Just do your work.
Files file themselves.
The Mac file organizer that reads your files — and never deletes them
Downloaded PDFs, screenshots, invoices and scans land in a folder and turn into a pile of IMG_4821 and download(3).pdf. Filewise reads what's inside — right on your Mac — renames them to something you'll recognize and files them where they belong. What it can't read, it leaves untouched.
Developer ID signed & notarized — opens without warnings
~6MB, installs in seconds; downloads the local AI model on first launch (~2GB, one time — skipped if Apple Intelligence is on). The model and your files stay on your Mac, never uploaded.
Or via Homebrew: brew install --cask gatheon-apps/tap/filewise
Want to hear about updates? Drop your email:
Old name → new name + the reason, laid out as a table for you to review. Nothing happens until you say so.
It never touches delete. Worst case, a file is just renamed or moved — always recoverable.
On garbled or weak signals it stays quiet and keeps the original name — better unchanged than wrong.
It shows you everything before it acts
- ✓A plan first. Old name → new name + the reason, in one table. Nothing moves until you've seen it and said yes.
- ✓Auto only when confident. It renames silently only when the signal is strong; whenever it's unsure it queues the file for one glance from you — it never guesses on your behalf.
- ✓Keeps the name when unsure. Anything garbled or weak stays as-is and waits for your review — better unchanged than wrong.
They make you sort. Filewise recognizes for you.
The clearest way to choose is to rule out what doesn't fit. Here are the three real alternatives, side by side — no put-downs, just the contrast.
Hazel: you write the rules
You hand-write if-then rules for every file type, and they break the moment a file looks different. Filewise reads the content and decides for itself — you write zero rules.
Finder tags / smart folders: metadata only
They sort by tags you apply by hand, or metadata like date, type and size — they can't read what's inside. Filewise reads the invoice amount, who issued the document, the report's topic, then names it.
Another AI renamer: makes up a name
It will confidently invent a name from garbage — the classic failure of these tools. Filewise keeps the original name when it can't read, never fabricates, never deletes, every step undoable.
A few things it does for you
Reads the content, not just the name
PDFs, Office and plain text read directly; scans and screenshots via Apple Vision OCR (Chinese + English). It reads what's actually written — amounts, who issued the document, topics — instead of guessing from the filename.
An invoice gets named like an invoice
An invoice, a signed contract, a report — each is a scene it can recognize. It names them all as when_who_type_about, so re-reading the filename brings the moment back.
Set it and forget it
It lives in the menu bar, watching where files land — Downloads, Desktop. The moment one arrives it reads it in the background and has a plan ready; you just glance and confirm.
Fully local, learns your way
Reading, naming and learning all happen on your Mac — file contents never touch the cloud. Correct it once and it remembers, fitting your habits more every day.
You do the work. It does the rest.
It lands
A download finishes, you take a screenshot, you drag something in — right where files already show up, Filewise catches it and starts reading in the background.
It understands
Local AI works out the scene: who, what type, what it's about, what to call it, which folder it belongs in.
You nod
The action center puts the next step in front of you — a ready plan when it's confident, and when it's not, it asks only the one thing only you know. One glance to confirm.
Always undoable
Every change goes in an undo log — one click rolls it all back. All it ever does is rename and move, so worst case is always recoverable.
When it earns its keep
Important files, handled safely
Contracts, bills, invoices, scanned IDs, reports — the files you can't afford to lose and that pile up fastest in Downloads. Filewise names them by who, when and what type, files them where they belong, never deletes them, and keeps the contents on your Mac. That's how the product works, not a marketing line.
A Downloads folder that stays clean
"download (3).pdf", "untitled.docx" — junk names get read and renamed the moment they land, and filed where they belong. Downloads stays clean without a weekend cleanup.
Screenshots & scans, back to life
Apple Vision OCR reads the text inside screenshots and scans, turning blind names like Screenshot and IMG_4821 into searchable, meaningful ones — no more opening them one by one to find the right one.
I'm its user zero.
