Updated May 2026
C
Codex Review · 2026 — Pricing, Features & Alternatives
OpenAI's coding agent: CLI, IDE, cloud and GitHub, included with ChatGPT
4.5
/5 · 0
Codex is OpenAI's coding agent (launched May 2025, powered by the GPT-5 Codex model family). It runs on four surfaces: an open-source terminal CLI, a VS Code IDE extension, a cloud agent inside ChatGPT, and a GitHub bot. You delegate a task: it reads the repo, writes the code, runs tests in isolated sandboxes that can execute in parallel, fixes failures and opens a pull request. Codex is not sold separately: it is included in ChatGPT plans (limited Free, Go ~$8/mo, Plus $20/mo, Pro $100-200/mo), or billed per use through the OpenAI API.
4.5
/5
Our verdict
Codex is an excellent choice for developers on chatgpt plans who want to delegate complete coding tasks (fixes, features, reviews, prs) to an autonomous agent.
Best for: Developers on ChatGPT plans who want to delegate complete coding tasks (fixes, features, reviews, PRs) to an autonomous agent
Try CodexFeatures of Codex
End-to-end autonomous agent
Delegate a task: Codex writes the code, runs tests, fixes failures and opens a pull request
Four work surfaces
Terminal CLI, VS Code extension, cloud agent in ChatGPT and GitHub bot, your choice
Isolated, parallel sandboxes
Each cloud task runs in an isolated container, several tasks in parallel without conflicts
Configurable open-source CLI
Open-source terminal client, steerable via AGENTS.md and extensible through MCP
Native GitHub integration
Creates branches, commits and pull requests directly on your GitHub repositories
Included with ChatGPT
Codex access in every ChatGPT plan (limited Free, Go, Plus, Pro), no separate subscription
Pros and Cons
Pros
- Included with ChatGPT: no extra cost for Plus/Pro subscribers
- Parallel execution in isolated sandboxes, automatic pull requests
- Open-source, configurable CLI (AGENTS.md, MCP)
- Four surfaces covered: terminal, IDE, cloud, GitHub
Cons
- The best models stay tied to the OpenAI ecosystem
- Usage caps per rolling window depending on the plan
- Younger CLI than rivals like Claude Code or Aider
Use Cases
Delegating coding tasks (bugfixes, features) Opening and reviewing GitHub pull requests Multi-file refactoring in an isolated sandbox Coding agent in the terminal (CLI) or in the IDE Running tasks in parallel in the cloud
Sponsored Get connected
Global mobile data for travelling employees — one plan, 200+ countries, no roaming surprises.
Ready to try Codex ?
OpenAI's coding agent: CLI, IDE, cloud and GitHub, included with ChatGPT
Start for freeAlternatives to Codex
Frequently Asked Questions
Is OpenAI Codex free?
Codex is included in every ChatGPT plan at no extra cost, including the Free tier with limited usage. For regular coding work you will want ChatGPT Go (about $8/month) or Plus ($20/month); Pro tiers ($100-200/month) raise the limits roughly 5x to 20x. Alternatively, the Codex CLI can run against your own OpenAI API key and you simply pay per token.
What is OpenAI Codex?
Codex is OpenAI's coding agent, launched in May 2025 and powered by the GPT-5 Codex model family. You delegate a task in plain language: it reads your repository, writes the code, runs tests in isolated cloud sandboxes that can execute in parallel, fixes failures and opens a pull request. It works across four surfaces: an open-source terminal CLI, a VS Code extension, a cloud agent inside ChatGPT and a GitHub bot.
Codex vs Claude Code: which one should you pick?
Both cost about $20/month to start, but they differ in shape. Codex wins on bundle value (it is included with ChatGPT), parallel cloud sandboxes and a GitHub-native pull request flow across CLI, IDE, web and mobile. Claude Code is the more mature terminal agent, with a larger effective context for huge codebases, fine-grained sub-agent control and deep MCP/SKILL.md extensibility. Teams already paying for ChatGPT should try Codex first; developers who live in the terminal on large repositories tend to prefer Claude Code.
Is OpenAI Codex open source?
The Codex CLI is open source: you can inspect the code, configure behavior through AGENTS.md files and extend it via MCP servers. The underlying GPT-5 Codex models are proprietary and only accessible through your ChatGPT plan or the OpenAI API. So the tooling layer is open, the models are not.
Does Codex work with JetBrains IDEs?
There is no native JetBrains plugin. Codex officially supports VS Code through its extension, and the terminal CLI is editor-agnostic, so it pairs fine with IntelliJ, PyCharm, Neovim or any other editor via the command line. Cloud tasks run inside ChatGPT on web and mobile, and the GitHub bot handles pull requests without any IDE at all.
How much does OpenAI Codex cost?
Codex has no standalone price: usage rides your ChatGPT plan. Free gets limited access, Go costs about $8/month, Plus $20/month, and Pro tiers run $100-200/month with roughly 5x to 20x the Plus limits. Business and Enterprise are billed per seat. If you run the CLI on an API key instead, you pay standard OpenAI API token rates for the GPT-5 Codex models.
What's the difference between ChatGPT and Codex?
ChatGPT is OpenAI's general-purpose assistant; Codex is its coding agent. ChatGPT answers questions, writes and reasons across any topic, while Codex is built for autonomous development work: it reads your repository, edits files, runs tests in isolated sandboxes and opens pull requests. Codex runs as a terminal CLI, a VS Code extension, a cloud agent inside ChatGPT and a GitHub bot, and access is bundled into the same ChatGPT subscription.
Can Codex access my computer?
The Codex CLI runs locally on your machine, but you stay in control: it asks for approval before running commands outside its sandbox, and you can set it to read-only, workspace-write or full-access mode. Cloud tasks never touch your computer; they run in isolated containers on OpenAI infrastructure with a copy of your repository. If you want zero local footprint, use only the cloud agent in ChatGPT.
