Home › Rules › Antigravity vs Claude Code vs Codex

Google Antigravity vs Claude Code vs Codex: which one should a beginner use?

Prices, plans and limits from Google's, Anthropic's and OpenAI's own pricing pages and docs, checked Nov 3, 2026. Sources below.

• Comparison
• Antigravity
• Claude Code
• Codex
• Beginners
• Pricing

Short answer: all three are AI coding agents that read your project, change files and run commands for you. Google Antigravity has a $0 plan with weekly limits, and you sign in with a regular Google account. It comes as a desktop app, a command-line tool and editor add-ons. Claude Code has no free plan; it comes with Claude Pro ($20 a month). OpenAI Codex comes with ChatGPT; OpenAI lists it on the Free and Go plans in the desktop app ("subject to rollout"), and Plus ($20) adds the command line, editor add-on and web. Already paying Google, Anthropic or OpenAI? Try the tool that comes with that plan first. All three read an AGENTS.md rules file, so you can switch later without rewriting your rules.

Comparing Cursor or GitHub Copilot instead? See Cursor vs Claude Code vs Codex and GitHub Copilot vs Cursor vs Claude Code. This page swaps in Google Antigravity.

Side-by-side breakdown

Prices are per month in US dollars, before tax. Google lists the Google AI plan prices on its Google One plans page; the Antigravity pricing page names the plans but doesn't show their prices. All plans have usage limits; see usage limits.

Which should I pick?

Pick Google Antigravity if

• You want to start for $0. Google's plans page lists "access to all product features, such as scheduled tasks and the CLI" on every plan (two extra commands, /boost and /teamwork, need a paid plan). The $0 plan's limit refreshes weekly, so you may hit it and have to wait.
• You like seeing a plan before anything changes. Antigravity's /plan command has the agent look around your project, ask questions, and write an Implementation Plan you can comment on before you click Proceed.
• You already pay for Google AI Pro or Ultra, for example for Gemini or the extra Google storage. Those plans raise your Antigravity limits at no extra cost.
• You're building a website and want the agent to test it. Antigravity has a browser helper (/browser) that drives Google Chrome to test and debug pages, and can record the session as a video.

Pick Claude Code if

• You already pay for Claude Pro, Max or Ultra. Claude Code is included in every paid Claude plan and shares the same usage pool as your chats.
• You want to use it inside the editor you already have. Claude Code has an add-on for VS Code and Cursor, and a JetBrains plugin, plus a desktop app if you'd rather not use a terminal.
• You want to customize a lot. Anthropic's docs cover CLAUDE.md rules, skills, hooks (commands that run before or after the agent acts) and MCP connections to other apps.

Pick Codex if

• You already pay for ChatGPT Plus or Pro. Codex is included in the desktop app, the CLI, the editor add-on, the web and iOS.
• You want a $0 or $8 way in. OpenAI lists Codex in the desktop app on Free and Go, "subject to rollout", so it may not show up for everyone yet.
• You want tasks to run in the cloud or automatic code review on GitHub. OpenAI lists "cloud-based integrations like automatic code review and Slack integration" on Plus and up.

Still not sure? Antigravity and Codex both have $0 options, so you can try both on the same small project. Give each one the same AGENTS.md (below) and the same first task, then keep the one whose answers you understand best.

Usage limits, in plain words

None of the three gives you a fixed number of messages. How far your plan goes depends on how big the job is.

• Antigravity: on the $0 plan you get "a meaningful quota, refreshed weekly." Google AI Pro gets a quota that refreshes every five hours until you hit a weekly limit. Ultra gets the highest quota, also every five hours, and the highest weekly limit. Google says limits follow "the amount of work done by the agent," so simple tasks go further. On Pro and Ultra you can buy AI credits for extra use; a setting called AI Credit Overages decides whether they get used (Never or Always). Google says bring-your-own-key for extra limits isn't supported, but the CLI can run on a Gemini API key instead of a Google sign-in (requests then go straight to the Gemini API).
• Claude Code: it uses your Claude plan's usage, shared with your chats. Anthropic says most accounts have a limit that resets on a rolling five-hour window, and paid plans add weekly limits. Max gives 5x Pro's usage every five hours and every week; Ultra gives 10 times Pro's weekly usage with no five-hour limit for verified accounts. When you run out, you can wait, upgrade, or turn on usage credits at API rates. You can check in Settings > Usage. More detail: Claude Pro vs Max vs Ultra.
• Codex: OpenAI says it depends on the model, the size of the task, and whether it runs on your computer or in the cloud. Plus is counted per five-hour window, with weekly limits that may also apply. OpenAI says Pro plans currently have no five-hour limit. Plus and Pro users can buy extra credits instead of upgrading, and anyone can run local chats with an API key at API prices. In the CLI, /status shows what you have left.

Each company says its limits can change. Check your usage screen in the first week so you know your pace.

