Drop your PDF here
or click to browse Β· max 25 pages (free) / 200 pages (pro)
About this tool
Chat with any PDF using AI. Upload a document, ask questions in natural language, and get answers grounded in the document's content β with page references so you can verify.
How it works
- Upload a PDF β text is extracted locally using pdf.js. No upload, no server.
- Your question + relevant document chunks are sent to an AI model (GPT-4o-mini or Claude 3.5 Haiku) for answering.
- The AI is instructed to answer ONLY from the document, with page-number citations.
- Click any citation to jump to the relevant page in the document.
Privacy
- Text extraction happens locally in your browser.
- Only the relevant chunks (not the full document) are sent to the AI for each question.
- We pass "do not train" flags explicitly. Your documents are never used for model training.
- For maximum privacy, set your own API key in settings β your questions go directly to OpenAI/Anthropic, not through our servers.
Best for
- Reading long contracts, research papers, and reports
- Extracting specific facts from dense documents
- Quickly finding the section you need without reading the whole thing
Frequently asked questions
How accurate are the answers?
The AI is constrained to answer only from your document. For most factual questions (what is X, when did Y happen, who is Z), answers are highly accurate. For subjective questions (is this contract fair?), the AI summarizes but flags it as opinion.
Why is there a page limit?
Each question is sent with relevant document chunks, and we cap chunks to keep responses fast and within model context limits. Free plan: 25 pages. Pro: 200 pages. For longer docs, use the AI Summary tool first to get a condensed version.
Can I use my own API key?
Yes. Go to settings (gear icon, top right) and paste your OpenAI or Anthropic API key. Your questions then go directly to that provider β our servers never see the document chunks or your questions.