The invisible engine behind a billion documents - and the open-source company that proved giving code away could still pay the bills.
There is a decent chance that a document you printed, converted or opened today passed through code written by a small San Francisco company called Artifex Software. Its name rarely appears on a screen. Its products - Ghostscript, MuPDF and the PyMuPDF family - sit underneath printers, phones, enterprise apps and, increasingly, AI systems, quietly turning PostScript and PDF into pixels and text.
Artifex builds document-processing technology: software that renders, converts, extracts and manipulates PDF, PostScript, PCL and XPS files. It ships this technology three ways - as source-available SDKs, as APIs, and as hosted SaaS - so a developer can embed the engine in a device, call it over the web, or pull it into a Python script.
The company traces its roots to 1988, when L. Peter Deutsch - a UC Berkeley computer science PhD who spent three decades at Xerox and Xerox PARC - started an open-source PostScript interpreter, reportedly with encouragement from friends at Adobe. He called it Ghostscript, following the convention that GPL projects often begin with the letter "G." Artifex itself was founded in 1993 to handle the commercial licensing of that project.
More than three decades later, that same lineage of code is estimated to run in over a billion deployments worldwide. The technology is boring in the best sense: it is the kind of infrastructure that only gets noticed when it breaks, and it rarely does.
PDF and PostScript look simple on the surface and are anything but underneath. Fonts, layout, color, compression and dozens of edge cases make faithful rendering and clean extraction genuinely hard. Artifex's answer is a set of engines that have been hardened over 35 years of real-world files.
Artifex sells primarily to developers, OEMs and enterprises who embed its technology rather than end consumers. The company cites more than 40,000 customers and 150-plus OEM partners. Named licensees and users over the years have included some of the largest names in technology.
Print manufacturers need a reliable interpreter to turn a print job into a page - Ghostscript. Mobile and enterprise apps need to view, edit and render documents on constrained devices - MuPDF and SmartOffice. Python developers need to pull text and tables out of PDFs at scale - PyMuPDF. And a new wave of AI teams needs clean, structured text to feed retrieval and language models - PyMuPDF4LLM.
Artifex's catalog is really a handful of core engines packaged for different runtimes and jobs - from the browser to the data pipeline.
PostScript and PDF interpreter with a rendering engine and graphics library. The industry workhorse for PDF, PostScript, PCL and XPS.
Fast, lightweight, customizable PDF/XPS rendering and manipulation toolkit for developers - renderer, viewer or SDK.
The MuPDF engine compiled for JavaScript, running in the browser and Node.js.
MuPDF packaged for .NET applications, distributed via NuGet.
High-performance Python bindings to read, extract and manipulate PDFs and other document formats.
Extracts text and tables while preserving reading order and structure for RAG/LLM pipelines - roughly 10x faster, no GPU.
Combines PyMuPDF and PyMuPDF4LLM with support for Office documents and HWP files.
Mobile document productivity suite to view, edit, create and share MS Office and PDF documents (OEM/BYOD).
Web APIs and no-code integrations (Zapier, n8n) for extraction, parsing, conversion, form filling and barcodes.
HTML-to-PDF conversion at scale on the Prince engine, with a 99.99% uptime guarantee.
Artifex was among the first companies to show that open source could be a durable business, not a charity. Its core projects are released under the AGPL for open-source use; anyone shipping proprietary, SaaS or embedded software pays for a commercial license. That dual-license model has funded free software for more than three decades.
Revenue flows from three places: OEM licensing (device and software makers embedding the engine), commercial SDK and API subscriptions, and hosted SaaS such as PDF.co and DocRaptor. Public estimates put annual revenue in the low single-digit millions with a team of around 21 - a lean operation relative to its reach.
Artifex competes in the document-SDK space against commercial vendors like Adobe, Foxit, Apryse (formerly PDFTron), iText and Nutrient (PSPDFKit), as well as open-source libraries such as pdfium and Poppler. Its distinctive position is the combination of a genuinely open codebase, deep OEM trust built over decades, and a rendering pedigree that few competitors can match.
Figures self-reported by Artifex; bars are illustrative scale, not to exact proportion.
Artifex's deepest expertise is in page description languages and rendering - the low-level work of interpreting a document and drawing it faithfully. That know-how, refined on print jobs since the late 1980s, turns out to be exactly what modern AI needs.
Retrieval-augmented generation and large language models are only as good as the text they are fed. PDFs - contracts, filings, manuals, research - are a notoriously hard source. PyMuPDF4LLM applies Artifex's extraction engine to that problem, pulling text and tables while preserving reading order and document structure, and doing it roughly ten times faster than heavier approaches without requiring a GPU.
The company is SOC 2 Type II compliant and operates across North America, Europe and Asia. Its GitHub organization hosts projects with more than 10,000 stars and over 200 contributors, a reminder that the open-source community remains part of the engineering engine, not just the marketing story.
L. Peter Deutsch starts an open-source PostScript interpreter under Aladdin Enterprises.
Ghostscript ships, reshaping how documents are interpreted and printed.
The company is created to handle commercial licensing of Ghostscript.
Tor Andersson begins MuPDF, based on Raph Levien's Libart rendering library.
First MuPDF release built on the new Fitz graphics library after Artifex acquires the project.
Python bindings turn MuPDF into a go-to library for PDF extraction.
Seoul-based ePapyrus acquires Artifex; Jeong Hee Kim becomes CEO of the combined company.
PyMuPDF4LLM, PyMuPDF Pro, MuPDF.js and MuPDF.NET extend the stack toward LLM/RAG and new runtimes.
Document-processing technology - most notably Ghostscript, MuPDF and PyMuPDF - offered as SDKs, APIs and SaaS for rendering, converting and extracting PDF and other documents.
Yes. Core products are released under the AGPL for open-source use, while proprietary, SaaS or embedded deployments require a commercial license under a dual-license model.
Developers, OEMs and enterprises. Artifex cites 40,000+ customers and 1B+ deployments, with licensees that have included HP, Kyocera, Ricoh, Xerox, Google, Oracle and NVIDIA.
PyMuPDF4LLM and PyMuPDF Pro extract text and tables while preserving structure, producing clean input for RAG and LLM pipelines without needing a GPU.
Artifex was acquired by Seoul-based ePapyrus Inc. in April 2022, with Jeong Hee Kim serving as CEO of the combined company.