Web app vs desktop app
Why choose the desktop app?
The free web app is genuinely good: your documents never leave your browser, no account, nothing to install. For a quick redaction it's all you need. When the work gets heavier, the desktop app goes further in three ways.
Everything runs in your browser, nothing is uploaded, and it's free forever. Many people never need more than that.
Sharper detection
The desktop app runs the full detection stack — larger entity-recognition models plus an optional on-device AI layer the browser can't load. It catches more names, rare identifiers, and context-dependent details a lightweight in-browser model can miss.
Faster on real workloads
Native on-device processing instead of a browser sandbox means large PDFs, scanned pages (OCR), and whole batches of files process dramatically faster — and you can handle many documents at once instead of one at a time.
Organize by matter or client
Group documents into folders per matter or customer so the same person always gets the same code across every file in that folder. Keep each client's or case's work separate and consistent — the web app handles only one client or matter at a time.
Web app vs desktop app, side by side
Both run entirely on your own machine and neither uploads a document. The difference is how much work they can take on.
| Capability | Web Starter | Desktop app |
|---|---|---|
| Price | Free forever | 7-day trial, then Basic or Pro |
| Getting started | Nothing to install, no account | Install and sign in |
| Where documents go | Nowhere — stays in the browser | Nowhere — stays on the device |
| Detection stack | Lightweight in-browser model | Full models plus an optional on-device AI layer |
| Scanned pages | Reads them, a page at a time | Reads them far faster, across whole batches |
| Multiple documents | One at a time | Whole batches at once |
| Detection tuning | Not available | Match strength, custom expressions, manual regions |
| Consistent codes across files | One matter at a time | Folders share one code registry |
| Teams | Not available | Seats, audit log and compliance dashboard on Pro |
| Automation | Not available | Local REST API and CLI |
| Runs on | Any modern browser | Windows 10/11 64-bit, macOS 12+ |
When the web app is all you need
A handful of documents, one client or case at a time, and no need to tune what gets detected. It is free forever, there is nothing to install, and nothing you open is uploaded. Most people trying to strip personal data out of a file before pasting it into an AI tool never need more than this.
When the desktop app earns its keep
Volume is the usual trigger: batches instead of single files, scanned pages that need reading, and folders of documents where the same person must resolve to the same code every time. The other trigger is control — deciding how aggressively to match, adding your own expressions, drawing regions the models missed, and automating the whole thing through the CLI or the local API.
See exactly what each plan includes.
Scanned pages are pictures of text, so the usual redaction tricks quietly fail on them. Our guide covers what text recognition changes, and how to redact scanned PDFs and images so nothing is left underneath.
Detection quality is measurable. We benchmarked promptShield against Microsoft Presidio on 14 documents across 7 languages and published the script, the test files and the raw results.