If your work means renaming hundreds or thousands of documents by hand — legal teams, insurance agents and adjusters, accountants, and every other document-heavy professional — point Mylo Prism at the folder. It reads each file on your own Mac — client and claim files never leave the device — and proposes names like “20260214 - Settlement Agreement - Client v Opposing Party”. You approve every name before a single file changes.
Apple-notarized · Apple Silicon, macOS 14+ · 14-day free trial, no account · one-time price, no subscription · support@myloprime.com
Three steps. The AI does the tedious first draft; you stay in charge of every name.
Choose a folder — a thousand files is fine. Before any AI runs, Prism fingerprints every file and shows you exactly which ones are identical duplicates (they're skipped, never deleted). Then the pages are first converted to text by on-device text recognition (OCR), so even courthouse scans are read without any cloud, and the vision AI running on your Mac reads the actual pages — you watch each page appear on screen as it's read — including a zoomed pass over headers and reference numbers so identifiers like case, claim, or policy numbers are read digit by digit. A crash-safe checkpoint lets big jobs resume where they left off.
Every proposed name sits next to the live document itself, with the evidence: the date, type, parties, and reference numbers the AI found — and a one-line plain-English summary of what each document is. Anything it couldn't read is flagged in orange instead of guessed. Every field is editable, each value shows the exact words on the page it was read from — highlighted on the document, and flagged if the quote can't be found — and each row is labeled with the engine that read it: on-device, or the cloud account you chose.
One click renames the approved files. Every batch is written to a rename log you can export for the file, and any batch can be undone in one click. Name collisions resolve automatically.
Lawyers were first — but the problem isn't legal, it's volume. If your day includes a folder of Scan_0034.pdf and IMG_8841.png that has to end up filed correctly, Prism was built for you. A few of the people it's built for:
Turn a discovery dump into a filing-ready folder — dates, document types, matters, and case numbers read straight off the caption, named the way your firm already files.
Scan_0034 (2).pdf → 20260314 - Motion to Dismiss - Client v Opposing Party.pdfEstimates, photos, correspondence, and claim forms land under one scheme keyed to the claim — no more hunting through IMG_ files to find the right policyholder.
IMG_8841.png → 2026-02-14 - Auto Claim - Policyholder - Damage Photo.pngInvoices, receipts, and statements arrive named by scanner and inbox. Prism reads the vendor, date, and document type and files them the way your books expect.
document(7).docx → 20260131 - Invoice - [Vendor Name].docxReal estate & title, mortgage & lending, medical records and healthcare admin, HR, paralegals — anywhere accurate, consistent filenames matter more than anyone has time for.
Scan_0102.pdf → 2026-02-14 - Signed Disclosure - [File Reference].pdfEvery firm and office already has a naming convention. Prism doesn't impose one — you define your own. Field order, separators, and date format are all yours: pick a preset or build your own template, and a live preview shows the exact result before you commit. Set the same convention on every Mac in your practice and the whole team files documents identically.
Choose a date format (YYYYMMDD, MM-DD-YYYY, 2026-07-02, and more), the separators between fields, and the order fields appear — then reuse it on every batch.
Cloud renaming tools upload every document to someone else's server. When those documents hold client confidences, claim files, medical records, or financial data, that's a confidentiality question on every single file — Rule 1.6 for a lawyer, and the same duty of care for anyone handling regulated or personal information. Mylo Prism's default mode never uploads anything — the reading happens on your own machine.
