That’s more or less what I intend. I want to make different functions configurable and enable/disable in settings, within reason. I haven’t designed the plugin as modular from the beginning, perhaps I should have. But the functions are interconnected enough and I didn’t want to have it require its own plugins.
I use both Codex (OpenAI) and Claude (Anthropic, of course), both directly rather than through alternative harnesses/environments like Cursor. I get better use out of my subscription dollars this way, and in general I haven’t found the short-term advantages of other tools to make up for their various downsides. Staying in each respective company’s own native coding tools and ecosystem has served me better and better over time, and I use MCPs and skills to make up the difference (e.g. experimenting with novel memory skills backed by on-disk markdown files, or ygrep for file search, etc.).