server, AI or accounting judgment.