Which AI models you get

Antigravity is the only one of the three whose model picker includes another company's models. Google's model table (checked Nov 3, 2026) lists Gemini 3.8 Flash on every plan, including $0. Gemini 4 Argon needs Google AI Pro or Ultra. Claude Sonnet 5.5 and Claude Opus 5.5 are listed on Ultra and on paid Pro (not on a Pro trial). On $0, the table also lists Claude Sonnet 4.6, Claude Opus 4.6 and GPT-OSS-120b, but says those "will be removed on November 18, 2026." One thing to know: Google's plans page lists "access to third-party models" only under Ultra, while the model table shows Claude 5.5 on paid Pro too. If Claude inside Antigravity is the reason you'd pay, check the model picker before you upgrade.

Claude Code uses Anthropic's Claude models, and Pro adds "more Claude models" over Free. Codex uses OpenAI's models: Free and Go list GPT-6 Luna in the desktop app, and Plus lists GPT-6.1 Sol and GPT-6 Luna.

How rules and instructions files map across the three

A rules file is a plain text file in your project that tells the agent how you like to work. It's read at the start of every task, so you don't have to repeat yourself. Here's where each tool looks:

So one AGENTS.md in your project root works in all three. The catch is Claude Code: if your project has both a CLAUDE.md and an AGENTS.md, Anthropic's docs say Claude reads "your CLAUDE.md files only." The fix is to put @AGENTS.md at the top of your CLAUDE.md, which pulls the shared rules in. Codex stops adding files once they reach 32 KiB in total by default, so keep the file short.

Here's a starter you can copy. Fill in the brackets:

# AGENTS.md

## About this project
- A small web app for [who it's for] that lets them [the main thing it does].
- Built with [your stack, for example Next.js + Supabase].

## How to work
- Before you change anything, tell me your plan in plain English.
- Change one thing at a time, then tell me how to check it.
- Ask me before adding a new package or deleting any file.
- Never put API keys or passwords in the code. Use environment variables.

## Commands
- Install: npm install
- Run locally: npm run dev
- Tests: npm test

And if you also want Claude-only rules, this small CLAUDE.md keeps everything in one place:

@AGENTS.md

## Claude Code only
- Use plan mode for any change that touches more than 3 files.

Skills work the same way: a folder with a SKILL.md file. Antigravity and Codex both read .agents/skills/; Claude Code wants .claude/skills/. Our SKILL.md examples for every tool page shows how to share one skill between them. For deeper guides, see AGENTS.md examples, CLAUDE.md examples, Google Antigravity rules (triggers and the .agents/rules/ folder) and Google Antigravity skills.

Asking before it acts

All three can run commands on your computer, so it's worth knowing what they do by default. Antigravity's App asks your permission before running terminal commands outside its sandbox, and its CLI asks before it writes files, runs commands or uses web tools, unless you've allowed that action. Claude Code and Codex have their own permission settings and a plan step you can turn on; our plan mode guide shows how in each tool. Whatever you pick, keep your project in Git (or a copy) so you can undo a bad change.

What about Cursor?

Cursor is a full code editor with an AI agent built in. It has a $0 plan with "limited agent requests," and Plus is $8 a month for Grok models only (checked Nov 3, 2026). It reads AGENTS.md too, and Codex and Claude Code both have add-ons that run inside it. If you want a normal editor window with your files on the left, compare it in Cursor vs Claude Code vs Codex.

Never coded? A browser builder may fit better

All three tools here work on code in a folder on your computer (or in the cloud for Codex). If you'd rather describe an app and watch it appear in your browser, with hosting included, start with a builder like Lovable, Bolt, Base44 or Replit. We compared them with real prices in Lovable vs Bolt vs Replit and Base44 vs Lovable. You can move the code to one of these agents later.

Related

• Cursor vs Claude Code vs Codex (real prices)
• GitHub Copilot vs Cursor vs Claude Code (real prices)
• Claude Pro vs Max vs Ultra (which plan for Claude Code)
• Google Antigravity rules (examples, triggers, where the files go)
• Google Antigravity skills (skills folder, workflows vs skills)
• SKILL.md examples for every tool
• AGENTS.md vs CLAUDE.md (which file each tool reads)
• AGENTS.md examples
• CLAUDE.md examples
• Plan mode guide (6 tools + 6 plan prompts)
• Free rules file generator

Sources

• Google Antigravity: Pricing
• Google One: Google AI plans and pricing
• Google Antigravity docs: Plans
• Google Antigravity docs: Choose a model
• Google Antigravity docs: Choose your Antigravity
• Google Antigravity docs: Get started
• Google Antigravity docs: Authenticate
• Google Antigravity docs: Features
• Google Antigravity docs: Plan (/plan)
• Google Antigravity docs: Rules
• Google Antigravity docs: Agent skills
• Claude: Pricing
• Claude Code docs: Overview
• Claude Code docs: Memory (CLAUDE.md and AGENTS.md)
• Claude Code docs: Skills
• OpenAI docs: Codex pricing
• OpenAI docs: Codex CLI
• OpenAI docs: Codex IDE extension
• OpenAI docs: Custom instructions with AGENTS.md
• OpenAI docs: Build skills
• Cursor: Pricing
• Cursor docs: Rules

All prices, plans, limits and file locations checked Nov 3, 2026 on each company's own pricing and docs pages. Prices are in US dollars, before tax, and change often. If something here no longer matches, tell us at [email protected].