What about scanned PDFs?

Pages without their own text are read by AI automatically, at the same 1 credit per page. They take a little longer than text pages.

Do you keep my documents?

Files are kept for 24 hours so you can parse other pages or ask another question without sending them again, then deleted. You can delete one yourself at any time.

Should I use Parse or Scrape for a PDF?

Scrape reads a document link alongside the web pages you're already scraping. Parse is for documents as the main job: uploads, page ranges, per-page output, estimates, batches and files up to 500 MB.

Parse API

Documents into Markdown, or the fields inside them

Send a PDF, Word, Excel or PowerPoint file, by upload or link, and get Markdown page by page, or JSON fields from a prompt or schema. Scanned pages are read too.

parse…

invoice_numberinvoice_datedue_datetotalcurrencyline_items

Formats

Thirteen formats, one request

The format is detected from the file's content, not its name. Formats without pages count each 3,000 characters of output as a page.

PDF

DOCX

DOC

XLSX

PPTX

PPT

ODT

ODS

ODP

RTF

EPUB

CSV

HTML

Text read as it is. Scans read by AI.

Each page is read from its own text when it has one, so text PDFs are fast and exact. Scanned pages are read by AI, at the same 1 credit per page. Each page in the result says which way it was read.

autoThe default. Pages with their own text are read directly; AI reads only the scanned ones.
fastNever uses AI. Scanned pages come back with an error and aren't charged.
ocrAI reads every page, for documents whose text layer can't be trusted.

Three ways to send a document

Upload a fileUp to 50 MB in the request, or up to 500 MB through a prepared upload.

Send a linkAny document on the web, like a report or a paper.

Reuse a fileIdParse other pages of a file, or ask another question, without sending it again.

Short documents answer in the same request; longer ones return a job ID within 20 seconds. The SDK waits for you.

What you get

Markdown page by page, or your fields

Every parse returns the document as Markdown. Add a prompt or a schema and you also get the fields as JSON. Real runs.

Request

POST /parse

{ "url": "https://arxiv.org/pdf/1706.03762", "maxPages": 3 }

The first 3 of 15 pages, 3 credits. includePages adds each page's Markdown on its own, with how it was read.

What came back

response

{
  "status": "completed",
  "fileName": "1706.03762",
  "format": "PDF",
  "totalPages": 15,
  "pagesParsed": 3,
  "creditsUsed": 3,
  "markdown": "### Provided proper attribution is provided, Google hereby grants permission to\n\nreproduce the tables and figures in this paper solely for use in journalistic or scholarly works.\n\n## Attention Is All You Need\n\n#### Ashish Vaswani\n…"
}

Before you parse

Know the cost first, for free

An estimate counts pages and credits without parsing or charging, and keeps the file so the real parse uses its fileId instead of a second upload.

dryRun: true

{
  "status": "estimate",
  "fileName": "attention.pdf",
  "format": "PDF",
  "totalPages": 15,
  "pagesToParse": 2,
  "credits": 2
}

Deleted after 24 hours

Uploaded files are kept a day so you can parse other pages, then deleted. Delete one sooner with a single call.

Failures cost nothing

A password-protected or broken document comes back as failed with a plain reason, and isn't charged.

Batches of 50

Send up to 50 documents in one batch, links and uploads mixed, up to 2,000 pages each.

What it costs

1 credit per page

Scanned pages included. Plus AI tokens when you ask for fields. Formats without pages count each 3,000 characters of output as a page.

See plans

Costs nothing

Estimates before you parse
Documents that fail and pages that can't be read
Checking status and deleting files

Limits

Up to 2,000 pages per document
Uploads to 50 MB, or 500 MB prepared
50 documents per batch

Use cases

Built with Parse

Docs into RAG

Crawl a documentation site into clean Markdown, code blocks intact.

Learn more

Reports and documents

Find PDFs with search, see what they cost, then parse them into fields.

Learn more

All use cases

Every workflow on one page.

See them all

FAQ

Questions, answered

Parse

PDF, Word (DOCX and DOC), Excel (XLSX), PowerPoint (PPTX and PPT), OpenDocument files, RTF, EPUB, CSV and HTML. The format is detected from the file's content, not its name.

The other tools, same API key

SearchWeb, news, research and GitHub results as data MapEvery URL on a site, in seconds ScrapeOne page or 50, as Markdown or JSON CrawlOnly the pages you describe

You build the product.
Spidra handles the data work.

Start with 300 free credits, about 300 pages. No credit card required.

Using an AI agent? Connect the MCP server