PDF & DOCX to Markdown Converter for AI Agents
Pricing
from $1.00 / 1,000 page converteds
Convert PDF, DOCX, HTML and text documents from URLs into clean Markdown with headings, lists, tables, links and metadata, plus optional heading-aware RAG chunks and OCR for scanned pages. No browser, no external API. Pay per page converted ($0.001).
Documents to convert, one per line: public http(s) URLs of PDF, DOCX, HTML, Markdown or plain-text files, Apify key-value-store references in the form kvs:/// (use kvs://default/ for this run's own store), or data: URLs with inline base64 content. Each entry becomes one document record. The prefilled PDFs are small public files used by Apify's daily automated test.
Also emit one chunk record per heading-aware chunk of the Markdown (recordType: chunk) with chunkIndex, startPage, endPage, headingPath and tokensEstimate — ready for embedding pipelines. Each chunk is charged separately ($0.0002).
Target size of each chunk in estimated tokens (~4 characters per token). Blocks such as tables and paragraphs are kept whole where possible; headings start new chunks once a chunk is half full.
How much of the previous chunk's tail is repeated at the start of the next chunk, in estimated tokens. Capped at half the chunk size.
When a PDF page has no extractable text (scanned or image-only), render it and run English OCR (tesseract, bundled, no external API). Pages that would need OCR are always listed in needsOcrPages; this toggle decides whether they are actually recognised. OCR'd pages cost $0.005 each on top of the page price and take 2–5 s per page.
Also write each document's full Markdown as .md into the run's default key-value store and record the key as kvsKey. Useful for documents whose Markdown is too large to read comfortably from the dataset.
Insert comments between PDF pages so chunks and citations can reference page numbers. Turn off for Markdown without any HTML comments.
Files larger than this are rejected with a TOO_LARGE error item instead of being downloaded. Raise it up to 50 MB for large scans; bigger files need more memory.
Maximum time to download one file before it is recorded as a TIMEOUT error item.
