Everything a document manager should do — and the AI to actually do it.
Capture, understanding, filing, finding, and asking. Every capability below runs entirely on your Mac.
Capture
Get documents in any way you like — dragged, imported, or scanned straight from the Inbox.
Drag & drop anything
Drop files or folders onto ShabuBox — any file type, multi-select, even onto a note.
⌘O importer
Bulk-import from anywhere on disk with a standard file picker.
Direct ScanSnap USB
Native support for the Fujitsu ScanSnap iX1300 — press Scan and go.
Network scanners
Any eSCL / AirScan scanner via Apple ImageCaptureCore, plus a universal scan-to-folder watch fallback.
Scan options
PDF or JPEG, duplex, color / gray / B&W, 150 / 300 / 600 dpi.
The Inbox
Everything new lands in one place to review, confirm, or file in a batch.
Understanding
On-device models read each page and figure out what it is and what it contains.
Apple Vision OCR
The same tech as Live Text reads every page. PDFs use embedded text with a Vision fallback for scans.
Document-type ID
Invoices, statements, tax forms, EOBs, contracts — it knows the difference.
Entity extraction
People, organizations, places, and postal addresses via GLiNER2 — mergeable with aliases.
Auto-naming
Files are renamed from their contents at ingest, with dates and currency amounts extracted.
Filing
A classifier that learns from where you file — and knows when to stay out of the way.
Learning auto-filer
Learns your system and files new arrivals the same way — no rules to write, no uploads.
Confidence bands
Auto-files when sure, suggests when close, and abstains when it isn’t. You stay in control.
Tags & smart collections
Tag freely and build smart collections that update themselves from simple criteria.
Folder Health
A running view of what needs your attention — misfiled, unfiled, or duplicated items.
Finding
Find anything by any means — a keystroke, a word on a page, or a date.
⌘K search everything
One palette across files, scanned text, notes, entities, collections, and reminders.
Full-text index
Every OCR’d word of every page is searchable — not just filenames.
Semantic search
Optional, experimental local embeddings (~67 MB) fused with full-text. Off by default.
Calendar view
See your documents on a timeline by their extracted dates.
Asking
Ask a collection a question and get a grounded answer you can verify.
Cited answers
Every claim carries an [n] citation to the exact source passage.
Three-tier generation
Apple Intelligence first, a local open MLX model as backup, extractive passages as a last resort — it never dead-ends.
RAM-gated models
The picker only shows models your Mac can run — from ~1.7B up to ~27B parameters.
One optional download
A richer local chat model (Qwen3-4B, ~2.3 GB) is a one-time, optional download.
Beyond documents
A place for the writing and follow-ups that live alongside your paperwork.
Notes with wiki-links
Markdown notes that link to each other and to your documents — a lightweight personal wiki.
Reminders
Set follow-ups that sync with Apple Reminders, so nothing slips.
Your data
Real files in real folders. The app is an index over your documents, not a cage around them.
Real folders, real files
Your documents stay as ordinary files in ordinary folders. The app never destructively changes them.
Rebuildable index
The database is just an index — it can be rebuilt from your files at any time.
Optional iCloud sync
Turn on sync to use your own iCloud Drive. The developer runs no servers.
Maintenance tools
Rebuild index, regenerate thumbnails, rescan, and repair the library when you want to.
See it work on your own paperwork.
Free for 30 days, full-featured. Then $24.99 to own it.