Starting today you can use Codex in Claude Code 👀 /plugin marketplace add openai/codex-plugin-cc Try it out today with: /codex:review for a normal read-only Codex review /codex:adversarial-review for a steerable challenge review /codex:rescue to let codex rescue your code Enjoy Codex-ing!

if codex is so good like y'all are claiming, why does it need to run inside claude code?

The adversarial review is really cool. I built exactly that in our open source plugin. /code:plan-with-codex a Claude plan, codex review, Claude analyze findings, update plan with what’s valid, pushback on what isn’t, codex re-analyzes pushbacks and also re-checks plan against code, this loops until codex approves plan

claude code and codex finally shaking hands is the crossover we needed

the adversarial review mode is the one I'm most interested in. getting the same codebase challenged by a different model with different training priors is better than any single model self-reviewing. the blind spots in Claude and Codex probably overlap less than you'd expect. that's where the real value is.

hey @garrytan - in case you missed it, this might be of interest to you & gstack community

Bro @blackboxai is already having this feature where you can use claude and codex together

this is getting out of hand

this is cool👀 but i prefer using multi agent on Blackbox

love the progress… but managing plugins vs just using @blackboxai is a different experience 👀

Set this up last Tuesday. /codex:rescue rewrote our authentication layer overnight. By Wednesday it had removed the login screen. 17 users can't access their accounts. My VP of InfoSec called it 'an active incident.' I called it a successful deploy.

The real unlock isn't Codex in Claude Code, it's that plugin marketplaces turning AI tools into platforms means your stack compounds faster than any single model improvement ever could. Which plugin would actually move the needle for solo devs?

How is this different than just running: claude mcp add codex --scope local --transport stdio -- codex mcp-server

Or you can use Maestri and connect them together in their own harnesses

What the point ? By default official harness should work better with their's own model becuse of RL and SFT

they are both still used in their own harnesses, that’s the whole reason why this is effective. CC can hand off an issue to codex to work on in its own harness!

Aaaa so it is just claude code calling codex cli :)

openai building plugins for claude code. the moat is the ecosystem now.

This plugin is basically the first step toward what Blackbox CLI has been shipping for months

current situation:

thank you! can you add a ralph loop? i’ve been liking this

just brazen 🔥

LMAO claude inside codex next pls

this is a meaningful unlock — being able to route specific tasks to codex from inside a claude code session means you can start matching subtasks to the model that's actually best at them, rather than using one model for everything. curious how the context handoff works in practice when you're mid-session on a complex codebase.

/codex:adversarial-review is the sleeper feature here. Having Codex challenge your Claude-written code isn't just fun — it's actually a better code review loop than most teams run. Competing models reviewing each other is the new pair programming.

Bro relax

Starting today? Been doing this for the past year using a custom command

you can just use claudex just move to blackbox cli

@towheretobegin the best of both worlds 🔥

> /codex:rescue to let codex rescue your code Shots fired lol

This will be the most powerfully tool ever

this is cool but based on how anthropic acts they'll start baning accounts that install this plugin.

when your competitor becomes your plugin the war is over

The interesting part isn’t the plugin. It’s that the coding surface is going multi-model by default and the winning product won’t care whose logo solved the bug.

This is what an open AI ecosystem should look like. Competitors building plugins for each other's tools because devs benefit. The /codex:adversarial-review alone could save hours of code review time. Excited to try the rescue feature for debugging legacy code

Using blackbox for claude + codex multiagent workflows, loving it so far

"If you can't beat them, join them" Whatever, Codex team has done exceptional job for Python and Rust.. Not going to look back.