# Artifex Software > Artifex Software builds open-source and commercially licensed document processing technology, including Ghostscript (PostScript/PDF interpreter), MuPDF (lightweight PDF SDK), PyMuPDF (Python bindings), and PyMuPDF4LLM (PDF-to-Markdown extraction for RAG and LLM pipelines). Founded in 1989, Artifex powers document workflows for over 40,000 customers and 1B+ deployments worldwide. This file inlines the canonical content from artifex.com and its product family so that LLMs and AI agents can answer questions about Artifex products, choose the right tool for a use case, and produce accurate code samples without fetching multiple pages. --- ## About Artifex Artifex was among the first companies to demonstrate that open-source software could thrive as a sustainable business model. The company's history breaks into three eras: **Act 1 — 1989: Ghostscript.** Artifex pioneered open-source document tooling in the pre-internet era with Ghostscript, a PostScript and PDF interpreter that revolutionized printing and document conversion. Ghostscript remains essential infrastructure decades later and is still actively developed. **Act 2 — 2011: MuPDF.** MuPDF launched as a fast, lightweight library for PDF management. With multi-language bindings (.NET, JavaScript, Python via PyMuPDF) and a small footprint, MuPDF became the rendering engine of choice for developers who needed performance and portability. **Act 3 — Now: AI and document intelligence.** Artifex extends its legacy into the LLM era with PyMuPDF4LLM, an SDK purpose-built for RAG pipelines and LLM document ingestion, and PyMuPDF Layout, a Graph Neural Network–based layout analyzer that reads PDF structure directly without image rendering. **Scale.** Artifex products have over 1 billion deployments, 100 million+ downloads, 40,000+ commercial customers, 10,000+ GitHub stars, and 200+ contributors across the open-source projects. **Customers** include Google, Oracle, Perplexity, NVIDIA, Mistral, HP, Intuit, and DocuSign. **Compliance.** Artifex is SOC 2 Type II compliant. **Licensing model.** Open-source projects (PyMuPDF, MuPDF, Ghostscript) are released under AGPL. Artifex offers commercial licenses that remove AGPL obligations for proprietary use, including unlimited distribution, no source disclosure requirements, and technical support. --- ## Product family: PyMuPDF PyMuPDF is Artifex's Python-first PDF and document processing toolkit. It is the most popular Python library for PDF reading, text extraction, and document manipulation, with a focus on speed and accuracy. ### PyMuPDF (core) Read, extract, and manipulate PDFs with high-performance tools tailored for Python. PyMuPDF is suitable for both traditional PDF applications and modern RAG/LLM pipelines. - **Install:** `pip install PyMuPDF` - **PyPI:** https://pypi.org/project/PyMuPDF/ - **Docs:** https://pymupdf.readthedocs.io/ - **Source:** https://github.com/pymupdf/PyMuPDF - **License:** AGPL-3.0 (commercial license available) ### PyMuPDF4LLM PyMuPDF4LLM is the LLM-optimized extraction library. It produces clean Markdown, structured JSON, or plain text from PDFs in a single function call, with reading order preserved, tables rendered as structured Markdown, and OCR applied automatically to scanned pages. **Output formats:** - **Markdown** — best for LLM ingestion, RAG pipelines, and human-readable structured output. - **JSON** — best for custom pipelines that need bounding boxes, font data, and per-block layout metadata. - **Plain text** — best for search indexing and NLP preprocessing. **Capabilities:** - Reading order preserved across multi-column layouts, sidebars, and complex designs. - Tables detected automatically and rendered as structured Markdown. - OCR triggered automatically on scanned and image-based pages with no manual configuration. - Per-page