PDF Insight reads, classifies and merges tax slips on your own machine. By default, nothing is uploaded, nothing trains a model, and no provider receives the contents of your files.
Most document tools upload files to their servers to process them. PDF Insight does the opposite: the AI runs on your computer, so the document never has to travel to be read. A cloud speed lane exists, but it is optional and off by default.
Default
The AI model and OCR run on your machine. Documents are read, classified and merged on-device. Nothing leaves your computer, and everything works with no internet connection.
Optional, off by default
For near-instant processing, you can turn on an optional, clearly-labelled paid lane. Only then, and only if you enable it yourself, do documents leave your machine.
No jargon. Here is exactly what happens to a file you hand PDF Insight in the local tier.
Every page is read and classified on your own computer. OCR for scanned documents also runs locally, via Tesseract.
In the local tier there is no server and no cloud copy of your files. Documents are never transmitted anywhere.
Your documents are never used to train or improve any model, because they do not leave your machine.
The local model runs through Ollama, on your machine. No model provider receives your files in the local tier.
Files stay in your own folders. You decide what is kept, archived or deleted, whenever you want.
Your files live in your own file system, protected by your operating system's full-disk encryption (FileVault on macOS, BitLocker on Windows) when it is enabled.
Accountants, bookkeepers and preparers are accountable for their clients' tax information: SINs, income, addresses. Local-first reduces the risk at the source instead of trying to insure it after the fact.
No copy on a third-party server means no third-party copy to breach. Your exposure is limited to the one machine you already secure.
Data stays where your computer is. You do not have to guess which country a file is hosted in or track it across borders.
When a client asks where their documents are and who can reach them, the answer fits in one sentence: on your machine, and only you.
Local-first processing supports several principles at the heart of Quebec's Law 25 and PIPEDA: minimizing the collection and transfer of personal information, limiting who can access it, and keeping data in a jurisdiction you control. Because your files are not uploaded in the local tier, there is no transfer to a third party to document or to govern by contract for this step of your work.
We put this carefully: software alone does not make you "compliant." Compliance depends on your policies, your engagements, and how you use all of your tools together. PDF Insight is built to help you meet your obligations on this specific step, not to stand in for them.
Not legal or compliance advice. Obligations vary by province, client and engagement. Confirm specifics against your firm's policies, Law 25 and PIPEDA guidance, and legal counsel where needed.
Privacy is earned by being honest. Here are the things we do not assert.
Straight answers for firms evaluating a private, local document tool.
No, not by default. In the local tier, PDF Insight reads, classifies and merges files entirely on your own computer; they are never uploaded. The paid cloud speed lane (Cerebras) is optional and off by default; only if you turn it on do documents leave your machine.
No. In the local tier the model runs on your machine via Ollama, and your documents are never used to train or improve any model, because they never leave your computer.
Compliance is your firm's responsibility, not a box software ticks for you. That said, local-first processing directly supports several principles in Law 25 and PIPEDA: minimizing transfers of personal information, limiting access, and keeping data in a jurisdiction you control. This is not legal advice.
No, and we do not claim to be. Those certifications mostly attest to how a vendor protects data on its servers. In the local tier your documents are not on our servers, so there is no third-party copy to certify or to breach.
No. No automatic classifier is perfect. PDF Insight proposes an order that you review and adjust before exporting; you keep the final say. We do not promise perfect accuracy.
The local tier works fully offline. The licence check is offline-tolerant with a grace window, so the tool does not break on an air-gapped machine.
No credit card. Runs entirely on your computer, nothing to upload. macOS, Windows and Linux.
Needs Ollama (free) for the local AI — setup link in the app. Want to understand the workflow first? See how it works.