PDF to Markdown Converter — LLM-Ready Text & RAG Chunks

Convert PDF URLs to clean, LLM-ready Markdown: headings, paragraphs and lists rebuilt from the layout, running headers and page numbers removed, plus plain text, per-page content, RAG chunks and metadata. Finds PDFs linked from web pages. $3 per 1,000 PDFs + $0.20 per 1,000 pages.

Turn PDFs into clean, LLM-ready Markdown. Paste PDF links (or a web page that links to PDFs) and get one row per PDF. The document structure is rebuilt from its layout: headings (from font sizes, section numbering and the PDF's bookmarks), paragraphs re-joined across line and page breaks with hyphenation undone, bullet and numbered lists, and tables of contents. Running headers, footers and page numbers are removed, and so is rotated margin text. Each row also has title, author, subject, keywords, language, creation and modification dates, page count and word count, plus optional plain text, per-page content, or ready-to-embed RAG chunks that carry their heading path. Use it to feed RAG pipelines and vector databases, build chatbot knowledge bases from manuals, research papers, reports, regulations and policies, give AI agents a PDF-reading tool, or bulk-extract text from document archives. Simple pricing: $3 per 1,000 PDFs + $0.20 per 1,000 pages, so a 10-page PDF costs $0.005, with no compute bill on top. Scanned pages without text, failed downloads, invalid or locked files and duplicates are never charged.

What data does it return?

Every run outputs clean, schema-validated rows like these:

url	title	pageCount	wordCount	markdown
https://arxiv.org/pdf/1706.03762	Attention Is All You Need	15	5811	# Attention Is All You Need\n\n…\n\n## 1 Introduction\n\nRecurrent neural networks, long short-term memory…
https://nvlpubs.nist.gov/nistpubs/CSWP/NIST.CSWP.29.pdf	The NIST Cybersecurity Framework (CSF) 2.0	32	9027	# The NIST Cybersecurity Framework (CSF) 2.0\n\n…\n\n- 1. Cybersecurity Framework (CSF) Overview … 1
https://www.irs.gov/pub/irs-pdf/fw9.pdf	Form W-9 (Rev. March 2024)	6	6239	# Request for Taxpayer Identification Number and Certification\n\n…

How much does it cost?

Transparent pay-per-event pricing — you only pay for what you get, and the actor always respects your run budget: set a maximum total charge and it stops cleanly at the cap instead of overrunning.

Event	Price	Charged	Worked example
pdf-converted	$0.003	per PDF converted and delivered (all output formats and chunks included)	1,000 PDFs = $3.00 + their pages
page-converted	$0.0002	per page with text converted (scanned pages without text are free)	A 10-page PDF = $0.003 + 10 × $0.0002 = $0.005; a 100-page report = $0.023
apify-actor-start	$0.00005	per run per GB of memory (platform start fee)	One run at 4 GB = $0.0002

Input options

Field	Type	Required	Default	Description
urls	array	yes	—	Links to PDF files. You can also add a web page (e.g. a reports or publications page): every PDF it links to is converted.
outputMode	string	no	"document"	One row per PDF (whole document), or one row per page (handy for page-level citations).
outputFormats	array	no	["markdown"]	Which content fields to return. Same price whatever you pick.
chunkSize	integer	no	0	Also split each PDF into chunks (one-row-per-PDF mode only) of at most this many characters for embedding, split at headings and paragraphs, each with its heading path. 0 = no chunks. Around 1,500–4,000 suits most embedding models.
includePages	boolean	no	false	In one-row-per-PDF mode, also add a pages array with each page's Markdown/text and word count.
maxPagesPerPdf	integer	no	1000	Convert at most this many pages of each PDF (from the start). Longer PDFs are marked truncated.
maxPdfs	integer	no	1000	Stop after converting this many PDFs (including ones found on web pages).
findPdfLinks	boolean	no	true	When a URL is a web page rather than a PDF, convert the PDFs it links to (one level, robots.txt respected).
password	string	no	—	For password-protected PDFs you are entitled to open. Used for every PDF in the run.

Limits & honest notes

• Text is read from the PDF's text layer. Scanned PDFs (images of pages) have no text layer: they are reported and not charged. For those, use our OCR to Text actor (apify.com/actor-forge/ocr-to-text).
• Tables come out as lines of text in reading order, not as Markdown tables. Images and figures are skipped. Mathematical formulas come out as their plain characters.
• Headings are inferred from the layout. Font sizes, section numbers such as "3.1", and the PDF's bookmarks all count, but bold text at body size without numbering stays a paragraph.
• PDFs up to 100 MB and up to 5,000 pages each (1,000 by default). Password-protected PDFs need the password input.
• Web pages in the input are scanned for links to .pdf files (one level deep, robots.txt respected). If a site's robots.txt can't be fetched (HTTP 5xx or a network error, after one retry), the PDFs linked on that site are skipped (RFC 9309) and the page gets a free row saying why. Pages behind a login are not supported.
• Respects your run budget: set a maximum total charge and conversion stops cleanly at the cap, with no PDF charged twice.

Related ActorForge actors

• PDF OCR — for scanned PDFs with no text layer
• PDF to Text
• Word, Excel & PowerPoint to Markdown — Word, Excel and PowerPoint
• Website to Markdown

Changelog

• 2026-10-18 — Initial release: layout-aware Markdown (headings, paragraphs, lists, table of contents), header/footer removal, metadata, plain text, per-page and per-PDF rows, RAG chunks, PDF-link discovery on web pages.

---

Built and maintained by ActorForge. Reliability first: every actor in the portfolio is monitored by an automated nightly self-test, and breakages are typically fixed the same day.