Chat with your PDF
100% private — runs on your device, never uploaded. Works offline once loaded.
Upload a PDF and ask it questions in plain language. Answers are pulled straight from your document with the source page shown — and unlike other chat-with-PDF tools, everything runs on your device, so your file is never uploaded.
A private alternative to ChatPDF, Humata and AskYourPDF
Tools that let you "chat" with a document are genuinely useful, but almost all of them upload your PDF to their servers to process it. For a contract, a medical report, a financial statement or anything confidential, that is exactly the file you do not want to hand to a third party.
This tool does the same job entirely inside your browser. Your PDF is read, split into passages, indexed and searched on your own device using on-device AI models. Nothing is uploaded, so there is no server to trust in the first place.
How it works
When you open a PDF, the tool extracts its text and turns each passage into a numerical embedding with a MiniLM model. When you ask a question, it embeds the question the same way, finds the passages whose meaning is closest to it, and then runs an extractive question-answering model over those passages to pull out the exact answer.
Because the answer is quoted directly from your document — and shown alongside the page it came from — the tool cannot invent facts. If the answer is not in the PDF, it tells you so rather than guessing.
- Ask in plain language: "What is the notice period?", "Who signed this?"
- Every answer cites the source page and passage
- Nothing is uploaded — works offline once the models are cached
What it is good for (and its limits)
It shines at finding specific facts in long documents: dates, names, amounts, clauses, definitions and policies. It is ideal for contracts, manuals, research papers, reports and terms of service you would rather not upload anywhere.
Because it runs compact on-device models, it answers factual, extractive questions rather than writing long essays, and it needs a PDF with selectable text (use OCR first for scans). The first document also triggers a one-time model download that is then cached.
Frequently asked questions
Is my document really private?
Yes. Everything — text extraction, indexing and question answering — happens in your browser on your device. The PDF is never uploaded, which is the core difference from cloud chat-with-PDF services.
Can it answer any question?
It answers questions whose answer appears in the document. It retrieves the most relevant passages and extracts the answer from them, so it will not invent information that is not in the PDF.
Why is the first document slow?
The first time you use it, two small AI models download once (an embedding model and a question-answering model) and are then cached. Later documents and questions are much faster.
Does it work on scanned PDFs or images?
It needs selectable text. If your PDF is a scan, run it through the OCR tool first to add a text layer, then chat with that version.
How big a PDF can it handle?
It indexes up to a few hundred passages, which covers most everyday documents. Very large PDFs are truncated to stay fast and within your device memory.
Does it work offline?
Yes, once the models are cached. After the first run you can ask questions with no internet connection.
Advertisement