Claude Code vs Cursor vs Codex vs GitHub Copilot: what each costs and does

Claude Code vs Cursor is a choice between two ways of working with a coding agent: Claude Code is Anthropic's agent that runs in your terminal, your IDE or the web, and Cursor is a code editor built around an agent. Most people also weigh OpenAI's Codex and GitHub Copilot, so this page compares all four on what each company publishes: price, plans, where it runs, how usage is limited and what it says about your code. Everything was checked on October 6, 2026 against each vendor's own pages, and Cursor's plans again on October 29, 2026.

What this page is not: a quality ranking. We have not run all four on the same project, so we do not say which one writes the best code. How we checked: this comparison relies only on the vendors' own documentation, never on a tool's opinion of itself or of its competitors.

Claude Code vs Cursor vs Codex vs Copilot at a glance

Table: the four tools compared (October 6, 2026; Cursor row October 29, 2026)

Cursor and GitHub Copilot publish their prices in US dollars. For the Anthropic and OpenAI plans, euro prices are as shown on their pricing pages from Spain: Anthropic says its prices exclude tax, and OpenAI does not say whether its euro prices include it, so check the checkout total.

Claude Code vs Cursor: a terminal agent or an editor?

Anthropic describes Claude Code as "an agentic coding tool that reads your codebase, edits files, runs commands, and integrates with your development tools", available in the terminal, in IDEs, in a desktop app and in the browser. You install it with a command such as curl -fsSL https://claude.ai/install.sh | bash (macOS, Linux, WSL), irm https://claude.ai/install.ps1 | iex (Windows PowerShell), brew install --cask claude-code or winget install Anthropic.ClaudeCode, then run claude in a project folder (step by step in how to use Claude Code). Most surfaces require a Claude subscription or an Anthropic Console account.

Cursor is a different shape: you work inside its editor, and the agent is part of it. Its free plan offers limited agent and Chat requests. The paid plans now share one usage pool between Cursor and Grok, and the pricing page sells them as an X subscription that also includes features of the X social network. Plus includes Grok models only; from Premium, the plans add cloud agents and automations, code hosting, code reviews with Bugbot and third-party models, which are billed separately at their API list price.

They are not exclusive. Anthropic's documentation lists a Claude Code extension for VS Code that can also be installed in Cursor, and GitHub's Copilot Pro plan lists access to third-party agents, including Claude Code and Codex, although its comparison table lists delegating tasks to them from the cloud agent (a preview feature) as included from Pro+ and not in Pro. If you want to know what an agent is under the hood, read what an agent is.

What does each tool cost?

Claude Code has no plan of its own: it is included in every paid Claude plan and shares that plan's usage limits with your chats. Pro costs $20 a month, or $17 billed annually, and Max costs $100 or $200 a month; for heavy sessions Anthropic offers pay-as-you-go API credits through a Console account. The full plan breakdown is in Claude pricing.

Cursor lists Free ($0), Plus ($8 a month, Grok models only), Premium ($30, four times Plus's usage), Super ($100, 15 times), Ultra ($200, 50 times), Teams (from $30 per user per month) and Enterprise (custom). Cursor says all prices exclude applicable taxes and that third-party models are not included in the monthly price: they are billed by usage at their API list price, with spending caps you can set. It recommends Premium for newcomers, Super for daily agent users and Ultra for power users.

Codex comes with ChatGPT: OpenAI's pricing page says ChatGPT Work and Codex are included in the Free, Go, Plus, Pro, Business, Edu and Enterprise plans, with limited access on Free and Go and expanded usage from Plus. The plans and euro prices are in ChatGPT pricing.

GitHub Copilot has the most plans: Free ($0, with 2,000 completions a month and access to Haiku 4.5, GPT-5 mini and other models), Pro ($10 per user per month, with the cloud agent, code review, unlimited completions and access to third-party agents, though delegating to them from the cloud agent is listed from Pro+), Pro+ ($39, adding premium models including Opus and 4x or more of Pro's included usage) and Max ($100, with priority access to new models and 2.9x or more of Pro+'s usage). The paid plans include monthly credits of $15, $70 and $200. Students can use a separate Copilot Student plan.

How do the usage limits work?

Each vendor limits usage differently, which is why a plan price alone does not predict a bill. Claude resets limits on a rolling five-hour window with weekly limits on top. Cursor gives each plan a usage pool shared with Grok that refreshes weekly, sells top-offs when you run out and bills third-party models separately at their API list price. OpenAI says the number of Codex messages "depends on the model used, size and complexity of your tasks, and whether you run them locally or in the cloud". GitHub gives each paid plan a monthly credit balance and notes that the cloud agent consumes GitHub Actions minutes and that code review incurs Actions usage. For the unit behind all of these, see what a token is.

What do they say about your code?

Cursor states that when Privacy Mode is enabled, "code data is not used for training by us or our model providers", and that a team admin can enable it for everyone. On Anthropic's individual plans, model training is listed as an opt-out, and OpenAI lists an opt-out on all four individual ChatGPT plans, the plans that include Codex. If your code is confidential, read the vendor's data page and enable the setting before you open the project.

Which should you choose?

Without a controlled test, the useful way to choose is by how you work and what you already pay for:

• You already pay for ChatGPT Plus: try Codex first, it adds no subscription.
• You already pay for Claude Pro or Max: Claude Code is included, in the terminal or in your editor.
• You want the editor to be built around the agent: Cursor.
• You want the cheapest way into agent-style coding inside your current editor: Copilot Pro at $10, or the Free plan to start.
• Your team needs billing, SSO and admin controls: compare Cursor Teams, Copilot's business plans and the Claude and ChatGPT team plans.

Most of them have a free way in: give two of them the same real task from your project before you pay for either. For a wider list, including Google Antigravity, Mistral and local models, see best coding assistant.

What to check before you pay

Prices exclude tax on several of these pages, so read the checkout total. Credits, usage allowances and plan names change often: this page is reviewed monthly and every change is logged in the history below. For running models on your own computer instead of paying for a cloud plan, see how to run an LLM locally.