Cursor vs Claude Code vs Codex (2026): Price, Which Wins

Clients and founder friends keep asking me which AI coding tool to pay for. In 2026, Cursor, Claude Code and Codex are all good. They're just good at different jobs, and their pricing changed a lot this year: Cursor is now sold through The X Subscription, Claude added an Ultra plan, and ChatGPT Pro split into three tiers.

This guide compares Claude Code vs Cursor vs Codex on price, plans, usage limits and workflow, including the question I get most from heavy users: Cursor Ultra vs Codex Pro. I use Grok, ChatGPT and Claude every day in my client work, so this is written from that angle.

Every price below was rechecked on November 7, 2026 on the official pages: Cursor pricing, The X Subscription pricing docs, Claude pricing, Anthropic's Team plan article, Codex pricing, OpenAI's Pro tiers article and OpenAI's Plus article. Prices are in US dollars, exclude tax and change often, so confirm before you subscribe.

Quick answer: Pick Cursor if you want an AI code editor where you approve every diff — Premium is $30/month (4x Plus usage, cloud agents, Bugbot reviews), and Claude or GPT models inside Cursor are billed extra at API list price. Pick Claude Code if you want to hand off multi-file tasks from the terminal, IDE, desktop app or web — it comes with Claude Pro at $20/month ($17/month billed annually), Max at $100 or Ultra at $200. Pick Codex if you already pay for ChatGPT — it's included from Plus at $20/month, with Pro at $100, $200 or $500/month for heavy use. At the top end, Cursor Ultra ($200), Claude Ultra ($200) and ChatGPT Pro 200 ($200) cost the same; the difference is what the usage pool covers.

Cursor vs Claude Code vs Codex at a glance

The short version

• Cursor is an AI-first code editor. You stay in a VS Code-style IDE, see every change and accept or reject diffs. Best when you need to understand every line before it ships.
• Claude Code is Anthropic's agent. You describe the outcome, it reads the repo, runs commands you approve and edits across files. It runs in the terminal, VS Code and JetBrains, a desktop app and the browser.
• Codex is OpenAI's agent across the ChatGPT desktop app, CLI, IDE extension, web and iOS. Its edge is cloud tasks and GitHub @codex delegation and PR reviews.

You can mix them, too: Anthropic even ships a Claude Code extension for Cursor, so plenty of developers run Claude Code inside Cursor as their editor.

Cursor vs Claude Code vs Codex pricing

This is the part most people search for, so here are all three side by side, then each one in detail.

Cursor pricing (The X Subscription)

Since October 2026, new self-serve customers buy Cursor through The X Subscription, a single plan that covers Cursor, Grok and X Premium benefits. Cursor's own FAQ says subscriptions are sold only directly through X Subscription Corp. Existing Cursor plans continue, and switching is optional. From cursor.com/pricing and docs.x.ai:

• Free: $0 — limited agent and chat requests.
• Plus: $8/month ($80/year) — basic shared usage pool, Cursor agent, Grok Chat, Voice, Imagine and Build. Only Grok models are available on Plus.
• Premium: $30/month ($300/year) — 4x Plus usage, cloud agents and automations, code hosting via Origin, AI code reviews via Bugbot, latest Grok models, Grok Bot, and third-party models at API list price.
• Super: $100/month ($1,000/year) — 15x Plus usage and priority model access at peak times.
• Ultra: $200/month ($2,000/year) — 50x Plus usage and earliest access to new products and models.
• Teams: from $30 per user per month (Premium, Super or Ultra seats, mix and match), plus Enterprise at custom pricing.

The catch: third-party models are not in the pool. If you want Claude or GPT models inside Cursor, you pay for them separately at API list price from the first use, and you can set spending caps. For every plan detail, see my Cursor pricing guide.

