# Claude Code vs Codex vs Antigravity: which fits your work? > Claude Code, Codex and Google Antigravity compared on plans, models and daily workflow, plus how to run Claude Code and Codex on one repository. Updated: 2026-11-03 Claude Code, Codex and Google Antigravity are coding agents from Anthropic, OpenAI and Google. Each is built around its maker's models, so the right one follows the models you prefer and the plan you pay for, not a single winner. One 5abeer Bundle plan runs all three, so you can test each yourself. ## How do Claude Code, Codex and Antigravity compare? All three read and change your code to finish a task you describe. This table follows each maker's own docs, checked on 2026-11-03: | | Claude Code | Codex | Google Antigravity | | --- | --- | --- | --- | | Made by | Anthropic | OpenAI | Google | | Models with 5abeer | Claude | GPT | Gemini | | Where it runs | Terminal, IDE extensions, desktop app, browser | Terminal, IDE extension, ChatGPT desktop app, Codex cloud | Desktop app, terminal (`agy`), IDE extensions | | Official sign-in | A paid Claude plan or an Anthropic Console account | Any ChatGPT plan, or an OpenAI API key | A Google Account, Gemini Enterprise or a Gemini API key | Claude Code also accepts another API endpoint through `ANTHROPIC_BASE_URL`, and Codex another model provider in `config.toml`. That is how both run on a 5abeer key. ## Which plan does each one need? On the makers' own plans, as their pages read on 2026-11-03: - **Claude Code:** Anthropic includes it in Claude Pro, Max and Ultra, not in its no-cost plan. An Anthropic Console account works too. - **Codex:** OpenAI includes it in every ChatGPT plan, the no-cost one too. Usage depends on the model, the task and whether it runs locally or in the cloud. Extra local chats can run on an API key, at standard API rates. - **Google Antigravity:** Google lists an Individual level at no charge, and its Google AI Pro and Ultra plans raise the limits. The official OpenAI and Anthropic prices, each row with its source and the date we checked it: | Plan | Market and channel | Official price | Billing | Source | Checked | | --- | --- | --- | --- | --- | --- | | ChatGPT Go | United States · Vendor website | $8 | per month | https://openai.com/index/introducing-chatgpt-go/ | 2026-10-31 | | ChatGPT Plus | United States · Vendor website | $20 | per month | https://help.openai.com/en/articles/6950777-what-is-chatgpt-plus | 2026-10-31 | | ChatGPT Pro 100 | United States · Vendor website | $100 | per month | https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers | 2026-10-31 | | ChatGPT Pro 200 | United States · Vendor website | $200 | per month | https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers | 2026-10-31 | | ChatGPT Pro 500 | United States · Vendor website | $500 | per month | https://help.openai.com/en/articles/9793128-about-chatgpt-pro-tiers | 2026-10-31 | | Claude Pro | United States · Vendor website | $20 | per month | https://claude.com/pricing | 2026-10-31 | | Claude Pro | United States · Vendor website | $200 | per year | https://claude.com/pricing | 2026-10-31 | | Claude Max | United States · Vendor website | $100 | per month | https://claude.com/pricing | 2026-10-31 | | Claude Ultra | United States · Vendor website | $200 | per month | https://claude.com/pricing | 2026-10-31 | The vendors' own prices as published at the source on the date checked; they can change. Tax: per Anthropic's pages, some countries include tax in the displayed price and others add it at checkout. Google's prices are in [Google Antigravity explained](https://5abeer.ai/en/guides/google-antigravity-guide#price). For paid limits in more than one tool, you pay each maker separately. ## Which models does each one run? Claude Code runs Claude models and Codex runs GPT models. Antigravity is built on Gemini, though Google's own picker also offers some Claude models and OpenAI's GPT-OSS. In Claude Code and in Codex, you switch models during a session with `/model`, or when you start the tool with `--model`. These are the newest models in each, at our price per million tokens: | Model | Id | Input (ours) | Output (ours) | Context | | --- | --- | --- | --- | --- | | Claude Sonnet 5.5 | `claude-sonnet-5.5` | $0.60 | $3 | 1M | | Claude Opus 5.5 | `claude-opus-5.5` | $1.20 | $6 | 1M | | Claude Fable 5.1 | `claude-fable-5.1` | $3 | $15 | — | | Model | Id | Input (ours) | Output (ours) | Context | | --- | --- | --- | --- | --- | | GPT-6.1 Sol | `gpt-6.1-sol` | $0.60 | $3 | 922K | | GPT-6 Astra | `gpt-6-astra` | $3 | $15 | — | | GPT-6 Sol | `gpt-6-sol` | $0.60 | $3 | 922K | With 5abeer, Google Antigravity runs Gemini models, listed on [Models that run in Google Antigravity](https://5abeer.ai/en/models?app=antigravity). To choose within a family, read [Which Claude model to use](https://5abeer.ai/en/guides/which-claude-model) and [Which GPT model to use](https://5abeer.ai/en/guides/which-gpt-model). ## How do Claude Code and Codex differ day to day? Most differences are in file names and commands: | | Claude Code | Codex | | --- | --- | --- | | Settings file | `~/.claude/settings.json`, or `.claude/settings.json` in a project | `~/.codex/config.toml`, or `.codex/config.toml` in a trusted project | | Project instructions | `CLAUDE.md`, or `AGENTS.md` when there is no `CLAUDE.md` | `AGENTS.md` | | Add an MCP server | `claude mcp add` | `codex mcp add` | | Run one task without the interactive session | `claude -p ""` | `codex exec ""` | | Split work across agents | Subagents | Subagents | | Work away from your computer | Web sessions, GitHub Actions, GitLab CI/CD | Codex cloud, GitHub Actions | ## How do you run both on one repository? To keep one set of rules for both tools: - **One instructions file.** Keep your project rules in `AGENTS.md`. Codex reads it in each folder from the Git root down to your working folder, and closer files take priority. Claude Code, from v2.1.277, reads it only when there is no `CLAUDE.md`, `.claude/CLAUDE.md` or `CLAUDE.local.md` in your working folder or above it. If you also keep a `CLAUDE.md`, start it with an `@AGENTS.md` line so Claude Code reads both. - **Two files that break it.** Adding a personal `CLAUDE.local.md` is enough to make Claude Code skip `AGENTS.md`; the `claude-md-and-agents-md` value of **Project instructions** in `/config` loads both. And in any folder that has an `AGENTS.override.md`, Codex reads that file instead of the folder's `AGENTS.md`, while Claude Code never reads it, so put rules for both tools in `AGENTS.md`, not in an override. - **Or let Codex read your `CLAUDE.md`.** Add `project_doc_fallback_filenames = ["CLAUDE.md"]` to Codex's `config.toml`, and Codex reads `CLAUDE.md` in folders that have no `AGENTS.md`. By default it stops adding instruction files after 32 KiB in total. - **MCP servers go in two files.** Claude Code writes project servers to `.mcp.json` when you add them with `--scope project`; without it, a server is yours alone. Codex reads project servers from `.codex/config.toml`, only in a trusted project. Both files sit in the repository, so you can commit them. - **Moving from Codex.** Claude Code's `/import codex` copies the Codex instruction files, MCP servers, commands, subagents and skills on your computer into Claude Code. Add `--dry-run` to see the list first. It isn't available when telemetry or feature-flag fetching is off. ## Which one fits your work? Each fits a different starting point: - You prefer Claude's answers, or your team already keeps `CLAUDE.md` files: start with Claude Code. - You use ChatGPT every day and want the same account in the terminal and the ChatGPT desktop app: start with Codex. - You build on Gemini, or already pay for Google AI Pro or Ultra: start with Google Antigravity. To decide, give the same real task to two of them and compare the result, the time and how much of your limit or balance it used. ## What does each one cost to run? Claude Code and Codex run on a plan or on an API key billed per token. Antigravity runs on a plan, and only its CLI can also run on a Gemini API key. On a plan, your work counts against the plan's usage limits, and long agent sessions reach them sooner. On a key, every request is billed by its model and token count, and a coding agent sends far more tokens than you type, because each turn carries the conversation, the files it read and its command output. Per Anthropic's docs, when `ANTHROPIC_API_KEY` is set, Claude Code uses that key instead of your Claude plan, even when you are signed in. An interactive session asks you to approve it once. Use every model priced per token through the API for 70% less than its official price. For real numbers, read [The cheapest way to run Claude Code](https://5abeer.ai/en/guides/claude-code-cheapest-way) and [Codex CLI with an API key](https://5abeer.ai/en/guides/codex-cli-api-key). ## Can one 5abeer plan run all three? Yes. Every Bundle plan runs every official app, from the smallest to the largest. 