OpenAI Codex
OpenAI · OpenAI Codex: Coding Agent for Terminal, IDE, Cloud, and ChatGPT
Codex is OpenAI's coding agent that reads, edits, runs, and reviews code in your repository. It works through an open-source CLI, IDE extensions, cloud tasks that run in parallel, and the ChatGPT desktop and web apps, and is aimed at developers and teams already paying for ChatGPT or using the OpenAI API.
Best for
Developers and teams already paying for ChatGPT Plus, Pro, Business, or Enterprise who want one coding agent across the terminal, IDE, and parallel cloud tasks
Not ideal for
Teams that want a single agent billed across many model vendors, organizations that need a self-hosted cloud agent, or users who want cloud tasks without a paid ChatGPT plan
Who it's for
Software developers and engineering teams on ChatGPT plans or the OpenAI API who want local and cloud coding agents
Capabilities
- Open-source Codex CLI (Apache-2.0) for macOS, Linux, and Windows, installable via npm, Homebrew, or a standalone installer
- IDE extension for VS Code, Cursor, and Windsurf, plus Codex integrations in Xcode and JetBrains IDEs, with handoff of longer tasks to Codex cloud
- Codex cloud: each task runs in its own isolated environment with configurable dependencies, setup steps, and per-environment internet access, and returns a summary and diff for review
- Start cloud tasks from the web, GitHub, GitLab (beta), Linear, or Slack, and run several attempts in parallel
- Codex in the ChatGPT desktop app (macOS, Windows, Linux) with a browser, scheduled tasks, and computer use (region-limited)
- Code review against uncommitted changes, a specific commit, or a base branch, with prioritized findings
- AGENTS.md project instructions (created with /init), subagents, MCP servers, and reusable skills and plugins
- Non-interactive `codex exec` for scripts and CI, a GitHub Action, and a Codex SDK for TypeScript and Python
- GPT-6 Astra, Sol, and Luna models, plus support for any provider that implements the Chat Completions or Responses APIs
Limitations
- On ChatGPT plans, Free and Go only include GPT-6 Luna in the desktop app (subject to rollout). The CLI, IDE extension, web, and cloud integrations start at Plus ($20/month)
- Plan usage is measured per five-hour window, weekly limits may also apply, and consumption varies by model, task size, and local vs cloud execution. Going past included limits requires credits, priced by plan or agreement
- With an OpenAI API key instead of a plan, Codex works in the CLI, SDK, and IDE extension only; cloud features such as Codex cloud, GitHub code review, and Slack are not included
- Codex cloud requires connecting GitHub or GitLab, and GitLab support is still in beta
- Only the CLI is open source. Codex cloud runs on OpenAI infrastructure, and no self-hosted cloud option is documented
- GPT-5.5 retires from Codex on October 14, 2026, so workflows pinned to it need to move to GPT-6 models (GPT-5.5 stays on the OpenAI API)
Use cases
- Running the CLI in a local repository to implement a feature, run the test suite, and fix failures
- Kicking off several cloud tasks in parallel from GitHub issues or Linear tickets and reviewing each diff before merging
- Reviewing a branch against main with the code review command before opening a pull request
- Running `codex exec` or the GitHub Action in CI to automate fixes, triage, or release chores
- Starting a Codex cloud task from a Slack thread to investigate a bug and reviewing the resulting diff later
Our take
Codex's strength is breadth. The same agent runs in a local terminal, inside the editor, in parallel cloud environments, and in the ChatGPT apps, and local work can be handed off to the cloud. The open-source CLI and the `codex exec`, GitHub Action, and SDK options make it scriptable. The tradeoffs are ecosystem and predictability: plan-based access favors teams already on ChatGPT, Free and Go are limited to the desktop app, API-key use drops the cloud features, and usage limits vary with model and task size. Teams standardized on another model vendor will find less reason to switch.
Who should use it
Developers and teams already on ChatGPT Plus, Pro, Business, or Enterprise; engineers who want to delegate parallel tasks to the cloud from GitHub, Linear, or Slack; and teams that want an open-source CLI they can script in CI.
Who should skip it
Teams that want to switch freely between model vendors under one bill, organizations that need a self-hosted cloud agent, and users who want the CLI or IDE extension without paying for Plus or API usage.
