# How do Codex and Claude Code compare, and can you use both?

AutoRouter · Oct 23, 2026 · 9 min read

As of October 2026, Codex is OpenAI's coding agent and Claude Code is Anthropic's, and both run in a terminal, an editor, a desktop app and the cloud. Codex comes with every ChatGPT plan, and Plus at $20 a month adds GPT-6.1 Sol, the model OpenAI recommends for complex coding. Claude Code comes with every paid Claude plan from Pro at $20 a month, and its default model is Claude Opus 5.5. Which one suits you depends on the models you prefer and the plan you already pay for, and the two can work side by side in one repository. This post is by AutoRouter, the Mac app for Codex and Claude Code, which works in both, so read the comparison with that in mind.

## What are Codex and Claude Code?

Codex is OpenAI's agentic coding tool and Claude Code is Anthropic's, and each runs on several surfaces whose local versions share one configuration.

• - Codex runs in the ChatGPT desktop app for macOS, Windows and Linux, where you select Codex from the ChatGPT dropdown. It also runs in the Codex CLI, an IDE extension for VS Code, Cursor and Windsurf, and Codex Cloud. The CLI is open source.
• - Claude Code runs as a CLI on macOS, Linux and Windows, in the Code tab of the Claude desktop app for macOS and Windows (with a Linux beta), in extensions for VS Code and JetBrains IDEs, and on the web at claude.ai/code, where tasks keep going after you disconnect. Anthropic calls the CLI the most complete surface for terminal-native work, and scripting and the Agent SDK are CLI-only.

Both connect to outside tools through MCP, run subagents, load skills, review code, run tasks in the cloud and offer an SDK (Codex SDK, Agent SDK).

## How do Codex and Claude Code compare?

ChatGPT Plus and Claude Pro both cost $20 a month, and the two tools differ on models, higher tiers and how limits work. The figures come from each maker's docs, checked October 22, 2026.

## Which models does each one use, and what do they cost?

Codex has no single named default: with no model set in config.toml, it uses a recommended model, and OpenAI recommends GPT-6.1 Sol for complex coding and GPT-6 Luna for focused, repeatable tasks. Claude Code starts on Claude Opus 5.5, and Anthropic's model guide describes Sonnet as the model for daily coding tasks and Haiku for simple tasks. In both, /model switches models during a session and --model sets one at launch.

The spread inside each lineup is wide. For input and output tokens, GPT-6 Luna costs a twentieth of GPT-6.1 Sol in Codex credits, and Claude Haiku 5.5 costs a fortieth of Claude Opus 5.5 at API prices on prompts up to 100,000 tokens. OpenAI's estimates for Plus show a gap of about 20 times in plan usage too: 15 to 160 local messages per five hours on GPT-6.1 Sol, against 350 to 3,000 on GPT-6 Luna. Anthropic publishes no message counts, but its pricing page says usage depends on the model you choose.

So on either plan, the model you run decides much of how far the plan goes. The posts on Codex usage limits and Claude Code usage limits go through each plan in detail.

## Which one should you choose?

Choose by the plan you already pay for and the models you want to work with, because the two tools overlap on most features. Codex fits if you pay for ChatGPT, want OpenAI's models or want a CLI whose source you can read, and its Pro plans currently have no five-hour limit. Claude Code fits if you pay for Claude or want Claude models, and Claude Opus 5.5, its default, comes with the $20 Pro plan.

Usage surveys show both in wide use. In JetBrains' Developer Ecosystem Survey 2026 of more than 15,000 professional developers, about 39% used Claude Code at work in May to July 2026, and 16% used Codex, up from 3% in January. In The Pragmatic Engineer's 2026 survey of 906 respondents, published March 3, 2026, 70% used between two and four AI tools at the same time.

