I built an AI Coding tool optimized for my workflow. VSCode forks use too much memory and I am over the idea of having AI Agents rewriting files and also reviewing complicated diffs with the Accept/Reject UI.

I built Isogen which uses as much memory as a Chrome tab instead of a tool that can spike up to 1GB of RAM. You drag and drop or paste files into an isolated context and do fast generations file-by-file. This approach allows me to keep a strong mental model of the codebase. I also added a snapshot feature that keeps the history of the files and the generated output. File copies are saved locally with SQLite.

Bring Your Own Key for inference which allows for unlimited generations. The only models supported now are the fast ones from Gemini, ChatGPT, Claude and Grok.