| Hey HN I’ve been working on CodexLocal
— a privacy-first, offline AI coding assistant that runs entirely in your browser (no servers, no tracking, no data sent anywhere). It’s built with WebLLM and WebGPU, and supports RAG (retrieval-augmented generation) so it can be context-aware — even without internet access. Think of it as a local ChatGPT-style coding tutor, but one that never leaves your machine. Why I built it
Most AI coding assistants today are cloud-based — great for convenience, but not ideal for privacy-sensitive or educational settings. Bootcamps, schools, and dev teams often want to use AI without sending code or student data to third-party servers. CodexLocal aims to fill that gap. Current MVP features
-----------------
Works fully offline in your browser (WebLLM + WebGPU)
Context memory via local RAG
No login, tracking, or external API calls
Works on Chrome, Edge, and soon Safari
Light and dark themes What’s next
-----------------
File uploads for RAG context
Offline model caching
Classroom/enterprise deployment (commercial tier)
NPM package / SDK for integrating into private dev environments The ask
-----------------
Would love feedback on: Performance and model responsiveness in your browser
Whether this kind of privacy-first setup would fit your org or classroom
Any ideas for improving UX / RAG relevance Try it: https://codexlocal.com Demo video: https://www.youtube.com/watch?v=rnDmwW2x16s&feature=youtu.be
Free for personal use, future commercial tier planned for enterprise deployments.
Thanks for reading — feedback very welcome! |