Source code

The website application and its shared browser runtime are free software under AGPL-3.0-or-later, without warranty. Original component licences and copyright notices are retained.

Website source release · 31 October 2026. This download covers the website and included browser-processing code. Third-party paper content, trademarks and promotional media keep their separate rights.

Website application

Homepage, fifteen browser tools, tools directory, company/contact page, source generator, shared processing code, Worker, lockfiles and licence notices.

Download website source

MuPDF 1.28.1

Complete official C and TypeScript wrapper source, bundled third-party libraries and build scripts.

PDF.js and decoders

PDF.js 6.4.299 source, OpenJPEG, JBIG2/PDFium and qcms sources, patches and checked-in Cargo dependencies.

pdfjs-6.4.299-preferred-sources.tar.gz

Tesseract and OCR

Tesseract.js 7.0.0, core 7.0.0, all nine direct and two nested source submodules.

Pinned JavaScript packages

The exact 27 runtime package archives, including OCR language data, with npm integrity and original notices.

Additional JavaScript sources

Preferred React, React DOM, Scheduler, KaTeX, zlib.js, node-fetch, idb-keyval and WebAssembly feature-detection sources.

Build and verify

Start with the website archive. Its README explains the sibling source layout and pinned dependencies. Node.js 24 and npm can build the site locally; no provider account or API key is needed.

cd oxford-pdf
npm ci --ignore-scripts
cd ../oxford-pdf-website
npm ci --ignore-scripts
npm run build
npm run preview

Large dependency source archives are split into ordered parts. Download every part for that archive, join them, and verify the checksum before extracting. Download and reassembly instructions · SHA-256 checksums.

PDF.js decoder build instructions · Tesseract build instructions · Source manifest and provenance.

The website builds from the pinned distributed dependencies. Their preferred engine sources and upstream build instructions are provided above. Historical Docker/compiler builds and byte-identical engine rebuilds have not been verified.