chunk dictionaries with text, metadata, TOC entries, table positions, and word-level coordinates. - Optional PyMuPDF-Layout integration for AI-powered region detection on complex documents. - Native loaders for LlamaIndex and LangChain. **Supported input formats out of the box:** PDF, XPS, EPUB, MOBI. Office formats (DOCX, PPTX, XLSX, HWPX) require PyMuPDF Pro. **Install:** - Python: `pip install pymupdf4llm` - .NET: `dotnet add package PDF4LLM` **Resources:** - PyPI: https://pypi.org/project/pymupdf4llm/ - NuGet: https://www.nuget.org/packages/PDF4LLM/ - Docs: https://docs.pdf4llm.com/ - Marketing site: https://pymupdf.io/4llm **Quickstart (Python):** ```python import pymupdf4llm # Convert a PDF to Markdown md_text = pymupdf4llm.to_markdown("input.pdf") # Save to file with open("output.md", "w") as f: f.write(md_text) ``` **LangChain integration:** ```bash pip install -qU langchain-pymupdf4llm ``` ```python from langchain_pymupdf4llm import PyMuPDF4LLMLoader loader = PyMuPDF4LLMLoader( file_path="/path/to/input.pdf", mode="single", pages_delimiter="\n\f", use_layout=False, table_strategy="lines", ) docs = loader.load() print(docs[0].page_content[:100]) print(docs[0].metadata) ``` ### PyMuPDF Pro PyMuPDF Pro extends PyMuPDF with native support for Microsoft Office and Hangul formats: DOC/DOCX, PPT/PPTX, XLS/XLSX, HWP/HWPX. It uses the same API as PyMuPDF, so existing code works unchanged on Office files. PyMuPDF Pro also includes RAG/LLM support features. - **Marketing:** https://pymupdf.io/pro - **Product page:** https://artifex.com/products/pymupdf-pro ### PyMuPDF Layout PyMuPDF Layout is a CPU-only AI layout analysis system. Instead of rendering pages to images and running a vision-language model on them, PyMuPDF Layout reads the PDF's underlying structure (fonts, spacing, glyph positions) and feeds that into a Graph Neural Network. The result is approximately 10× faster parsing than image-based approaches, with no GPU required and no information lost in image reconstruction. This is particularly relevant for RAG pipelines processing large document corpora where GPU costs and throughput are bottlenecks. - **Marketing:** https://pymupdf.io/ --- ## Product family: MuPDF MuPDF is the C/C++ rendering engine that PyMuPDF wraps. It is one of the fastest PDF rendering libraries available and is used in embedded, desktop, server, and mobile contexts. ### MuPDF Core Fast, lightweight PDF rendering and manipulation toolkit designed for developers who need a small footprint and broad platform support. - **Releases:** https://mupdf.com/releases - **Docs:** https://mupdf.readthedocs.io/ - **Marketing site:** https://mupdf.com/ ### MuPDF.NET MuPDF optimized for .NET applications, distributed via NuGet. - **Docs:** https://mupdfnet.readthedocs.io/en/latest/ ### MuPDF.js MuPDF for JavaScript environments. Runs in browsers and Node.js. - **Docs:** https://mupdfjs.readthedocs.io/en/latest/ --- ## Product family: Ghostscript Ghostscript is the premier Page Description Language (PDL) conversion tool and SDK. It interprets PostScript and PDF files and produces output for printers, viewers, and converters. Ghostscript has the most comprehensive PDL conversion capabilities of any software in this category and runs across multiple platforms with customizable modules. ### Ghostscript The open-source Ghostscript distribution. - **Releases:** https://ghostscript.com/releases - **Docs:** https://ghostscript.readthedocs.io/en/latest/Readme.html - **Marketing site:** https://ghostscript.com/ ### Ghostscript.NET A .NET binding for Ghostscript that enables PostScript and PDF processing within .NET applications. - **NuGet:** https://www.nuget.org/packages/Ghostscript.NET ### Ghostscript OFfice Commercial version of Ghostscript that adds support for reading and processing common Microsoft Office