Chat with PDF

Ask questions about a PDF and get answers from Claude. PDF parsed in your browser — only extracted text reaches our server.

Chat with PDF is for the moment when you have a 40-page research paper, a 200-page legal contract, or a manual full of dense tables and you just want the answer to one specific question. Reading the whole thing is not an option; CTRL+F often fails because the question is conceptual ('what's the main finding?', 'is this term liable for X?', 'where are the salary numbers?').

Pickrack's Chat with PDF parses the PDF entirely in your browser using pdf.js — the same library that powers Firefox's built-in PDF viewer. Only the *extracted text* is sent to Anthropic's Claude API along with your question. The binary PDF never reaches our server. You can verify this in DevTools → Network: the /api/ai/chat-pdf call sends { documentText, question } as JSON, not the file.

Free, no signup. Daily quota: 30 chat turns per IP (generous because chat is per-turn, not per-document). Best for text-based PDFs (research papers, contracts, manuals). Scanned PDFs without OCR won't work — for those, use Pickrack's PDF to Markdown tool first to OCR the document, then paste the result.

Key features

• Browser-side PDF parsing — pdf.js extracts text from your PDF in your browser. The binary PDF never uploads to our server — only the extracted text + your question.
• Multi-turn chat — Ask follow-up questions; the chat history is sent with each turn so Claude has context. Last 12 turns kept (the rest is dropped to control token cost).
• Page markers in extracted text — Each page is prefixed with [Page N] so Claude can cite the page in its answer when relevant ('In page 7, the report states...').
• 80K char document limit — Roughly 30-50 pages of text-heavy PDF. Bigger documents are truncated with a notice — for those, summarize sections separately.
• Language follows document — Vietnamese PDF → Vietnamese answers; English → English. Mixed-language PDFs: Claude answers in the question's language.

How to use

1. Step 1: Upload a text-based PDF — Drop a PDF up to 30 MB. The browser extracts text using pdf.js — typically 1-5 seconds for a 50-page document.
2. Step 2: Ask your first question — Type the question into the input at the bottom. Press Enter or click Send. Examples: 'what are the 3 main findings?', 'summarize section 4', 'is the company liable for damages?'
3. Step 3: Read the answer — Claude responds with a concise answer, often quoting the relevant document passage. If the answer isn't in the PDF, it says so plainly — no hallucination.
4. Step 4: Follow up — Ask follow-up questions. The conversation history is included so Claude understands context. 'What about clause 5?' makes sense after asking about clause 4.
5. Step 5 (optional): Start over with a new PDF — Click 'New PDF' to clear the current document and start fresh. The previous PDF and chat are not persisted.

When to use

• Research paper Q&A — drop a 30-page paper, ask 'what's the main contribution?', 'what dataset did they use?', 'what's the effect size?'
• Contract review — upload a contract, ask 'what's the termination clause?', 'what happens if I miss a payment?'
• Technical manual — drop a device manual, ask 'how do I enter pairing mode?', 'what's the warranty period?'
• Resume / CV review — paste a candidate CV (as PDF), ask 'how many years of Python experience?', 'what's their largest project?'
• Academic study guide — upload a textbook chapter, ask 'what are the key formulas?', 'explain example 3 step by step'
• Privacy policy or ToS — drop a long ToS PDF, ask 'is my data shared with third parties?', 'can I delete my account?'

Frequently asked questions

Related tools