I deal with downloaded PDFs, signed contracts, invoices and screenshots every day. I got fed up with the pile of unrecoverable junk names in Downloads — and I'd never dream of feeding those files to a cloud AI. Filewise is something I built for myself first. Precisely because these files can't touch the cloud and can't be lost, I made "fully local, never deletes, always undoable" a hard rule, not a feature. It's still rough, but the safety floor is built in by design: rename and move only, never delete, every step undoable — and that won't get watered down just because the version is early.
You might be wondering
Will it delete my files?
Never. Filewise only renames and moves. It shows you a plan to preview first, writes an undo log for every step, and one click rolls everything back exactly. Deleting is something it simply never does.
Do my file contents get uploaded to the cloud?
No. Whatever engine it uses, reading, naming and learning all run offline on your Mac — not a single byte of file content leaves your machine, and there's no upload switch. By default it names files with a local Qwen model (set up on first launch); Macs with Apple Intelligence use the system model, and if you have Ollama it uses that — all offline on your Mac.
Does it handle Chinese documents and screenshots?
Yes. It uses Apple Vision OCR (Chinese and English) plus a local model. Structured details like amounts, dates and who issued the file are matched against verifiable signals — only confirmed matches are used in the name, otherwise it keeps the original and queues it for you. It never fabricates when it can't read. Once you have the DMG, try it on your own files with a dry run first; one click reverts if you're not happy.
Can I download it now? Is it free?
Free to download right now (Apple Silicon · macOS 14+ · Developer ID signed & notarized, opens without warnings). No Ollama required — grab the built-in model in-app for offline AI naming. A local tool doesn't make money off your data; it's free during beta. Want updates? Leave your email.
Always improving
Once installed, every update ships through in-app auto-update — no need to come back and download.
- “Connect to Server” is smoother: WebDAV no longer bounces to the browser or crashes, one dialog for address + username + password, plus SFTP; any mounted drive ejects in one click. Naming is more grounded: it checks content harder before naming — it won’t label an invoice with a similar-but-wrong company, and won’t invent a topic that isn’t there
- New “Connect to Server” (⌘K or the sidebar): mount SMB / AFP / NFS / FTP / WebDAV servers — credentials stay in the system Keychain, and you organize the moment it mounts
- Cloud drives, one click away: the sidebar’s “Cloud” section auto-lists whatever you have — iCloud, Dropbox, Google Drive, OneDrive, Nutstore — no setup
- Cloud drives supported: online-only files get a ☁︎ marker and are never silently downloaded when organizing — right-click “Download & name by content” to opt in
- Legacy invoices now recognized: older “增值税电子普通发票” PDFs whose text can’t be decoded fall back to page OCR; list view gained right-click “Copy Path” + ⌘⌥C
- Much sharper invoice recognition across ride-hailing, train, dining, hotel, flight and travel-agency e-invoices — total reads 价税合计, full-width train amounts parse, seller no longer swapped with buyer (verified 86/86 on real invoices)
- New Chinese e-invoice reader: pulls the seller, billed entity, expense type, tax-inclusive total and issue date straight into the filename
- Fixed images, screenshots and scans reading as empty on recent macOS — OCR and naming work again
- Sort by time now mixes files and folders by creation date, newest first — no longer buried under folders
- Faster local-brain setup: the model now downloads from the official mirror
- “Name it with AI” always responds now — a suggested name to accept, or an honest reason it couldn’t
- Sharper local naming (company, doc type, source); clear local-brain status; uses Apple Intelligence on supported Macs — no download
- Watch any folders you choose; skips source & dev files; new pre-read for instant suggestions
- Single / dual / tri / quad panes with drag-between; customizable naming structure; much better Chinese recognition
Just do your work.
Files file themselves.
Free to download right now — or leave your email and we'll tell you about new versions first. Fully local, never deletes, every step undoable.
The same principle — local-first, always findable. Gathering the scattered into a whole.
Snap it, forget it — find it back with one word
Snap a picture, the event lands in your calendar
Hundreds of screenshots, sorted wave by wave
Screenshots you took, found back with one word
Hold to talk, real-time transcription you can use right away