OpenAI Just Put a Trojan Horse Inside Claude Code

OpenAI just released an official plugin that runs inside Claude Code — a coding tool built by Anthropic, their direct competitor.

It's called codex-plugin-cc. Within days of launch, it had over 3,700 stars on GitHub. Thousands of developers installed a tool made by one AI company inside the product of another AI company, and nobody blinked.

If you're a business owner watching the AI tool landscape and trying to figure out who to bet on, this move tells you more than any product announcement or press release could.

What the Plugin Actually Does

Claude Code is a command-line coding tool where developers write software with AI assistance. It's built by Anthropic — the company behind Claude. As we covered in 73% of Developers Now Use AI Coding Tools Daily, these tools have become the default way software gets built.

OpenAI's new plugin adds three things to Claude Code:

Code review. After Claude Code writes or modifies code, the plugin sends that code to OpenAI's Codex infrastructure for a second opinion. Think of it like having one contractor do the work and a different inspector check it.

Adversarial review. This goes further than a standard check. The plugin actively challenges design decisions and looks for weaknesses in the code Claude produced. It's not just asking "does this work?" — it's asking "is this the right approach?"

A Review Gate. This is the interesting one. The plugin can block Claude Code from finalizing changes until OpenAI's Codex completes its review. One company's product can now hold up another company's product.

Why This Is a Classic Business Move

This isn't about code review. There are plenty of ways to review code. This is a distribution strategy.

OpenAI looked at where developers spend their time and decided: instead of convincing them to leave Claude Code, just show up inside it. Get developers comfortable using Codex features without switching tools. Let them build a habit. Then, when OpenAI merges ChatGPT, Codex, and their Atlas browser into a single desktop app — which they're doing right now — those developers already know the Codex experience. The switch becomes a slide, not a jump.

It's the same playbook as putting a free sample table inside a competitor's store. You're not trying to burn down the store. You're trying to make sure every customer walking out has tasted your product.

What This Tells You About the AI Market

For businesses choosing AI tools, this move reveals three things worth paying attention to.

First, no single AI tool is going to win. If OpenAI thought they could simply out-build Claude Code, they wouldn't bother making a plugin for it. The fact that they're embedding inside a competitor's product means they've accepted that developers use multiple tools. We outlined this dynamic in Evaluating AI Vendors Without the Hype — the market is fragmenting, not consolidating.

Second, your AI tools are going to get tangled. Today it's a Codex plugin inside Claude Code. Tomorrow it could be Google Gemini features embedded in Microsoft Copilot, or Anthropic tools running inside OpenAI's app. The lines between vendors are blurring. For businesses, this means your "AI stack" might involve three competing companies without you realizing it.

Third, the Review Gate pattern is worth watching. One company's tool holding up another company's tool is new. If this pattern spreads, it creates dependency chains. Your developer might be using Claude Code, but their workflow could stall because an OpenAI server is slow. That's a different kind of vendor risk than most businesses think about.

The Bigger Picture: Bundling Has Started

OpenAI isn't just making plugins. They're merging ChatGPT, Codex, and their Atlas web browser into one desktop application. That's a chat assistant, a coding tool, and a web browser — all in one app, all feeding data to the same company.

This is the same pattern that played out in enterprise software for decades. Start with a product people love. Add adjacent features. Bundle them together. Make it inconvenient to use anything else. SaaS stocks are down 80% in part because AI agents are replacing individual tools — and bundling accelerates that.

The difference in AI is the speed. Traditional enterprise software bundling took years. AI companies are doing it in months. The Codex plugin launched, gained thousands of users, and now those users are primed for the bundled desktop app — all within weeks.

What This Means for Your Business

If you're not a software company, you might wonder why any of this matters.

It matters because the tools your developers use shape what you can build, how fast you can build it, and who you depend on. If your team is using Claude Code with the Codex plugin, you now have a dependency on two competing AI companies. If either one changes their pricing, their terms of service, or their API availability, your development workflow is affected.

The practical takeaway: ask your team what AI tools they're actually using. Not just the primary one — the plugins, the extensions, the add-ons. Know where your dependencies are before one of these companies makes a business decision that disrupts your workflow. We covered the framework for deciding what to build versus buy in AI agent teams — the same logic applies to developer tools.

And if you're evaluating AI vendors right now, accept that the landscape is shifting underneath you. The company you choose today might end up embedded inside a competitor's product by next quarter. That's not a reason to avoid choosing — it's a reason to stay flexible, avoid long lock-in contracts, and keep your options open.

The Move Nobody Saw Coming

A year ago, the idea that OpenAI would ship a product designed to run inside Anthropic's tool would have sounded absurd. These companies are in a head-to-head race for AI dominance.

But platform strategy has always been about pragmatism over pride. OpenAI would rather have their code running inside every developer's workflow — even a competitor's — than win on principle and lose on distribution.

It's a smart move. And it's a signal that the AI tool market is entering a phase where the battle isn't just about who builds the best model. It's about who shows up where the developers already are.

Related: Evaluating AI Vendors Without the Hype and AI Agent Teams: Buy vs. Build.