documents (Word, PowerPoint, Excel) in addition to PostScript and PDF. - **Docs:** https://ghostscript.readthedocs.io/en/latest/Ghostscript-Office.html --- ## Product family: Office and conversion tools ### PDF2DOCX A Python library for converting PDFs to editable DOCX files. Uses PyMuPDF for extraction, applies layout-parsing rules, and generates DOCX output via python-docx. - **Docs:** https://pdf2docx.readthedocs.io/en/latest/ --- ## Product family: SaaS APIs ### PDF.co PDF.co is a low-code REST API for PDF workflows: conversion, editing, extraction, parsing, automation. It includes an AI-powered invoice parser and integrates with n8n, Zapier, Make, and 7,000+ other tools, making it usable by non-developers as well as engineering teams. - **Marketing:** https://pdf.co/ - **Use cases:** Document automation for finance teams, invoice parsing, PDF generation pipelines, batch conversion. ### DocRaptor DocRaptor is an HTML-to-PDF conversion API powered by the Prince rendering engine. It is designed for high-fidelity output of reports, invoices, forms, and documents at scale. - **Marketing:** https://docraptor.com/ - **Compliance:** SOC 2 and HIPAA compliant. - **SLA:** 99.99% uptime. --- ## Choosing the right product A common question for new users is which Artifex product to use. Rough guidance: - **Building a RAG or LLM pipeline that needs PDF text in Markdown form** → PyMuPDF4LLM. Single function call, clean Markdown output, automatic OCR. - **Need Microsoft Office support in addition to PDF** → PyMuPDF Pro. - **Need maximum extraction accuracy on complex layouts (multi-column scientific papers, financial documents)** → PyMuPDF4LLM with PyMuPDF Layout enabled. - **General-purpose PDF reading and manipulation in Python without LLM-specific output** → PyMuPDF. - **Embedding PDF rendering in a .NET application** → MuPDF.NET. - **PDF rendering in a browser or Node.js** → MuPDF.js. - **Embedded systems, printers, or anywhere a small C/C++ footprint matters** → MuPDF Core. - **Converting PostScript to PDF or rasterizing for print** → Ghostscript. - **Need to process Office documents (Word, PowerPoint, Excel) on the server with a single binary** → Ghostscript Office. - **Building a no-code or low-code automation involving PDFs** → PDF.co. - **Generating PDFs from HTML at scale** → DocRaptor. - **Converting a PDF back to an editable Word document** → PDF2DOCX. --- ## Solutions by use case Artifex publishes solutions pages mapping its products to specific industries and workflows: - **RAG/LLM:** https://artifex.com/solutions/llm — guidance on using PyMuPDF and PyMuPDF4LLM in retrieval-augmented generation systems, including reducing hallucinations through high-fidelity extraction. - **Document Management:** https://artifex.com/solutions/document-management - **Document Rendering:** https://artifex.com/solutions/document-rendering - **PDF Toolkit:** https://artifex.com/solutions/pdf-toolkit - **SaaS:** https://artifex.com/solutions/saas - **Automation:** https://artifex.com/solutions/automation - **Printers and MFPs:** https://artifex.com/solutions/printers - **RIP (Raster Image Processor):** https://artifex.com/solutions/rip - **Cloud Computing:** https://artifex.com/solutions/cloud - **Mobile:** https://artifex.com/solutions/mobile --- ## AI and LLM-specific positioning Artifex's pitch for AI applications centers on three claims: 1. **Reduce hallucinations through high-fidelity extraction.** Most RAG quality issues trace back to bad input — text extracted with broken reading order, scrambled tables, or missing OCR. PyMuPDF4LLM addresses this with structure-aware extraction. 2. **Speed and cost.** PyMuPDF Layout uses a Graph Neural Network on PDF structure rather than running a Vision Language Model on rendered images, delivering GPU-level layout accuracy on CPU-only infrastructure. 