Smart File Organizer
Drop a batch of files. On-device analysis reads each document, suggests a real name (not “scan_0423.pdf”), sorts it into a category, and flags exact and near-duplicates — then hands you renamed copies in a ZIP. Nothing is uploaded.
Smart Organizer
Drag & drop here
or click to browse
Analysis
| Original | Suggested name (editable) | Category | Duplicates |
|---|
Real On-Device Document Intelligence — No Upload, No Account
Every step of this tool is a real algorithm running in your browser. File contents are read with Mozilla's pdf.js, each document is fingerprinted with SHA-256 to catch exact duplicates, word-overlap similarity catches near-duplicates (like a re-export of the same contract), keyword analysis classifies each file as an invoice, receipt, contract, resume, report, letter, or form, and the most prominent first-page text plus any detected date builds a meaningful filename — so scan_0423.pdf becomes invoice-2026-05-14-acme-consulting.pdf.
The suggested names are editable before you export, and the ZIP you download contains untouched copies of your originals — only the names change. Everything runs on your device via open-source browser libraries — your files are never uploaded. You can verify it yourself.
Frequently Asked Questions
It is honest, deterministic document analysis: cryptographic hashing for duplicates, word-overlap similarity for near-duplicates, keyword scoring for classification, and layout heuristics (largest first-page text, detected dates) for naming. No cloud model is involved and nothing leaves your device — which also means it is fast and works offline.
Files with byte-for-byte identical content are flagged as exact duplicates even if their names differ. Documents whose extracted text overlaps more than ~82% are flagged as similar — typically re-exports or minor revisions of the same document.
PDFs with a text layer and plain-text files. Scanned PDFs without text can still be checked for exact duplicates by their hash — run them through our OCR tool first to give them a text layer. Images are included in duplicate checking too.
No. The downloaded ZIP contains exact copies of your original files under the new names. Your originals on disk are never touched.
Treat suggested filenames as a draft, not a filing decision
Local classification can make a messy folder easier to review, but business context still belongs to the person organising the records.
Use a stable naming pattern
A practical pattern is date, counterparty, document type and version: 2026-07-11_Acme_SOW_v2.pdf. Put the most useful sorting field first and avoid characters that Windows, macOS or cloud drives treat differently.
Check extracted clues
Scans, letterheads and repeated boilerplate can mislead automated classification. Confirm the document date, parties and version against the actual content before accepting a name. An amendment should not be filed as the agreement it modifies.
Keep a reversible review step
Work on copies until the naming scheme is proven. Duplicate detection can identify identical or highly similar files, but it does not decide which copy has the correct signatures, annotations or retention status.
Organisation guidance reviewed 11 July 2026 · How guidance is maintained