Strengths
- One agent across the CLI, IDE, cloud tasks, and the ChatGPT apps, with local work that can be handed off to the cloud
- Open-source CLI under Apache-2.0
- Parallel cloud tasks that can be started from GitHub, Linear, or Slack
- Included in ChatGPT plans many developers already pay for, or billed at API rates with an API key
- Scriptable through `codex exec`, a GitHub Action, and TypeScript and Python SDKs
Weaknesses
- Free and Go plans only get GPT-6 Luna in the desktop app
- Usage limits (five-hour windows, with possible weekly limits) depend on model and task size, which makes capacity harder to predict
- API-key access excludes cloud tasks and the GitHub and Slack integrations
- Cloud tasks need GitHub or GitLab, and GitLab is in beta
- No self-hosted option for the cloud agent
OpenAI Codex pricing
Free
$0
Billed monthly
- GPT-6 Luna at Standard speed in the desktop app, subject to rollout
Go
$8/month
Billed monthly
- Same Codex model access as Free
Plus
$20/month
Billed monthly
- Codex on the web, in the CLI, in the IDE extension, and on iOS
- GPT-6 Sol and GPT-6 Luna
- Cloud integrations such as automatic code review and Slack
Pro
From $100/month
Billed monthly
- 5x or 20x more Codex usage than Plus
Business
$20/user/month (annual) or $25/user/month (monthly)
Billed monthly
- Minimum 2 users
- Larger virtual machines for cloud tasks
- SAML SSO, MFA, and admin controls
Enterprise & Edu
Contact sales
- Priority request processing
- SCIM, EKM, and user analytics
Note: Usage is estimated per five-hour window, weekly limits may also apply, and consumption depends on the model, the size and complexity of tasks, and whether they run locally or in the cloud. After reaching included limits, users can buy credits; credit prices and discounts depend on the plan or agreement. Using Codex with an OpenAI API key bills at standard API token rates instead and covers the CLI, SDK, and IDE extension, but not cloud features such as Codex cloud, GitHub code review, or Slack.
Technical specs
API pricing
With an OpenAI API key, Codex usage in the CLI, SDK, and IDE extension is billed at standard API token rates instead of plan limits; cloud features are not included
Available models
Where OpenAI Codex excels
Clearing a backlog of small, well-scoped issues
Codex cloud gives each task its own isolated environment and can run several in parallel, returning a summary and diff for each so a developer can review and merge them one by one.
Pre-merge review of a feature branch
The code review command checks uncommitted changes, a commit, or a whole branch against its base and returns prioritized findings before a human reviewer looks at the pull request.
Automating repetitive fixes in CI
`codex exec` and the GitHub Action run Codex non-interactively inside pipelines, and the TypeScript and Python SDKs let teams build it into internal tools.
OpenAI Codex vs. competitors
OpenAI Codex vs. Claude Code
Claude Code is Anthropic's coding agent for Claude models across the terminal, IDEs, desktop, and web. Codex is OpenAI's equivalent built around GPT-6 models, with an open-source CLI and cloud tasks that can be started from GitHub, Linear, or Slack. The choice usually follows which model family and subscription your team already uses.
OpenAI Codex vs. Cursor
Cursor is an AI code editor (a VS Code fork) with its own agent and cloud agents. Codex is an agent you add to your existing tools, including a Codex extension that runs inside Cursor.
OpenAI Codex vs. GitHub Copilot
GitHub Copilot combines inline autocomplete, IDE agent mode, and a cloud coding agent tied to GitHub, and lets you pick models from several vendors. Codex centers on OpenAI's models and spans the terminal, IDE, and ChatGPT apps, with GitLab (beta), Linear, and Slack as task entry points.
OpenAI Codex vs. OpenCode
OpenCode is an open-source coding agent designed to work with 75+ model providers. Codex's CLI is also open source and can point at other providers, but Codex cloud, the desktop app, and plan-based usage are tied to OpenAI.
Frequently asked questions
What is OpenAI Codex?
Codex is OpenAI's coding agent. It reads and edits code, runs commands and tests, and reviews changes through the Codex CLI, an IDE extension, Codex cloud (isolated environments for parallel tasks), and the ChatGPT desktop and web apps.
How much does Codex cost?
Codex is included in ChatGPT plans: Free ($0) and Go ($8/month) get GPT-6 Luna in the desktop app (subject to rollout), Plus ($20/month) adds the web, CLI, IDE extension, iOS, and cloud integrations such as automatic code review and Slack, Pro starts at $100/month with 5x or 20x Plus usage, and Business is $20/user/month billed annually ($25 monthly, 2+ users). Enterprise and Edu are sales-led. With an OpenAI API key, usage in the CLI, SDK, and IDE extension is billed at standard API token rates, without cloud features.