5abeer's Bundle plans are Pro, Max (formerly Max 5x) and Ultra (formerly Max 20x). Each runs Claude Code, Codex and Google Antigravity on your own official account, through the 5abeer AI app on Mac, Windows or Linux. They are not the Anthropic or OpenAI plans with the same names. Our prices, as on the [pricing page](https://5abeer.ai/en/pricing), in your currency: Current 5abeer Bundle plan prices are on the pricing page: https://5abeer.ai/en/pricing 5abeer's Max has 5 times the limits of 5abeer's Pro, and 5abeer's Ultra has 20 times. Each plan has one 5-hour limit and one weekly limit that cover all your apps, as in [Usage limits](https://5abeer.ai/en/docs/subscriptions/limits.md). We don't promise limits matching Anthropic's or OpenAI's plans. The [Claude Code](https://5abeer.ai/en/apps/claude-code), [Codex](https://5abeer.ai/en/apps/codex) and [Google Antigravity](https://5abeer.ai/en/apps/antigravity) pages show what you get in each. Plans are for the supported apps. To use the models from your own code, create an API key. ## How do you use them with 5abeer keys? To pay per token instead, give each tool its own key. A 5abeer key works with one provider's models, so Claude Code needs a Claude key and Codex a GPT key, and both draw on the same balance, as in [One key per provider](https://5abeer.ai/en/docs/start/keys.md#per-provider). Claude Code uses our Anthropic base URL, and Codex our OpenAI one: | Format | Base URL | Key header | | --- | --- | --- | | OpenAI | `https://api.5abeer.ai/v1` | `Authorization: Bearer ` | | Anthropic | `https://api.5abeer.ai` | `x-api-key: ` | | Gemini | `https://api.5abeer.ai/v1beta` | `x-goog-api-key: ` | The steps are in [Claude Code setup](https://5abeer.ai/en/docs/apps/claude-code.md#manual) and [Codex CLI setup](https://5abeer.ai/en/docs/apps/codex.md#manual). On a plan, the 5abeer AI app sets up each tool for you, with no key and no files to edit. ## FAQ ### Is Claude Code better than Codex? Neither wins every task. The better choice depends on which maker's models do better on your own work, and only a test on a real task shows that. ### Can I use them without a paid plan? Two of them, on their makers' terms. OpenAI includes Codex in its no-cost ChatGPT plan, and Google lists an Individual level of Antigravity at no charge. Claude Code needs a paid Claude plan or an Anthropic Console account. ### Can I use Claude Code and Codex on the same project? Yes. Both can read one `AGENTS.md`, while each tool keeps its own settings and MCP servers. The details are in [How do you run both on one repository?](https://5abeer.ai/en/guides/claude-code-vs-codex#one-repo) ### Does time in Codex count against my Claude Code limits on a 5abeer plan? Yes. Every app you turn on with one subscription counts against its 5-hour and weekly limits, so a long Claude Code session leaves less for Codex, and the other way round. ### Can I pay for only Claude Code or only Codex? Yes. 5abeer's Claude plan runs Claude Code and the other Claude apps, its GPT plan runs ChatGPT Desktop and Codex, and its Gemini plan runs Google Antigravity. On an API key, you pay only for the requests you send. ### Does a 5abeer plan work in Claude Code on the web or on my phone? No. A 5abeer plan works in the desktop and terminal apps through 5abeer AI on Mac, Windows or Linux, not in the phone apps or on the websites. ## Get started [One subscription for every AI app](https://5abeer.ai/en/guides/one-subscription-all-ai-apps) shows how to turn on each app. Or pick a Bundle plan and give Claude Code and Codex the same task. ## Sources Checked on 2026-11-03: - Anthropic: Claude Code [overview](https://code.claude.com/docs/en/overview), [settings](https://code.claude.com/docs/en/settings), [environment variables](https://code.claude.com/docs/en/env-vars), [CLAUDE.md and AGENTS.md](https://code.claude.com/docs/en/memory), [MCP](https://code.claude.com/docs/en/mcp), [commands](https://code.claude.com/docs/en/commands) and [Claude plans](https://claude.com/pricing) - OpenAI: Codex [CLI](https://learn.chatgpt.com/docs/codex/cli), [configuration reference](https://learn.chatgpt.com/docs/config-file/config-reference), [AGENTS.md](https://learn.chatgpt.com/docs/agent-configuration/agents-md), [MCP](https://learn.chatgpt.com/docs/extend/mcp), [subagents](https://learn.chatgpt.com/docs/agent-configuration/subagents), [GitHub Action](https://learn.chatgpt.com/docs/github-action) and [pricing](https://learn.chatgpt.com/docs/pricing) - Google: Antigravity [getting started](https://antigravity.google/docs/getting-started), [models](https://antigravity.google/docs/models), [plans](https://antigravity.google/docs/plans) and [pricing](https://antigravity.google/pricing) > 5abeer is an independent service, not affiliated with OpenAI, Anthropic or Google.