Should I remove support for Codex? Read further to understand why I am asking this.

[REPOZY]

Codex users will always have a materially different (and "weaker") experience than Claude Code users. Not because of implementation choices, but because of permanent platform-level gaps that OpenAI has to close.

Claude code offers things that Codex doesn't, and this makes it hard to optimize this plugin even further.

So my question to you is: Should I focus on creating a plugin for Claude Code users only and get rid of the Codex/OpenCode/Gemini CLI support, or should I continue supporting all platforms, even thought users using anything other than Claude Code won't be able to experience the full potential of what this plugin has to offers.

Please have a look at the attached screenshot.

Hooks-enabled Codex on macOS/Linux gets roughly 70% of Claude Code's automation value; Windows native gets skills only.

Codex is and will remain a reduced-capability platform for this plugin. The correct framing for y'all is not "same plugin, different shell", it's "substantial subset of features, skills work fully, lifecycle automation is partial and platform-gated."

Please do share your opinion by casting your vote.

Thank you!

[REPOZY]

I've pushed a big update (v6.5.0) that should now fix ALL the issues you may encounter using this plugin on Codex.
But the main problem can't be resolved: Codex simply doesn't support the same hooks than Claude Code.

So it is highly recommended using this optimized plugin on Claude Code, this will give you the best experience.

The release notes:
https://github.com/REPOZY/superpowers-optimized/releases/tag/v6.5.0

[ivan-vilches]

Aquí tienes la versión en inglés, manteniendo el tono técnico-colaborativo y el cierre simpático:

---

Personally, I wouldn’t remove support for Codex, OpenCode, and other providers and leave only Claude. The model ecosystem moves very fast, and what is the best option today might not be tomorrow. Keeping multi-provider compatibility makes the project more resilient and attractive to a wider group of users.

Also, many people simply don’t use Claude because of the cost. A lot of users rely on Copilot, OpenRouter, local models, or other more affordable alternatives, and removing that support could significantly reduce adoption of the project.

Even if Claude is the main or recommended provider, keeping an abstraction layer for other models gives the fork more long-term flexibility and avoids depending on a single platform.

And speaking completely honestly (and a bit in self-defense 😄), I use OpenCode, so if that support disappears I’d be out of the game.

By the way, do you have a Discord server for the project? I’d be happy to join.

[REPOZY]

Thank you for sharing your thoughts with us @ivan-vilches! I agree with you that multiplatform is key here, it's just really stupid that all these different platforms don't use the same hooks system. Makes it hard to make the tool perform as well on all platforms!

I have pushed some big updates to make sure that the plugin works as good as possible on other platforms such as Codex (Which as many issues regarding the hook system this plugin has before).

You stated you use this plugin on OpenCode, how is your experience using this plugin on that platform? Any issues you've come across?

Thank you!