Is Codex open source?
The Codex CLI is open source under the Apache-2.0 license on GitHub (openai/codex). Codex cloud and the ChatGPT apps are proprietary OpenAI services.
Which models does Codex use?
OpenAI recommends GPT-6 Astra for complex work, GPT-6 Sol for complex coding and agentic workflows, and GPT-6 Luna for efficient, high-volume tasks. GPT-5.5 retires from Codex on October 14, 2026. Codex can also point at any model provider that supports the Chat Completions or Responses APIs.
Which editors does Codex work with?
The Codex extension runs in VS Code and compatible editors such as Cursor and Windsurf. Xcode and JetBrains IDEs provide their own Codex integrations. From the IDE you can work locally or hand a longer task to Codex cloud.
How is Codex different from ChatGPT?
ChatGPT is OpenAI's general assistant. Codex is its coding agent, which works in your repository: it edits files, runs commands, reviews diffs, and runs cloud tasks against connected GitHub or GitLab repos. Codex is available inside ChatGPT's desktop and web apps and as a standalone CLI and IDE extension.
Integrations & fit
Alternatives to consider
Anthropic
Developers and teams who want deep codebase reasoning, multi-file changes, and persistent project instructions in a terminal-friendly or IDE-native workflow
PaidAnysphere
VS Code developers who want multi-file AI editing, autonomous cloud agents, and model flexibility — not a plugin layered on top of an existing editor
Freemium
Amp
Developers and teams who want to run many background coding agents in parallel on remote machines or their own runners, and check on them from any device
FreemiumGitHub (Microsoft)
Developers on GitHub who want AI completion, chat, and agents embedded in their existing IDE and GitHub workflow
Freemium
Anomaly
Developers who want a terminal-based AI coding agent with no vendor lock-in, the ability to bring any LLM provider, and undo control over agent file changes
FreeAbout OpenAI Codex
Codex is the coding agent inside OpenAI's product line, separate from general ChatGPT chat. You can run it locally or delegate work to the cloud. Locally, the Codex CLI (Apache-2.0, on macOS, Linux, and Windows) and the IDE extension for VS Code, Cursor, and Windsurf edit files and run commands in your working tree. A /permissions setting controls when Codex may edit files or run commands without asking. Xcode and JetBrains IDEs have their own Codex integrations. In the cloud, each task gets its own isolated environment connected to GitHub or GitLab (GitLab is in beta). You can run several tasks in parallel, start them from the web, GitHub, GitLab, Linear, or Slack, and review a summary and diff before merging. The same agent also runs inside the ChatGPT desktop app on macOS, Windows, and Linux, which adds computer use (limited to some regions), a browser, and scheduled tasks. Codex supports AGENTS.md project instructions, subagents, MCP servers, skills and plugins, a dedicated code review command, non-interactive `codex exec` runs for CI, a GitHub Action, and a Codex SDK for TypeScript and Python. The recommended models are GPT-6 Astra, GPT-6 Sol, and GPT-6 Luna, and Codex can also point at any provider that supports the Chat Completions or Responses APIs. Access comes through ChatGPT plans or an OpenAI API key. On a ChatGPT plan, Free and Go only get GPT-6 Luna in the desktop app, and the web, CLI, IDE extension, and cloud integrations start at Plus ($20/month). Plan usage is measured per five-hour window (weekly limits may also apply) and depends on the model, task size, and whether work runs locally or in the cloud. An API key covers the CLI, SDK, and IDE extension at API token rates but does not include cloud features such as Codex cloud, GitHub code review, or Slack. That makes Codex easiest to justify for people already on a ChatGPT plan. Teams that want one bill across model vendors may prefer a provider-neutral agent.
Updates from OpenAI Codex
Codex CLI 0.157.0 made fullscreen transcripts the default, starts the background server automatically, added a shortcut to fork conversations, and made /import available in remote sessions. Version 0.157.1 followed on September 26.
OpenAI began rolling out GPT-6 Sol and GPT-6 Luna to Codex and ChatGPT Work at lower token prices than their GPT-5.6 predecessors. Codex CLI 0.156.0 shipped the same day with an optional fullscreen UI, voice conversations on by default, and a /usage dashboard.
Are you the founder? Claim this listing →