OpenAI Codex review

What OpenAI Codex is

Codex is OpenAI’s coding agent. Instead of only suggesting lines, it can read your repository, plan a change, edit files, run commands and tests in a sandbox, and propose a diff or pull request. You can work with it interactively in your terminal or IDE, or hand off tasks to run in the cloud while you do something else.

Codex now lives alongside ChatGPT’s other agentic features: ChatGPT Work and Codex share the same pricing, credits and usage limits, and the ChatGPT desktop app on Mac and Windows is a main surface for projects, files and long-running work.

Key features

Codex CLI and IDE extension

Use Codex from your terminal and scripts, or inside your editor. Configure it with config.toml and give project-specific instructions with AGENTS.md.

Codex Cloud

Run coding tasks in the cloud, in parallel, and review the results.

Code review

Understand pull requests, investigate findings and review local changes; automatic GitHub code review is available on Plus and above.

Safety controls

Sandboxing, approvals, network controls and rules for which commands Codex can run outside the sandbox.

Subagents, skills and SDK

Use subagents and custom Codex agents, build skills, and control Codex programmatically with the SDK or embed it with the App Server.

Speed modes

Fast and Ultrafast modes trade more usage for speed (Ultrafast on Pro $500).

Pricing breakdown

From OpenAI’s ChatGPT/Codex pricing docs, checked October 16, 2026.

Plan	Price	Codex access
Free	$0	Quick coding tasks (GPT-6 Luna, Standard speed, subject to rollout)
Go	$8/mo	Lightweight coding tasks
Plus	$20/mo	Web, CLI, IDE, iOS; code review and Slack; GPT-6.1 Sol and GPT-6 Luna; extend with credits
Pro	$100, $200 or $500/mo	More usage, no five-hour limit; Ultrafast on $500
Business	$20/user/mo annual (2+ users) or $25 monthly	Larger cloud VMs, admin controls, SAML SSO, no training on business data by default
Enterprise & Edu	Contact sales	Priority processing, SCIM, EKM, RBAC, analytics
API key	API rates	CLI, SDK, IDE only; no cloud features

Fast mode uses included usage at 2.5x the Standard rate; Ultrafast at 8x. OpenAI notes that API token prices are separate from subscription usage.

Best use cases

• Feature work: “add pagination to the orders API with tests.”
• Bug triage: reproduce, fix and explain an issue.
• Code review: automatic review comments on every PR.
• Migrations: upgrade dependencies or refactor modules in parallel tasks.

How to get started

1. Sign in with your ChatGPT account (or an API key).
2. Install the CLI or IDE extension, or open the desktop app.
3. Add an AGENTS.md describing build, test and style rules.
4. Start with a small, testable task and review the diff.
5. Connect GitHub to enable cloud tasks and code review.

Tips and tricks

• Be explicit about acceptance criteria and test commands.
• Keep tasks scoped—one feature or bug per task.
• Use approvals for commands that touch the network.
• Check /status to track remaining limits.

Example task:

In the /api/orders route, add cursor-based pagination (limit default 50, max 200), update the OpenAPI spec, and add unit tests. Run npm test and fix any failures.

Pros and cons

👍 Pros

• Works locally and in the cloud
• Included with ChatGPT subscriptions
• Automatic GitHub code review
• Strong sandboxing and approval controls
• Subagents for parallel work

👎 Cons

• Usage limits are hard to predict
• Heavy use needs Pro ($100+)
• Cloud features unavailable with API keys
• Rapid model changes require migrations

Who should (and shouldn't) use OpenAI Codex

✅ A good fit if you…

• Delegate well-scoped coding tasks
• Want automatic PR reviews
• Already pay for ChatGPT

⛔ Look elsewhere if you…

• Want an AI-first IDE (consider Cursor)
• Build apps without coding (consider Lovable or Replit)
• Need inline completions only (consider GitHub Copilot)

OpenAI Codex alternatives

• Claude Code — Anthropic's agentic coding tool that plans, edits, tests and ships code from your terminal, IDE, Slack or the web. (from $17/mo (Claude Pro, annual; $20 monthly))
• Cursor — AI coding agent and code editor that builds, tests and ships software from the desktop, CLI, Slack and GitHub. (from $8/mo (Plus); $80/yr)
• GitHub Copilot — AI pair programmer with code completion, chat, a cloud coding agent, code review and access to third-party agents. (from $10/user/mo (Pro))
• Devin Desktop (formerly Windsurf) — Cognition's AI IDE, the successor to Windsurf, with an agent command center for Devin and third-party coding agents. (from $20/mo (Pro))
• Replit — Build and publish apps by describing them to Replit Agent, all in the browser. (from See replit.com/pricing (price not confirmed))

FAQ

Is OpenAI Codex free?

Yes, on a limited basis. ChatGPT Free lets you explore Codex on quick coding tasks, and Go ($8/month) covers lightweight coding tasks.

How much does Codex cost?

Codex is part of ChatGPT plans: Plus is $20/month, Pro is $100, $200 or $500/month, and Business is $20 per user/month billed annually (2+ users) or $25 billed monthly. Enterprise and Edu are through sales. You can also use an API key and pay API rates.

What are Codex usage limits?

Limits depend on the model, task size and complexity, and whether tasks run locally or in the cloud. Plus and Business have five-hour windows (weekly limits may also apply); Pro currently has no five-hour limit. Check your usage dashboard or run /status in the CLI.

Where can I use Codex?

In the ChatGPT desktop app, the Codex CLI, IDE extensions, on the web and iOS, through Codex Cloud, and programmatically via the Codex SDK.

Does Codex review pull requests?

Yes. Plus and higher include cloud integrations like automatic code review on GitHub and a Slack integration. These aren't available when you use Codex with an API key.

What changed with GPT-5.5?

GPT-5.5 retired from Codex with ChatGPT sign-in on October 14, 2026. Plans now use newer models such as GPT-6.1 Sol and GPT-6 Luna; API-key sessions weren't affected.

Sources

This is an editorial review based on research of official OpenAI Codex pages and documentation, not hands-on benchmark testing. Prices are in USD and were checked on October 16, 2026; vendors change plans often, so confirm on the official pricing page before buying.

1. Codex pricing (ChatGPT docs)
2. Models (ChatGPT docs)
3. ChatGPT and Codex docs index