PaperClip

PaperClip is a Chrome browser extension built for AI researchers to save and retrieve details from machine learning, computer vision, and natural language processing papers, blog posts, and news articles. Users highlight text on a page - including sites like Hugging Face, arXiv, Wikipedia, TensorFlow, OpenAI, and Weights & Biases - and right-click to "memorize" it. Saved snippets are indexed locally and can later be found through a similarity search rather than exact keyword matching. All processing happens on-device using an in-browser AI model, so no text or data is sent to an external server, and the extension works fully offline once installed. Users can search, reset, or delete their saved data at any time, and the storage footprint is small - the developer cites roughly 0.12MB for 24 saved items as an example. PaperClip was built by Hugo Duprez using SvelteKit and Transformers.js. It functions as a personal reference tool for researchers who read many papers and want a lightweight way to recall specific findings later, rather than a full note-taking or citation-management system. The extension does not disclose pricing, and no company, funding, or public API details are stated on its site.