PDF & Document to Text/Markdown (DOCX, HTML)

Turn public PDF, Word (.docx), HTML and text files into clean Markdown or plain text, ready for LLMs, RAG pipelines, search indexes and data entry.

• PDF: text in reading order, with headings detected from font size, bullet and numbered lists, de-hyphenation, and | between columns that sit far apart. Rotated margin text and watermarks are skipped. You also get metadata (title, author, dates, producer), page counts, optional per-page output, and a scanned-PDF flag for files with no text layer.
• DOCX: headings, lists, bold/italic, links and tables.
• HTML: main-content extraction; navigation, header and footer are dropped.
• TXT / MD / CSV / JSON: passed through as UTF-8.
• Chunking for embeddings: paragraph-aware chunks with overlap.
• Big outputs are stored in the key-value store and linked from the dataset row, so you never hit item size limits.

No OCR: image-only PDFs are flagged as no-text and are not charged.

Use cases

• RAG pipelines: turn PDF and DOCX links into Markdown chunks with overlap, ready for embedding.
• AI agents: give an agent a document URL and get readable text back.
• Search indexing: extract text and metadata (title, author, pages) from document libraries.
• Research and compliance: bulk-convert filings, reports and policies.
• Data pipelines: a document-to-text step for Zapier/Make/n8n.

Input

{
  "startUrls": [{ "url": "https://arxiv.org/pdf/1706.03762" }, { "url": "https://example.com/report.docx" }],
  "outputFormat": "markdown",
  "includePages": false,
  "chunkSize": 2000
}

Output sample (shortened)

{
  "url": "https://arxiv.org/pdf/1706.03762", "status": "ok", "documentType": "pdf",
  "title": null, "totalPages": 15, "pagesProcessed": 15, "words": 6268, "characters": 40210,
  "likelyScanned": false, "fileSizeBytes": 2215244, "sha256": "…",
  "metadata": { "author": null, "creator": "LaTeX with hyperref", "producer": "pdfTeX-1.40.25", "creationDate": "2024-04-10T21:11:43.000Z" },
  "markdown": "# Attention Is All You Need\n\n…\n\n### Abstract\n\nThe dominant sequence transduction models are …",
  "chunks": ["# Attention Is All You Need …", "…"]
}

The dataset has two views: Overview and Content.

Pricing (pay per event)

Event	Price	When
document-converted	$0.002 per document	A document was downloaded and produced non-empty Markdown/text.

These are never charged: download errors (404, timeouts), unsupported or legacy formats (.doc), oversized files, password errors, and scanned PDFs with no text. Platform usage is extra. Memory defaults to 1 GB; raise it for very large PDFs.

Cost example: converting 1,000 PDFs costs $2.00 in events, plus platform usage (typically a few cents for average-size PDFs).

Limits

• Only publicly reachable URLs. No logins or cookies.
• Complex layouts (multi-column scientific papers, forms, tables) come out as readable text, not pixel-perfect structure.
• Legacy .doc, .ppt, .xls and images are not supported.

Use it from AI agents (MCP), Zapier, Make, n8n or the API

AI agents (Claude, Cursor, VS Code, ChatGPT and other MCP clients). Add Apify's hosted MCP server with just this Actor as a tool:

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=steadyline/pdf-document-to-markdown"
    }
  }
}

Sign in with Apify (OAuth) on first connect, or add an Authorization: Bearer <APIFY_TOKEN> header. Then ask your agent, for example:

Use steadyline/pdf-document-to-markdown to convert these 5 PDF reports to Markdown, split into 1,500-character chunks, and summarize each.

The agent sees this Actor's input schema and output fields, so it can call it without extra prompting. Agents that use the general Apify MCP server can also find it with search-actors.

This Actor uses pay-per-event pricing with limited permissions, so it can also be run through Apify's agentic payments (no Apify account needed for the agent). See the Apify docs.

Zapier, Make and n8n. Use the official Apify app: Zapier "Run Actor" action, Make "Run an Actor" module, or the n8n Apify node. Choose steadyline/pdf-document-to-markdown, paste the input below, then use "Get dataset items" (or the run's dataset) to send results to Google Sheets, Slack, a CRM or a database.

API (any language). Run it and get the results in one call:

curl -X POST "https://api.apify.com/v2/acts/steadyline~pdf-document-to-markdown/run-sync-get-dataset-items?token=$APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"startUrls": [{"url": "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf"}]}'

Or with the Python client (pip install apify-client):

from apify_client import ApifyClient
client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("steadyline/pdf-document-to-markdown").call(run_input={"startUrls": [{"url": "https://www.w3.org/WAI/ER/tests/xhtml/testfiles/resources/pdf/dummy.pdf"}]})
for item in client.dataset(run["defaultDatasetId"]).iterate_items():
    print(item)

Works well with other Steadyline Data tools

• Website → RAG knowledge base: find document URLs with Site Discovery, convert them here, and fetch HTML pages as Markdown with HTTP Request Runner.
• Related: Site Discovery (sitemaps, robots.txt, llms.txt), Bulk HTTP Request Runner, App Store Reviews & App Details.

FAQ

Does it do OCR? No. Scanned (image-only) PDFs are detected and flagged with likelyScanned: true so you can route them to an OCR tool.

Which formats are supported? PDF, DOCX, HTML and plain text from public URLs. Password-protected PDFs work if you supply pdfPassword.

What does it cost? $0.002 per document converted. Failed downloads are not charged.

Can I run it on a schedule? Yes. Create an Apify Schedule (or a saved Task with a schedule) and connect a webhook or integration to receive results.

Need a feature or found a bug? Open an issue on the Issues tab of this page or email steadylinedata@proton.me. Built and maintained by Steadyline Data.

Acceptable use

Convert only documents you have the right to access and process. Respect copyright and the source site's terms. You're responsible for how the extracted content is stored and used.

---

Built and maintained by Steadyline Data. PDF parsing uses Mozilla pdf.js (Apache-2.0); DOCX uses mammoth (BSD-2-Clause).