No AI — local or cloud — reads every scan perfectly, and a wrong date, case number, or claim number on an important file is a real problem. That's why nothing is renamed without your review, uncertain reads are flagged, unreadable numbers are omitted rather than guessed, and every batch is logged and reversible. The AI drafts; you approve.
| Document content | Read and processed locally by an on-device model (via Ollama). Never transmitted in local mode. |
| Network use #1 | One-time download of the local AI model (~7 GB) from Ollama's registry, with your consent at setup. |
| Network use #2 | Update check (app version only — no document data). |
| Network use #3 | One-time download of the on-device speech model (≈466 MB) from myloprism.com the first time you transcribe a recording on macOS 14–25 (macOS 26+ uses Apple's built-in speech engine — no download). No document data is sent. |
| Network use #4 | Only if you opt in: pages you choose are sent to the cloud AI vendor you selected — Anthropic (Claude), OpenAI, or xAI (Grok) — under your own API key and their terms. Never more than one vendor per file, never without an explicit choice. |
| Network use #5 | Only if you sign in with a firm account (an option for Hepworth Legal and Mylo Prime staff, not something a purchased licence uses): your work email address and an opaque per-Mac identifier are sent to Mylo Prime to issue and renew your licence, at sign-in and about once a day after. No document data, no file names. |
| Telemetry | None. No analytics, no crash uploads, no content collection. |
| License check | Offline cryptographic verification — the signature on your licence is checked on your Mac, and a purchased licence needs no account and no network at any point. A firm sign-in re-confirms with Mylo Prime about once a day and keeps working for 3 days offline; even then, the check itself still happens on your Mac. |
| Compliance claims | We don't claim "HIPAA compliant" — that's an organizational matter. We state the architecture and let your compliance team draw conclusions. |
Software that touches client files shouldn't need a monthly fee to keep working. Buy it once — updates within v1 are free. Every purchase starts with the 14-day free trial.
Not in local mode — reading, naming, logging all happen on-device. The only way a document leaves your Mac is if you add your own cloud API key (Claude, OpenAI, or Grok) and explicitly choose that engine — for a whole run or one file at a time. The app labels which engine read every file so the choice is always visible, and documents carrying privilege or confidentiality banners ask for an extra, named confirmation first.
It renders the key pages of the scan — plus a zoomed pass over headers and reference boxes — and first runs Apple's on-device text recognition (OCR) over them, so most legible scans become real text right on your Mac. Pages OCR can't read go to a vision-capable AI model running locally via the free Ollama app, which reads them like a person would: parties and document type from the page, reference numbers (case, claim, or policy) digit by digit, the date from the signature or header block. Setup is two guided clicks; no terminal.
Nothing is renamed without your review. Proposals appear next to the original names, uncertain reads are flagged, every field is editable, and unreadable reference numbers are omitted rather than guessed. Each field also shows the exact words it was read from, found and highlighted on the page — a citation that can't be found is flagged instead of trusted. After renaming, each batch is logged and can be undone in one click. For a stubborn scan you can re-run just that file with your own cloud account (Claude, OpenAI, or Grok).
Instantly — your key appears on the confirmation page the moment payment completes, along with activation instructions. Keep it with your records; activation is offline and there's no account.
PDF (digital and scanned), Word (.docx and .doc), rich text, plain text, Markdown, CSV, HTML, and email (.eml); images (JPEG, PNG, HEIC, TIFF, WebP, GIF, BMP); and — through a built-in preview renderer — Excel, PowerPoint, Keynote, Pages, Numbers, and OpenDocument files. Audio and video recordings (.m4a, .mp3, .wav, .mp4, .mov, and more) are transcribed on your Mac and named from what was said.
Both, and both stay under your control. Point Prism at a scanner-output or intake folder and new files queue into Review automatically — watching never renames anything by itself, and every queued file still waits for your approval. And a naming format can optionally file each approved document into subfolders (for example Client/Matter/Year): the destination is shown per file before you approve, and undo puts every file back where it was.
Yes. Prism keeps the Mac awake during long jobs and notifies you when the batch is ready. Nothing is renamed while you're away — the results wait in Review, and a per-batch digest can be exported as a permanent record of every change.
The App Store forbids license keys and the update mechanism we use, so — like the other Mylo Prime apps — it ships as an Apple-notarized direct download: same Gatekeeper security checks, opened with a normal double-click.