# Extract SDK: PDF extraction for AI in your own code

Source: https://pdfluent.com/extract/sdk/

Read every table, heading, figure, scan, page on your own machine.

A library that turns a document into text in reading order, tables, headings, figures and chunks a model can use. It runs inside your own software.

[Install the SDK](https://pdfluent.com/docs/python/)[See the licence](https://pdfluent.com/sdk/pricing/)

Markdown

JSON blocks

Chunks

Element ids

Tables as cells

Figure crops

Page numbers

Reading order

See it work

## Three steps. Then it is just your code.

terminal

```
1# Python2pip install pdfluent34# Node.js5npm install @pdfluent/node67# Rust8cargo add pdfluent
```

Runs on your machine

What comes back

## Tables that stay tables. And the rest of the page too.

01

### Tables, cell by cell

Rows and columns come back as cells, so every number keeps its header.

02

### Figures, found

Charts and images are found and cropped, with the page they came from.

03

### Scans, read

A scanned page goes through OCR and comes back as the same blocks as any other page.

04

### More than PDF

In Python, CSV, TSV, JSON, XML, RTF, HTML, Markdown, plain text and e-mail come back as the same blocks a PDF page gives.

Speed

## Fast on one CPU.

See the full results

Speed, quality fast

165 *

pages of text per second, on one CPU

61.4 *

pages of Markdown per second, on one CPU

* Text extraction (quality fast) in SDK 1.2.0: 165 pages per second on one CPU against 85 for 1.1.0, measured in one run on one c3-standard-8 over 40 public arXiv papers (1,050 pages), 21 October 2026. Scientific papers only; no statement about scans or forms.

* Markdown extraction (quality fast) in SDK 1.2.0: 61.4 against 45.2 pages per second on one CPU, same run and machine. The Markdown output differs between the two versions on 20 of the 40 documents (a few hundred bytes), because 1.2.0 also contains extraction changes; the timings therefore compare slightly different output.

Quality

## Ahead on quality.

See the full results

Quality, quality auto

0.9282 *

PDFluent SDK 1.4.0, quality auto, scores this overall on opendataloader-bench: higher than every engine on its published leaderboard as of 3 November 2026. Highest on 2 of the 3 measures (reading order and heading hierarchy), second on table structure.

opendataloader-bench, overall score 0 to 1, one scale

1. PDFluent 1.4.0 0.9282
2. opendataloader [hybrid] 0.907
3. nutrient 0.885
4. docling 0.882
5. marker 0.861
6. unstructured [hi_res] 0.841
7. edgeparse 0.837
8. opendataloader 0.831
9. mineru 0.831
10. pymupdf4llm 0.732
11. unstructured 0.686
12. markitdown 0.589
13. liteparse 0.576

PDFluent 1.4.0 with quality auto, page headers and footers kept. Other engines as published in the benchmark README on 2 November 2026. Footnote

* Overall score 0.9282 (SDK 1.4.0, quality="auto", keep_furniture=True, layout, table and formula models installed) on all 200 documents, with the benchmark’s own harness and evaluator, measured 11 November 2026 on the final macOS arm64 release wheel. Competitor figures are taken from the benchmark’s public README as read on 3 November 2026 (highest overall there: 0.907, opendataloader [hybrid]), not re-run by us. The leaderboard engines ran on their own setups while PDFluent was measured on macOS arm64, and table structure depends on the platform. PDFluent is not on that leaderboard. Other engines have submissions open that are not yet on the leaderboard. One run, no confidence interval. The PDFluent row keeps page headers and footers because the benchmark’s ground truth counts them as content. PyMuPDF and MuPDF are trademarks of Artifex Software, Inc. PDFluent is an independent product of Innovation Trigger B.V. and is not affiliated with or endorsed by Artifex.

Good to know

## The small print, up front.

Would you rather not run anything? Extract Hosted is coming soon: the same engine behind one request.

[See Extract Hosted](https://pdfluent.com/extract/hosted/)

Licence

Free to use under the AGPL-3.0. A commercial licence is the alternative if you do not want to publish your own source.

Languages

Rust, Python, Node.js, WebAssembly, Java and.NET, plus a command-line tool.

Element ids

In Python, every block carries an element_id, and the same input gives the same ids.

Agents

pdfluent-mcp is an MCP server for Claude Desktop, Claude Code, Cursor, VS Code and Codex. It runs over stdio on your machine; your documents are not uploaded. [Set it up](https://pdfluent.com/mcp/)

From PyMuPDF

A PyMuPDF-compatible layer is available as a Preview.*

* The PyMuPDF-compatible layer is a Preview in SDK 1.4.0. PyMuPDF and MuPDF are trademarks of Artifex Software, Inc. PDFluent is an independent product of Innovation Trigger B.V. and is not affiliated with or endorsed by Artifex.

Questions

## Four questions.

### What does it cost? +

Nothing under the AGPL-3.0. If you do not want to publish your own source, there is a commercial licence at a published price.

### Which tier should I use? +

fast when you only need text and Markdown at speed. balanced when the tables matter: it adds them with installed models.

### Can my agent use it? +

Yes. The local MCP server reads a PDF as Markdown, text, tables or blocks, on your own machine.

### What if I do not want to install anything? +

Extract Hosted is coming soon. It will be the same engine behind one request.

## Try it on a document you know.

Install it, open one file you can judge, and read what comes back.

[Install the SDK](https://pdfluent.com/docs/python/)[Or call it as an API, coming soon](https://pdfluent.com/extract/hosted/)

Need to edit PDFs instead? [PDF API](https://pdfluent.com/pdf-api/sdk/)
