I upgraded to Grok's $300 plan to try out Grok Bot, and ended up finding something else that's worth your time more.
I run Codex Review and Codex Execute inside Claude Code on real app builds. Claude Code plans it, GPT-5.6-sol argues with the plan, then it gets built.
So I ran the same jobs through Grok's bridge instead.
Grok was faster, and more efficient at the planning as well as the building. Whether that's the bridge, the API or the model, I genuinely don't know yet. I'm moving more of my work over anyway.
The bridge is free: github.com/xai-org/grok-build-plugin-cc
Paste that link into Claude Code and it installs the commands itself. Check, review, critique and delegate, out of the box. I built my own /grok on top of it so it picks the right model for whatever I'm asking.
You do need a paid Grok plan for this. I'm on the $300 SuperGrok Heavy, but $30/mo does most of it.
Full walkthrough is in the classroom: