Document to Markdown: PDF, Word, PowerPoint, Excel to MD

Clean Markdown from PDF, Word, PowerPoint, Excel, HTML and EPUB links, ready for RAG, LLMs and search. $3 per 1,000 documents.

Give it a list of document URLs and get one dataset row per document with the full text as Markdown: headings, lists, links and tables as Markdown tables. Powered by Microsoft's open-source MarkItDown.

• One actor for all common office formats: PDF, DOCX, PPTX (slide by slide), XLSX and XLS (every sheet as a table), HTML, CSV, JSON, XML, TXT and EPUB. Mixed lists are fine.
• Built for pipelines: the Markdown is in the dataset row (markdown), with file name, type, size, SHA-256, page count and the first heading as title. Feed it straight into a vector database or an LLM prompt.
• Fast: in our test, 17 links (Word, PowerPoint, Excel, EPUB, HTML and 9 PDFs including 75- and 100-page research papers) were done in 3 minutes 20 seconds on the default 1 GB; the 10 short files took 7 seconds together.
• You pay only for documents with text. Broken links, unsupported files (images, ZIPs), files over your size limit and PDFs without a text layer are listed with the reason and not charged.
• Safe with untrusted files: each document is converted in an isolated worker with a time limit; links and redirects to private or internal network addresses are refused.

Measured on the Apify platform on 20 October 2026 (UTC), 1 GB memory: 17 URLs, 14 converted, 2 internal addresses blocked, 1 image skipped (not charged).

Use with AI agents (MCP)

Add it as a tool: https://mcp.apify.com?tools=batchpane/document-to-markdown

Input the agent should send: {"documentUrls": [{"url": "https://example.com/report.pdf"}, {"url": "https://example.com/slides.pptx"}]}

Results: one row per URL; markdown holds the text. Short documents take a few seconds each, so agents can usually wait for the result; for long lists check the run later (get-actor-run) and read the rows (get-dataset-items).

Example prompt:

Use batchpane/document-to-markdown to convert https://arxiv.org/pdf/1706.03762 to Markdown and summarise the paper's method section.

Input

Field	Default	Notes
Document URLs	–	[{"url": "..."}] or a public link to a .txt/.csv list (requestsFromUrl); up to 5,000 per run
Maximum file size	25 MB	1-100 MB; larger files are not charged
Also save each result as a .md file	off	results over 1,000,000 characters are always saved as a file
Respect robots.txt	on	can only be turned off together with "I own these files or have the owner's permission"

Example input (tested on the platform):

{
  "documentUrls": [
    { "url": "https://raw.githubusercontent.com/microsoft/markitdown/main/packages/markitdown/tests/test_files/test.docx" },
    { "url": "https://raw.githubusercontent.com/microsoft/markitdown/main/packages/markitdown/tests/test_files/test.pdf" }
  ]
}

Output

{
  "url": "https://arxiv.org/pdf/1706.03762",
  "finalUrl": "https://arxiv.org/pdf/1706.03762",
  "fileName": "1706.03762v7.pdf",
  "fileType": "pdf",
  "bytes": 2215244,
  "sha256": "bdfaa68d8984f0dc02beaca527b76f207d99b666d31d1da728ee0728182df697",
  "pages": 15,
  "billedDocuments": 1,
  "title": null,
  "markdown": "3\n2\n0\n2\n ... Provided proper attribution is provided, Google hereby grants permission to\nreproduce the tables and figures in this paper ...",
  "markdownChars": 39675,
  "markdownTruncated": false,
  "markdownFileUrl": null,
  "httpStatus": 200,
  "status": "ok",
  "error": null,
  "convertedAt": "2026-10-20T11:26:44.932549+00:00"
}

status is ok, failed, blocked or skipped; for anything other than ok, error has a fixed code (NOT_FOUND, TOO_LARGE, UNSUPPORTED_TYPE, EMPTY_RESULT, ...) and nothing is charged.

Pricing

$0.003 per document ($3.00 per 1,000), plus Apify's standard start fee per run ($0.00005 per GB of memory). Compute is included.

Long PDFs count as one document per started 20 pages, because conversion time grows with the page count:

Example	Cost
1,000 Word/PowerPoint/Excel files or PDFs up to 20 pages	$3.00
One 75-page PDF	4 x $0.003 = $0.012
One 100-page PDF	5 x $0.003 = $0.015

Every row shows billedDocuments. You can set a maximum cost per run; the actor stops before it would go over it.

Good to know

• No OCR: scanned PDFs without a text layer give no text (EMPTY_RESULT, not charged). Images inside documents are not described.
• PDF layout: text comes out in reading order as plain paragraphs; complex multi-column layouts and tables in PDFs can come out as plain text rather than Markdown tables (tables in Word, Excel and PowerPoint stay tables). Margin text such as arXiv's vertical stamp appears as short lines at the start.
• Old binary formats .doc and .ppt are not supported (convert them to .docx/.pptx); .xls is.
• Files must be reachable by a direct link without login. Files behind a login, a paywall or a "click to download" page are not reachable (not charged).
• A document that takes longer than 3 minutes to convert is stopped and not charged.
• Safety stop: the run stops after 20 URLs in a row fail ("Stop after failures in a row"; 0 switches only this rule off). A second rule cannot be switched off: after 50 URLs, the run stops if fewer than 25% succeeded.

Responsible use

You choose the documents; use the text only where you are allowed to (copyright, licences, personal data inside the documents). The actor keeps nothing: results are stored only in your own Apify storage. Conversion by MarkItDown (MIT licence); not affiliated with Microsoft.

Support

Open an issue in the Issues tab with the document link and the run ID. We aim to answer within 7 days. Answers may be drafted with help of an AI assistant and are checked by a person.

Changelog

• 0.1: first version.