## How do you use Codex and Claude Code together?

Run both in the same repository and give them one set of instructions. Codex reads AGENTS.md before doing any work, and Claude Code v2.1.277 or later reads AGENTS.md when the repository has no CLAUDE.md. If you keep a CLAUDE.md, import the shared file at its top and add anything only Claude Code needs below it:

```markdown
@AGENTS.md

## Claude Code only

Run `pnpm test` before you finish a task.
```

With one instruction file, you can split the work in a few ways:

• - Send each task to the app with the model you want for it, or to whichever plan still has room under its limits.
• - Review one tool's work with the other. Codex's /review reads a diff and reports findings without changing your working tree, and Claude Code's /code-review reviews your branch on any plan.
• - Hand long tasks to the cloud. A Codex Cloud task keeps working while your computer is asleep, and Claude Code on the web keeps going after you disconnect.

## How does AutoRouter work in Codex and Claude Code?

AutoRouter, the Mac app for Codex and Claude Code, adds 20+ models to both apps, plus one more choice, AutoRouter, which picks a model for each turn by how hard the task is. It works with the Codex app, the Codex CLI and the Claude Code CLI on macOS 14 or later (Claude Code needs Node 22.18 or later), and it is in beta. You pick AutoRouter from the model menu in Codex or with /model in Claude Code, and choosing a specific model sends turns straight to it.

In Codex, OpenAI models run on your ChatGPT plan, and in Claude Code, Claude models run on your Claude plan. Among the models that clear the quality bar for a task, hard work goes to the most capable one your plans cover, and everything else goes to the one that costs you least, which is often your plan or a low-cost open model. The landing page's examples are illustrations: a change across six files goes to Claude Opus 5.5 on the Claude plan, a test that only fails on CI goes to GPT-6 Astra on the ChatGPT plan, and a commit message goes to a low-cost open model so the plan isn't touched. Every turn shows its model, what it cost and why, and you can add rules in plain English, such as sending UI prompts to frontier models. How AutoRouter picks a model covers the full order.

AutoRouter Free costs $0 a month and routes on your ChatGPT and Claude plans. AutoRouter Pro costs $20 a month with $20 of included usage each month and adds every other model, billed at the provider's price plus 10%.

## Questions

### Is Claude Code included in the Claude Free plan?

No. As of October 2026, Anthropic's pricing page includes Claude Code in every paid plan, and the Free plan doesn't include it, so the lowest-priced way in is Pro at $20 a month, or $17 a month billed annually. OpenAI lists Codex on every ChatGPT plan, though Free and Go get only GPT-6 Luna in the desktop app, subject to rollout.

### Can you use Codex or Claude Code without a subscription?

Yes, both accept pay-as-you-go billing. Codex with an API key charges usage at API pricing, offers the models your key can use and leaves out cloud features such as GitHub code review and Slack. Claude Code with an Anthropic Console account bills at standard API rates. AutoRouter Pro also works without a ChatGPT or Claude subscription and gives every model, with $20 of included usage each month.

### Which apps does AutoRouter work with?

AutoRouter works with the Codex app, the Codex CLI and the Claude Code CLI on macOS 14 or later, as of October 2026. More apps are on the way.

## Sources

• - OpenAI, Codex pricing and usage limits, checked October 22, 2026
• - OpenAI, Codex models
• - OpenAI, Quickstart
• - OpenAI, Codex CLI
• - OpenAI, Codex IDE extension
• - OpenAI, Codex Cloud
• - OpenAI, Open source
• - OpenAI, AGENTS.md
• - OpenAI, Code review
• - Anthropic, Claude Code overview
• - Anthropic, Platforms and integrations
• - Anthropic, Model configuration
• - Anthropic, How Claude remembers your project
• - Anthropic, Code Review
• - Anthropic, Claude pricing, checked October 22, 2026
• - Anthropic, What is the Max plan?
• - Anthropic, Use Claude Code with your Pro or Max plan
• - Anthropic, API pricing, checked October 22, 2026
• - JetBrains, AI Coding Agents: Adoption Trends, August 2026
• - The Pragmatic Engineer, AI Tooling for Software Engineers in 2026, March 3, 2026