Page-aware
Every chunk keeps its source PDF and page number for traceable retrieval.
AI document workflow
PDF to RAG Chunks Converter
Turn PDFs into structured, page-aware chunks that are easier to retrieve, cite and feed into AI systems. Batch multiple documents, tune the chunking strategy, and download JSON or a ready-to-share ZIP.
Drop PDFs here
Build one AI/RAG-ready dataset from one or many PDFs.
PDFUp to 25 files
Instant processingProtected by HTTPSClipboard paste
Smart chunk settings
Smaller chunks improve precise retrieval; larger chunks preserve more surrounding context. Start around 1,200 / 150 for general documents.
Batch-ready
Process a document set in one run and download individual JSON files inside one ZIP.
Provider-neutral
Export structured data without forcing a particular LLM, embedding model or vector database.
Why convert a PDF into RAG chunks?
PDFs are excellent for sharing finished documents, but they are not automatically shaped for retrieval. A long report can contain hundreds of pages, repeated headers, page breaks and unrelated sections.
A chunked dataset gives your application smaller units to index and retrieve. MAXScanner keeps the original filename and page reference with every chunk, so your application can retain a clear path back to the source.
There is no single perfect chunk size for every document. The tool exposes chunk size and overlap instead of hiding the decision behind one opaque preset.
How it works
01
Add one or more PDFs
02
Extract readable page text
03
Split text with your chosen overlap
04
Download JSON or ZIP
Frequently asked questions
What is a PDF RAG chunk?+
A RAG chunk is a smaller, traceable section of a document that an AI system can retrieve. MAXScanner keeps the source filename and page number with each chunk.
Can I process multiple PDFs?+
Yes. Add up to 25 PDFs in one run. MAXScanner creates one combined dataset and a ZIP containing an individual .rag.json file for each source PDF.
What chunk size should I use?+
Around 1,200 characters with 150 characters of overlap is a sensible starting point for general documents.
Does this create embeddings?+
No. This tool prepares structured chunks and does not lock you into a particular embedding provider or vector database.
Are scanned PDFs OCRed?+
Image-only pages without a text layer may produce empty chunks. Use OCR PDF first when the source is scanned.
What can I do with the JSON?+
You can load it into a vector database, build document search, prepare LLM context, create citation-aware retrieval or feed it into your own ingestion pipeline.
Related MAXScanner workflows: PDF to Markdown, PDF to JSON, OCR PDF, Extract Tables.