Home/Articles
September 12, 2026

Claude Code vs. Codex vs. Grok Build: Which One for Non-Developers?

Three companies now sell a coding agent you can talk to instead of a text editor you have to learn: Anthropic's Claude Code, OpenAI's Codex, and xAI's Grok Build. For a non-developer, the differences that matter are not benchmark scores. They are whether you need a terminal, which account bills you, and what happens the first time you try it. This comparison sticks to facts documented in the Pick Your Agent course's five walkthroughs, one per tool and surface.

The one thing all three share

Every path in the Pick Your Agent course sends the same three prompts, on purpose, so you can compare honestly instead of guessing:

Prompt 1: Explain what is in this folder in plain English. Do not change anything. Prompt 2: Create a file named hello.txt in this folder. The file should contain the sentence: I am set up. Prompt 3: Add a second line to hello.txt that says: I can change files too. Show me the change before you save it.

Run the same three prompts against two different tools in the same empty test folder, and you are looking at the same work, not two different demos.

Claude Code: a terminal agent with a graphical option

Claude Code is Anthropic's coding agent. It has two surfaces built by the same company: a desktop app with a Code tab, a visual diff, and Accept or Reject buttons, and a terminal version you install with one pasted command. Both require a Claude Pro, Max, Team, Enterprise, or Console account; the free Claude.ai plan does not include Claude Code access.

The desktop app's Manual permission mode is the safest place for a first-timer to start: Claude proposes each change and waits for approval before touching a file. The terminal version adds something neither of the others documents today: /voice, which lets you speak your prompts instead of typing them, with a documented push-to-talk and tap-to-talk mode. The two surfaces share configuration, so moving from one to the other later costs nothing.

Codex: lives inside the ChatGPT app you may already have

Codex is not a separate download. It lives inside the ChatGPT desktop app, reached by toggling from Chat to Work, and it also ships as a terminal program you install with one command. Both surfaces bill against whichever ChatGPT plan you sign in with; OpenAI's pricing page lists Codex access across Free, Go, Plus, Pro, Business, Enterprise, and Edu, with usage rising by plan.

If you already pay for ChatGPT, the desktop app is the fastest path to a working coding agent of the three, because you are not creating a new account or judging a new subscription on its own. Codex's desktop app is also the only one of the three with documented voice conversation (ChatGPT Voice, on Plus and above), which is different from Claude Code's voice dictation: one is a live back-and-forth conversation, the other transcribes your speech into a prompt before sending it.

Grok Build: a terminal-only agent, honestly labeled

Grok Build is xAI's coding agent, described by xAI as "an early beta." As of this writing, it has one shape: a terminal program. There is no Grok Build desktop app and no web surface documented on xAI's own pages, and any installer claiming to be one on another site should be treated as unverified. It requires a SuperGrok or X Premium Plus subscription.

If a graphical window with Accept and Reject buttons is what makes you comfortable, Grok Build is not that today. What it does offer that is worth knowing: it reads Claude Code's own project setup, including CLAUDE.md and its skills and MCP servers, with what xAI's own docs describe as "zero configuration needed," so a project you already set up for Claude Code is not wasted if you try Grok Build against it.

What actually differs, in plain terms

  • No-terminal option: Claude Code and Codex both have one. Grok Build does not, as of this writing.
  • Voice: Claude Code's terminal has /voice dictation. Codex's desktop app has ChatGPT Voice, a live conversation. Grok Build documents neither.
  • Billing: each one bills against the subscription you sign in with (Claude Pro and above, a ChatGPT plan, or SuperGrok/X Premium Plus), or separately against an API key if you choose that route instead. Those are two different bills; pick on purpose.
  • The instructions file: Codex and Grok Build both read a plain AGENTS.md file with no setup. Claude Code reads CLAUDE.md instead. Writing one file covers two of the three tools automatically.

How to actually decide

Match the tool to a subscription you already pay for first. That single filter removes two of the three options for most people before any feature comparison matters. If you pay for neither ChatGPT nor Claude yet, start with whichever desktop app's free trial or entry plan is cheaper for what you need, run the three-prompt test, and judge it on the folder in front of you rather than on a comparison chart, this one included.

Every one of the five full walkthroughs, install commands and all, is free to run: start with Path A if you want the no-terminal Claude Code route, or Path C for the ChatGPT desktop route.

Keep learning
Session: Path A: Claude Code in the desktop appSession: Path B: Claude Code in the terminalSession: Path C: Codex in the ChatGPT desktop appSession: Path D: Codex in the terminalSession: Path E: Grok Build in the terminal (the only surface today)Course: Pick Your Agent
Frequently asked
Which one should a non-developer install first?

Whichever one matches a subscription you already pay for. If you already pay for ChatGPT, the ChatGPT desktop app's Codex surface is the shortest path: no terminal, and the account is one you already have. If you already pay for Claude, the Claude Code desktop app is the equivalent no-terminal path.

Do I have to use a terminal to use any of these?

No. Claude Code and Codex both ship a graphical desktop app with a visual diff and Accept or Reject buttons, so a non-developer never has to open a terminal. Grok Build is the exception: as of this writing it only ships as a terminal program, with no desktop app or web surface documented.

Can I run more than one of these agents?

Yes, and comparing two head to head is exactly what the Pick Your Agent walkthroughs are built for. Each one uses the same three first prompts, so running two back to back on the same test folder shows you the real difference in behavior rather than a marketing claim.

Want the full walkthrough, not just the guide?

The free first session in the library walks through the whole setup on screen. Or see what All-Access unlocks on the pricing page.

Start the free first sessionSee pricing