Privacy-first AI. Simple. Effective.

AI for your documents —
without sending them
anywhere.

Can't use ChatGPT with your client files? We get it. Here's the fix.

A ready-to-run bundle of open-source tools. Download, install, done. No cloud. No subscriptions. No drama.

Get PrivateDoc AI — CAD$89 → Buy Basic — CAD$19 →
PrivateDoc AI — CAD$89
Full product. RAG system, entire document library, all models, separate workspace per client or project.
Basic Version — CAD$19
One model, one file at a time. Try the interface, test response speed, see if it fits your workflow.
Who it's for

If your documents can't go to the cloud — this is for you.

⚖️

Lawyer

Use AI to work through contracts and commercial documents — without your client data ever leaving your computer.

🏢

Business Owner

Put AI to work on your business — strategies, deals, internal docs. Your intellectual property stays yours.

🏥

Doctor / Clinic

Process patient records and test results with AI — personal health data stays on your machine, full stop.

📊

Economist

Analyze financial documents and data with AI — complete confidentiality, no exceptions.

Works for anyone who handles sensitive information — not just the four above.

How it works

Three steps. Then you're done.

1

Download the installer

One file. Works on any modern Windows or Mac.

2

Run the script

Everything installs automatically in 30–40 minutes. No technical background needed.

3

Work with your documents — offline

Upload PDFs, Word files, spreadsheets, text files. Ask questions. Get answers. No internet required.

Pricing

Two versions. One clear upgrade path.

Start with the basics, or go straight to the full RAG system. Either way, your data stays on your machine.

Full version
PrivateDoc AI
CAD$89
One-time payment. No subscription.
  • Installer script (Windows now / Mac coming soon)
  • Full RAG system — ask questions across your entire document library
  • Access to all models — current and future additions included
  • Separate workspace for each client or project — files never mix
  • Load your documents and start working immediately
  • Quick-start guide
  • Complete user guide
  • 30 days email support — reply within 24 hours
Get PrivateDoc AI — CAD$89

Start here

CAD$19

Launch price — limited time

Installs permanently on your machine. Works offline. No subscription. No expiry.

One model of your choice from the current catalog. Try the speed, the interface, the quality — before committing to the full version.

Get the Basic Version — CAD$19
Want the full version? Pay CAD$89 and get PrivateDoc AI.
Already bought the basic? Your payment counts toward the upgrade.
Уже купили Quickstart за $19? Доплатите $70 — получите полный PrivateDoc AI.
Pick your model

Find the right model for your machine

Answer two questions — get a recommendation and a download link.

What operating system are you on?

All available models

Model Params RAM Context Best for OS Access
qwen2.5:7b★ recommendedAlibaba · 5 GB 7B16 GB128K Lawyers, accountants, HR, consultants. Best balance of quality and speed. ⊞ Windows ✓ Available
qwen2.5:7b★ recommendedAlibaba · 5 GB 7B16 GB128K Lawyers, accountants, HR, consultants. Best balance of quality and speed. ⌘ Mac ✓ Available
gemma3:4bGoogle · 3.3 GB 4B8 GB128K 8 GB RAM machines. Document summarization, single-file Q&A. ⊞ Windows ✓ Available
gemma3:4bGoogle · 3.3 GB 4B8 GB128K 8 GB RAM machines. Document summarization, single-file Q&A. ⌘ Mac ✓ Available
dolphin3:8bDolphin · 4.9 GB 8B16 GB32K Medical staff, researchers. No built-in content restrictions. ⊞ Windows ✓ Available
dolphin3:8bDolphin · 4.9 GB 8B16 GB32K Medical staff, researchers. No built-in content restrictions. ⌘ Mac ✓ Available
qwen2.5:14bAlibaba · 9 GB 14B32 GB128K Powerful workstations. Complex multi-document analysis. ⊞ / ⌘ ⏳ Coming soon
deepseek-r1:7bDeepSeek · 4.7 GB 7B16 GB64K Tax consultants, financial analysts. Shows reasoning steps before the answer. ⊞ / ⌘ ⏳ Coming soon
medllama2:7bMedical · 3.8 GB 7B16 GB4K Doctors, clinics. Trained on medical texts — understands clinical terminology. ⊞ / ⌘ ⏳ Coming soon
About this product

Open-source tools. Pre-built for you.

PrivateDoc AI is built on two open-source projects — Ollama and AnythingLLM. Both are free. Both are public. Anyone can read every line of their code.

We took these tools, tested them across different hardware setups, worked through all the edge cases, and packaged everything into a bundle that just works on the first try.

The installer scripts are open too — you're welcome to read them before you run anything.

🦙

Ollama

The AI engine. Runs models locally. No Docker required.

★ 95,000+ GitHub stars · MIT license
🤖

AnythingLLM

Chat interface + RAG + document workspaces. One .exe install.

★ 54,000+ GitHub stars · MIT license
Questions

FAQ

The script does two things: downloads Ollama from ollama.com and AnythingLLM from anythingllm.com. That's it. Both are open-source and widely used. The script itself is open — you can read every line before you run it.
Yes. Once installed, everything runs completely offline. Nothing is transmitted anywhere. Your files never leave your computer — not to us, not to anyone.
Yes. Setup requires internet — about 7 GB of downloads in total. After that, the whole thing runs offline. Permanently.
No. The installer handles everything automatically. If something goes wrong, email us — we'll get back to you within 24 hours.
Basic (CAD$19): Local AI without the RAG system. One model of your choice from the current catalog. Installed permanently on your machine — no expiry.

Full version (CAD$89): Full RAG system — load entire folders of documents and ask questions across all of them. Access to all models in the catalog, including new ones as they're added. 30 days email support included.
Email us at privatedeskai@gmail.com. We respond within 24 hours. If we can't get it running on your machine within 30 days — full refund, no questions.
Windows 10 or 11. 8 GB RAM minimum, 16 GB recommended. 10 GB free disk space. No GPU required. Mac support coming soon.
Depends on your hardware. CPU only (no graphics card): 20–60 seconds per response. With an NVIDIA GPU: 5–10 seconds. The first response after a break is always slower — the model reloads into memory. That's normal.
It means the software is written in public — anyone can read exactly how it works. Ollama and AnythingLLM are open-source the same way Linux or Firefox are. No hidden functions. No black boxes.