AI buildersOct 26, 20263 min read

Codex vs Claude Code: which one to use after Lovable

Codex and Claude Code do the same job. You open a project folder, describe a change and the agent edits files and runs commands. Codex is from OpenAI. Claude Code is from Anthropic.

vsTwo agents, one Lovable app

Either one can take over an app you built in Lovable, and the step-by-step guide has the same 6 to 19 steps for both. So the question is what you pay, how fast you hit a limit and how much the agent does without asking you.

Codex and Claude Code side by side

	Codex	Claude Code
Made by	OpenAI	Anthropic
Where it runs	The ChatGPT desktop app, the terminal, your code editor and the web	The Code tab in the Claude desktop app, the terminal, your code editor and the web
Free plan	Yes, but only GPT-6 Luna in the desktop app, subject to rollout	No. Claude Free does not include Claude Code
Cheapest plan with the terminal	Plus, $20 a month	Pro, $20 a month, or $17 a month billed yearly ($200 up front)
For heavy use	Pro, $100, $200 or $500 a month	Max, from $100 a month, with 5x or 20x the usage of Pro
Pay per use	An OpenAI API key at API rates, without the cloud features	A Claude Console account with API credits at API rates
Usage limits	A five-hour limit on Plus, and weekly limits may apply. Pro has no five-hour limit for now	A five-hour and a weekly limit on Pro and Max, shared with Claude chat
Default on your computer	Edits and runs commands in the project folder on its own, in a sandbox without network	Auto mode: a classifier checks each action instead of you
Notes for the agent	AGENTS.md	CLAUDE.md, or AGENTS.md when there is no CLAUDE.md

Prices checked on 26 October 2026, in US dollars, without tax. They change often, so check the pricing pages before you pay.

What you pay

Codex comes with every ChatGPT plan, the free one too. But on Free ($0) and Go ($8 a month) the pricing page lists only GPT-6 Luna in the desktop app, subject to rollout. Codex in the terminal, in your editor and on the web starts at Plus, $20 a month. The step-by-step guide opens the project in Codex in the terminal, so plan for Plus.

Claude Code starts at Pro, $20 a month, or $17 a month if you pay $200 for the year. There is no free option. Max starts at $100 a month and gives you 5x or 20x the usage of Pro.

At $20 a month it is a draw. What decides is what you already pay for. With ChatGPT Plus, Codex costs nothing extra. With Claude Pro, Claude Code costs nothing extra.

What happens when you hit a limit

This matters more than the monthly price. You notice it the day you run out in the middle of a migration.

Codex on Plus counts usage per five hours, and weekly limits may apply too. OpenAI gives estimates, not fixed numbers: 15 to 160 local messages per five hours with GPT-6.1 Sol and 350 to 3,000 with GPT-6 Luna. Pro plans have no five-hour limit for now. When you hit the limit during a turn, Codex can finish that turn. Then you wait, buy more credits or switch to an API key.

Claude Code on Pro and Max has a five-hour limit and a weekly limit. They are shared with Claude chat, so if you also use Claude for writing or research, it comes out of the same budget. When you hit the limit you can turn on usage credits, switch to a Console account with API credits or wait for the reset. Claude Code may finish the step it is on before it stops, but Anthropic says this is rolling out gradually.

If the ANTHROPIC_API_KEY environment variable is set on your computer, Claude Code asks whether to use that key. If you say yes, you pay API rates instead of using your subscription. Check this before you start.

How much it does without asking

Both work on their own by default. They keep you safe in different ways.

Codex locks itself in. In the default Auto preset it reads files, edits them and runs commands in the project folder on its own. The sandbox has no network and can only write in that folder. Codex asks before it edits anything outside the folder or runs a command that needs the network. In a new folder it may start read-only until you trust it. The /permissions command switches between Auto and Read Only.

Claude Code from version 2.1.283 starts in auto mode in the terminal and in VS Code. A separate model, the classifier, checks each action before it runs. By default it blocks things like production deploys and migrations, force pushes and downloading and running code. Anthropic says auto mode does not guarantee safety. Press Shift+Tab to switch to Manual mode. Then Claude Code asks before most file edits, shell commands and network requests.

For a Lovable app this is not theory. Your Supabase project is live, with real users in it. In Codex, a command that reaches that project needs the network, so it stops and asks. In Claude Code, the classifier decides unless you switch to Manual. If you want to see every step, use Read Only in Codex or Manual in Claude Code until you trust the agent.

Notes for the agent

Codex reads AGENTS.md before it starts work. The /init command creates one. By default it reads up to 32 KiB of these files combined.

Claude Code reads CLAUDE.md at the start of every session. /init writes a first version from your code. Anthropic recommends keeping each file under 200 lines. If the repository has an AGENTS.md and no CLAUDE.md, Claude Code reads the AGENTS.md instead.

So one AGENTS.md works for both agents, and your notes do not lock you in. If you wrote project knowledge in Lovable, Take your Lovable Knowledge to CLAUDE.md or AGENTS.md shows how to bring it along.

Which one writes better code?

Nobody agrees, and the answer changes with every model release. OpenAI retired GPT-5.5 from Codex with ChatGPT sign-in on 14 October 2026, so a test from last month may be about a model you cannot pick anymore.

For a typical Lovable app (React, Supabase and a few edge functions) the model is not what decides. The plan you already pay for, the limits and how much control you want decide. If you are not sure, use the one you have for a week. Your code is a normal Git repository, so switching later costs you an install and a sign-in.

Can you use both?

Yes. Both have extensions for VS Code and Cursor and both work on the same folder. Keep your notes in AGENTS.md, commit often and do not let both agents edit at the same time. If Lovable is still connected to the same repository, read Keep Lovable in sync when you edit in another tool first.

Getting started after Lovable

Connect your Lovable project to GitHub and get the repository onto your computer. Run a Lovable project on your computer shows how.
Install the agent. Codex: npm install -g @openai/codex. Claude Code on macOS or Linux: curl -fsSL https://claude.ai/install.sh | bash.
Go to the project folder and run codex or claude. Codex asks you to sign in with ChatGPT. Claude Code asks you to log in with your Claude account.
Run /init so the agent writes its notes file.

The step-by-step guides go further: database, logins, edge functions, keys, hosting and the domain. They ask a few questions about your app and show only the steps you need. If you are still deciding whether to leave Lovable at all, start with Lovable vs Codex or Should you leave Lovable for Claude Code?

Codex or Claude Code: which one fits you?

Pick Codex if you already pay for ChatGPT, you want the agent locked in a sandbox without network or you want no five-hour limit (Pro).
Pick Claude Code if you already pay for Claude and you use Claude chat lightly enough that the shared limit is not a problem. Go for Max if Pro runs out too often.