Feature request: official Claude Code / Codex compatibility

Hi SakanaAI team, thanks for the great work on AI-Scientist-v2.

I’d like to ask whether you would consider adding official compatibility with Claude Code and Codex, rather than requiring users to wire the workflow through provider/API integrations only.

More specifically, I’m wondering whether AI-Scientist-v2 could support a documented way to run through Claude Code / Codex style agent environments, for example via a lightweight adapter, skill/plugin, or CLI bridge.

A few concrete questions:

1. Is official Claude Code / Codex support on the roadmap?
2. Is there already an unofficial but recommended way to run AI-Scientist-v2 with these environments?
3. Are there technical blockers today, such as tool-calling assumptions, state management, sandboxing, or long-running experiment orchestration?
4. If partial compatibility already exists, could you share a minimal example or setup guide?

The motivation is that many users already have access to Claude Code or Codex and would prefer to reuse those environments, instead of separately managing model/provider integrations for the full pipeline.

Thanks!