Both cost $20/mo. The difference is what they do.
Anthropic's terminal coding agent
OpenAI's coding agent: CLI + autonomous cloud PRs
They are closer than ever: both run in the terminal, an IDE extension, a desktop app and the cloud, and both ship subagents, MCP, hooks and skills. Pick by the plan you already pay for and the models you trust. Codex comes with every ChatGPT plan, including Free, and defaults to GPT-5.6 Sol. Claude Code needs Claude Pro or higher and defaults to Sonnet 5 on Pro and Opus 5 on Max.
| Attribute | Claude Code | Codex |
|---|---|---|
| Vendor | Anthropic | OpenAI |
| Cheapest plan that includes it | Claude Pro, $20/mo or $17/mo billed annually; not on Free | ChatGPT Free, then $20/mo ChatGPT Plus |
| Top individual plan | Max 20x, $200/mo | Pro 20x, $200/mo |
| Team seats | Team Standard $25/mo, Premium $125/mo ($20 and $100 billed annually) | Business $25/user/mo, $20 billed annually |
| Default model | Sonnet 5 on Pro and Team Standard; Opus 5 on Max, Team Premium, Enterprise and the API | GPT-5.6 Sol at medium reasoning (the default Power setting) |
| Other models | Fable 5.1 and Haiku, plus Opus and Sonnet by alias | GPT-6 Astra, GPT-5.6 Terra and Luna; GPT-5.3-Codex-Spark on Pro |
| Subagents | Yes: custom subagents in their own context, run in the background by default | Yes: on by default, custom agents in TOML files, and an Ultra mode that splits work across them |
| Extend it with | MCP, hooks, skills, plugins, CLAUDE.md | MCP, hooks, skills, plugins, AGENTS.md |
| Where it runs | Terminal, VS Code and JetBrains, desktop app, web, the Claude mobile app, Slack, GitHub Actions and GitLab CI/CD | CLI, IDE extension, ChatGPT desktop app, web and cloud, iOS, Slack, Linear and a GitHub Action |
| Usage limits | Shared with Claude chat; a session limit that resets every 5 hours plus a weekly limit | Messages per 5-hour window that vary by model; local and cloud work share the allowance, and weekly limits may apply |
They are closer than ever: both run in the terminal, an IDE extension, a desktop app and the cloud, and both ship subagents, MCP, hooks and skills. Pick by the plan you already pay for and the models you trust. Codex comes with every ChatGPT plan, including Free, and defaults to GPT-5.6 Sol. Claude Code needs Claude Pro or higher and defaults to Sonnet 5 on Pro and Opus 5 on Max.
Codex has the lower floor. It is included in ChatGPT Free, paid ChatGPT plans start at $20/mo Plus, and OpenAI currently includes it in the $8/mo ChatGPT Go plan for a limited time. Claude Code is not on Claude's Free plan and starts with Claude Pro at $20/mo, or $17/mo billed annually. At the top both individual plans reach $200/mo, Max 20x for Claude and Pro 20x for ChatGPT. For teams, Claude Team Standard and ChatGPT Business both cost $25 per seat billed monthly, or $20 billed annually.
Claude Code defaults to Sonnet 5 on Pro and Team Standard seats and to Opus 5 on Max, Team Premium, Enterprise and the Anthropic API, with Fable 5.1 available but never the default. Codex's default Power setting runs GPT-5.6 Sol at medium reasoning, and you can pick GPT-6 Astra, GPT-5.6 Terra or GPT-5.6 Luna, plus GPT-5.3-Codex-Spark as a research preview on Pro.
Yes. Current Codex releases enable subagent workflows by default, you can define custom agents as TOML files, and Ultra mode splits a complex task across subagents in parallel. OpenAI notes that subagent workflows use more tokens than single-agent runs. Claude Code runs custom subagents in their own context windows with their own tools and permissions, in the background by default.
Claude Code shares its limits with Claude chat: Pro and Max have a session limit that resets every five hours plus a weekly limit, and usage credits let you keep going past it. Codex counts messages per five-hour window, and the number depends on the model: OpenAI estimates 10 to 100 GPT-5.6 Sol local messages per window on Plus and 200 to 2,000 on Pro 20x. Local and cloud work share the same allowance, weekly limits may also apply, and Plus and Pro users can buy extra credits.
Anthropic's terminal coding agent. It sits in the coding agents category, and ships in 4 tested stacks here.
OpenAI's coding agent: CLI + autonomous cloud PRs. It sits in the coding agents category, and ships in 1 tested stack here.
They are alternatives at the same step, so most stacks pick one or the other. No aistack.sh stack uses both today, though nothing prevents a custom workflow.
Beyond each other, Claude Code's closest alternatives are Cursor. Codex's closest alternatives are Cursor, GitHub Copilot, Windsurf.
