OpenAI Codex CLI exits beta
After nine months in preview, OpenAI's terminal agent is now generally available. Here's how it compares to Claude Code in real use.
Why it matters
OpenAI's Codex CLI is out of beta. Feature parity with Claude Code on the basics, fewer MCPs, faster on short tasks, slower on long ones. Worth having installed even if you're a Claude user.
Codex CLI, OpenAI's answer to Claude Code, hit general availability today. Short version: it's good, it's not yet great, and you should probably install it anyway.
What's good
- Fast on short tasks. "Rename this variable across the repo" returns in seconds.
- Clean install. One command, no config file required to get started.
- Familiar. If you've used Claude Code, the mental model maps directly.
What's not yet great
- Fewer MCPs. The ecosystem is 6 months behind Anthropic's.
- Slower on long tasks. Multi-step refactors feel sluggish compared to Claude Code with Opus 4.7.
- No equivalent of Claude's "skills" system. You can't easily add reusable personas or workflows.
Who should switch
If you're deep in the OpenAI ecosystem (ChatGPT Team, Codex, API) and want one vendor for everything — yes, switch. Otherwise, install it alongside Claude Code and use whichever fits the task.
Pricing
Bundled with ChatGPT Plus/Team at no extra cost. Standalone API pricing tracks GPT-5.
Our take
Two capable terminal agents is better than one. The pressure is good for everyone.
Source: openai.com
Newsletter
A short weekly email about AI tools and what's worth trying.
Free. No spam. Unsubscribe anytime.