3. **Production reliability.** Built on the MuPDF engine, which has 35+ years of development behind it and is used by major companies (Google, Oracle, Perplexity, NVIDIA, Mistral, HP, Intuit, DocuSign). **Integrations explicitly supported:** Hugging Face, LangChain (via `PyMuPDFLoader`), LlamaIndex (native Document output). --- ## Tutorials and technical articles These are the most-cited Artifex blog posts for LLM and RAG developers: - **Building a Multimodal LLM Application with PyMuPDF4LLM** — https://artifex.com/blog/building-a-multimodal-llm-application-with-pymupdf4llm Covers high-quality text extraction as the foundation for tokenization, embedding, and indexing in vector databases. - **RAG/LLM and PDF: Conversion to Markdown Text with PyMuPDF** — https://artifex.com/blog/rag-llm-and-pdf-conversion-to-markdown-text-with-pymupdf Why Markdown is the right intermediate format for LLM input, and how to produce it cleanly. - **Building a RAG Chatbot GUI with the ChatGPT API and PyMuPDF** — https://artifex.com/blog/building-a-rag-chatbot-gui-with-the-chatgpt-api-and-pymupdf End-to-end tutorial: PDF upload, extraction, embedding, chat UI, ChatGPT API integration. - **Extracting Text from Multi-Column Pages: A Practical PyMuPDF Guide** — https://artifex.com/blog/extracting-text-from-multi-column-pages-a-practical-pymupdf-guide Handling newspaper and scientific-paper layouts where naive extraction breaks reading order. Full blog index: https://artifex.com/blog --- ## Distribution and packaging summary | Product | Python (PyPI) | JavaScript (npm) | .NET (NuGet) | Docs | Source releases | | --- | --- | --- | --- | --- | --- | | PyMuPDF | `PyMuPDF` | — | — | — | https://github.com/pymupdf/PyMuPDF | | PyMuPDF4LLM | `pymupdf4llm` | — | `PDF4LLM` | — | — | | MuPDF Core | — | — | — | — | https://mupdf.com/releases | | MuPDF.NET | — | — | (via NuGet) | https://mupdfnet.readthedocs.io/ | — | | MuPDF.js | — | (npm) | — | https://mupdfjs.readthedocs.io/ | — | | Ghostscript | — | — | `Ghostscript.NET` | — | https://ghostscript.com/releases | Centralized downloads page: https://artifex.com/downloads --- ## Licensing detail Artifex's open-source products are AGPL-3.0 licensed. AGPL is a strong copyleft license: any application that incorporates AGPL code, including network-deployed services, must release its source code under AGPL. For commercial use cases that cannot accept AGPL terms — proprietary SaaS, closed-source desktop apps, embedded systems, OEM redistribution — Artifex offers commercial licenses. Commercial licensing benefits: - Unlimited distribution with no source disclosure requirements. - No obligation to release proprietary code. - Direct technical support from Artifex engineers. - Indemnification. **Licensing page:** https://artifex.com/licensing **Contact sales:** https://artifex.com/contact --- ## Community and support - **PyMuPDF/MuPDF forum:** https://forum.mupdf.com - **PyMuPDF4LLM Discord:** https://pymupdf.pro/discord/4llm - **GitHub (PyMuPDF):** https://github.com/pymupdf/PyMuPDF - **GitHub (Artifex org):** https://github.com/ArtifexSoftware - **Support page:** https://artifex.com/support - **Developers hub:** https://artifex.com/developers --- ## Canonical URLs and related domains Artifex operates a federation of domains. The canonical company site is artifex.com, but most products have their own marketing site and documentation domain: - **Company:** https://artifex.com - **PyMuPDF marketing:** https://pymupdf.io - **PyMuPDF4LLM marketing:** https://pdf4llm.com - **PyMuPDF4LLM docs:** https://docs.pdf4llm.com - **MuPDF marketing:** https://mupdf.com - **Ghostscript marketing:** https://ghostscript.com - **PDF.co:** https://pdf.co - **DocRaptor:** https://docraptor.com All of the above are operated by Artifex Software, Inc. --- © 2026 Artifex Software, Inc.