The Decoder / One term

Claude Code vs Cursor vs Codex: which one for a non-coder?

Claude Code and Codex are agents that run in a terminal and work on the files in a folder; Cursor is a code editor with an agent inside. For a non-coder, the terminal agents hide the code and the editor shows it.

By Mstimaj Numbers checked 30 September 2026 Updated 30 September 2026 Level Technical Tool Claude Code

Sign in to save your place and mark this done. Free, no card.

Three tools laid side by side on a dark workbench: a plain terminal window, a window full of visible code, and a second terminal

All three will build a small program from a plain description. The difference is what you look at while they work. Claude Code and Codex live in a terminal: you type a request, they edit the files and report back, and you never open the code. Cursor is a code editor with the same kind of agent in a side panel, so the code is always on screen. Pick by what you already pay for.

Key facts

  • $20per month for the cheapest paid route on each: Claude Pro, ChatGPT Plus, Cursor Pro
  • $0for Codex on a free ChatGPT account, with limited usage; Claude Code has no free route; Cursor's free tier is limited
  • 3shells Claude Code runs in natively on Windows: PowerShell, Command Prompt, Bash
  • WSLwhat OpenAI's own page still recommends for Codex on Windows, calling native support experimental
  • Anthropic Claude Code setup and costs pages; claude.com pricing; OpenAI Codex pricing and CLI pages; Cursor pricing and account docs; ChatGPT plan screen; checked 30 September 2026

What does each one actually do?

Claude Code is Anthropic’s agent. You open a terminal in a folder, describe the change, and it reads, edits, runs and checks, asking before anything risky. It also runs inside VS Code and as a desktop app, but the terminal is home. Codex is OpenAI’s equivalent: a terminal agent, a desktop app, and a cloud mode where the work happens on OpenAI’s servers and comes back as a finished change. Cursor is an editor, a modified VS Code, with an agent panel that can make the same kind of multi file changes while the code stays visible and editable by you.

Which one is free, and what do they cost?

Claude Code is the only one with no free route: it needs Claude Pro at $20 a month, Max from $100, or a pay per token API key. Codex comes with every ChatGPT plan, including free, with usage that scales up the tiers; Plus at $20 a month is the practical floor, and Pro runs $100, $200 or $500 depending on the usage tier you pick. Cursor has a free Hobby tier with limited agent requests, then Pro at $20, Pro Plus at $60 and Ultra at $200, and heavy use is billed on top from a monthly pool. If you already pay for Claude or ChatGPT, the matching agent costs you nothing extra.

The three tools compared, 30 September 2026
Question Claude Code Cursor Codex
Made by Anthropic Anysphere OpenAI
Where it runs Terminal, VS Code, desktop app Its own editor Terminal, app, cloud
Do you see the code? Only if you ask Always Only if you ask
Free route None Hobby tier, limited ChatGPT Free, limited
Cheapest paid Pro, $20 / month Pro, $20 / month Plus, $20 / month
Heavy use tier Max, $100 to $200 Pro Plus $60, Ultra $200 Pro, $100 to $500
Windows without WSL Yes, native Yes, native app Runs, but OpenAI recommends WSL
Standing rules file CLAUDE.md Rules AGENTS.md
Cheapest monthly price with real usage CHEAPEST MONTHLY PRICE WITH REAL USAGE Codex on ChatGPT Free $0 USD Cursor Hobby $0 USD Claude Code (Pro) $20 USD Cursor Pro $20 USD Codex (Plus) $20 USD
Vendor prices. The $20 floor is the same on all three; the differences are in what is free and what heavy use costs.

What are the usage limits?

None of the three sells unlimited use at $20. Claude Code shares the Pro plan’s five hour and weekly limits with the chat app; Anthropic states them as multiples of Pro rather than hours. Codex on Plus is published as a range of messages per five hours that depends on the model, from a few dozen for the strongest to a few thousand for the lightest, and OpenAI says Pro plans currently have no five hour limit. Cursor gives each plan a monthly pool and charges beyond it. For a beginner building one thing a week, none of the $20 tiers will be the bottleneck. For all day use, the $100 to $200 tiers exist on all three.

Which one on Windows?

Claude Code runs natively on Windows 10 and 11 in PowerShell, Command Prompt or Bash, with a one line installer and no WSL. Cursor is a normal Windows app. Codex is the one to check: OpenAI’s CLI page still lists Windows as experimental with WSL recommended, even though it runs in PowerShell in practice. If avoiding WSL matters to you, that decides it.

So which one for someone who does not code?

If you pay for Claude, Claude Code. If you pay for ChatGPT, Codex. If you pay for neither and want to try before spending, Codex on the free tier or Cursor’s free tier will show you the shape of the work. Choose Cursor over the other two only when you want to watch the code change and start reading it, which is the path from vibe coding to knowing what you built. Whatever you pick, the first thing to write is a plain text file of standing rules for it: CLAUDE.md for Claude Code, AGENTS.md for Codex, rules for Cursor. That file matters more than the tool.

Ask what did not make sense

Comments go through our own form and are read before they appear. Links are not allowed; describe it in words.

Comments are read before they appear.

The newsletter

Get the next term.

One email when something goes up. The argument, the film and the term, nothing else. One click to leave.

Have a question now? Ask Mstimaj, she knows the archive.