The library behind our on-device tools, the datasets behind our research, and the MCP server for AI builders are all public on GitHub. Read the code, reuse it, or open an issue, everything lives under one organization.
Each project is independent, documented, and open. Star the ones you find useful.
Client-side image processing library
Privacy-first image processing in the browser: convert and compress PNG, JPEG, WebP, AVIF, GIF, and ICO fully client-side. Files never leave the device.
npm i @cleanor/browser-image-toolsOpen datasets & reproducible benchmarks
Reproducible benchmarks and open datasets: image compression (JPEG/WebP/AVIF/JPEG XL), the HEIC-to-JPG size tax, cloud storage cost per GB, and phone storage capacity.
Zero-auth MCP server for AI builders
The source behind mcp.cleanor.app: a zero-auth hosted MCP server that lets AI builders optimize and convert images, pull cited storage and image-format data, and generate QR codes.
npm i @cleanor/mcpWordPress image optimizer plugin
A WordPress image optimizer that auto-converts Media Library images to WebP and AVIF for faster pages, better Core Web Vitals, and less storage. No account or API key.
Curated list of MCP servers
A curated collection of Model Context Protocol servers for AI builders exploring the wider MCP ecosystem.
Being able to read the code is the strongest version of “files never leave your device.”
The conversion and compression that runs on cleanor.app is the same open library you can read line by line, so the privacy claim is verifiable, not just stated.
Every figure in our studies is backed by open datasets and scripts, so anyone can re-run the benchmarks and check the numbers themselves.
The MCP server and image library are designed to be connected to and reused, so other tools and agents can build on the same foundations.
What people usually ask about our open source work.
All of our public repositories live under the cleanor-app organization on GitHub at https://github.com/cleanor-app. Each project has its own README, license, and issue tracker.
Yes. The repositories are public and open source. The browser-image-tools library and the storage datasets are built to be reused, and the MCP server is documented for AI builders to connect to directly.
The browser tools on cleanor.app run on the same open-source browser-image-tools library, so the image conversion and compression that happens entirely on your device is the exact code you can read and audit on GitHub.
The benchmarks and datasets behind our studies are published in the cleanor-storage-lab repository, so every number on the research pages is reproducible from open data and open scripts.
New repositories, benchmarks, and datasets land under the same organization. Star it to keep up.