Claude Code pricing and plans

There's no separate "Claude Code plan." Claude Code is included in every paid Claude plan and shares one usage pool with Claude chat on web, desktop and mobile. From claude.com/pricing:

A few details that matter for coding:

• Models: Pro, Max and Ultra include Opus, Sonnet and Haiku. Anthropic's top model, Fable, is paid with usage credits on Pro and included for up to half your weekly limit on Max and Ultra.
• Billing: Max and Ultra are monthly only; annual billing is for Pro and Team.
• When you hit a limit: wait for the reset, upgrade, or turn on usage credits to keep going at standard API rates. You can also run Claude Code on pay-as-you-go API credits through a Console account.
• Team plans cover 2 to 150 people (per Anthropic's Team article), and usage limits are per member, not shared.

For the full breakdown, including API token costs, read my Claude Code pricing guide and the wider Claude pricing guide.

Codex pricing (ChatGPT plans)

Codex is included in ChatGPT Free, Go, Plus, Pro, Business, Edu and Enterprise, and it shares usage with ChatGPT Work. From OpenAI's Codex pricing page:

• Free and Go ($8/month in the US): GPT-6 Luna at Standard speed in the desktop app, subject to rollout. Good for lightweight tasks.
• Plus ($20/month, monthly only): Codex on the web, in the CLI, in the IDE extension and on iOS; cloud integrations like automatic code review and Slack; GPT-6.1 Sol and GPT-6 Luna; extend usage with ChatGPT credits.
• Pro ($100, $200 or $500/month): longer Codex and ChatGPT Work sessions, GPT-6 Astra Pro reasoning, and Ultrafast mode on Pro 500.
• API key: Codex in the CLI, SDK or IDE extension, billed at API rates — good for CI, but no cloud features like GitHub code review or Slack.

One change to note: GPT-5.5 retired from Codex with ChatGPT sign-in on October 14, 2026. API-key sessions weren't affected.

Cursor Ultra vs Codex Pro (ChatGPT Pro)

If you're spending $100 to $200 a month, you're really choosing between Cursor Ultra/Super and ChatGPT Pro (which powers Codex Pro). I've added Claude Ultra too, since it sits at the same price.

My take on Cursor Ultra vs Codex Pro:

• If most of your coding runs on Grok models inside an editor, Cursor Ultra gives you a huge weekly pool (50x Plus) plus Grok and X perks. But if you mainly want Claude or GPT in Cursor, that spend sits outside the pool, so Ultra's 50x doesn't help much — you'll pay API list price on top.
• If you want long autonomous agent sessions on OpenAI models, ChatGPT Pro is the cleaner deal: Codex usage is included, there's no five-hour cap right now, and you get the rest of ChatGPT Pro. Start on Pro 100 and move up only if you hit the weekly limits.
• If you want the same thing on Claude models, Claude Max ($100) or Ultra ($200) is the match, and Ultra's $200 of monthly API credits is a real bonus if you also build on the Claude API.

The $100 tier is where most people should start: Cursor Super vs ChatGPT Pro 100 vs Claude Max. Only go to $200 when you hit limits most days.

Usage limits compared

None of the three gives you a fixed number of requests. Here's how each one works today:

• Cursor: each plan has an AI usage pool shared across Cursor and Grok that refreshes weekly. Premium is 4x Plus, Super 15x and Ultra 50x. Run out and you can buy top-offs or wait for the refresh. Third-party models are billed separately.
• Claude Code: limits reset on a rolling five-hour window, with a weekly limit on paid plans. Max is 5x Pro on both; Ultra is 10x Pro weekly with no five-hour limit for verified accounts. Your chats and Claude Code sessions draw from the same pool. Check Settings > Usage.
• Codex: OpenAI's own estimate for Plus is roughly 15 to 160 local messages every five hours on GPT-6.1 Sol, or 350 to 3,000 on GPT-6 Luna; weekly limits may also apply. Pro plans currently have no five-hour limit. Cloud tasks use more of your allowance than local messages, and Fast mode uses included usage at 2.5x the Standard rate. Use /status in the CLI to see what's left.

If you hit limits often on a $20 plan, the fix is usually the same: scope tasks tighter, trim your AGENTS.md or CLAUDE.md, use a lighter model for routine work, or move up to the $100 tier. With Cursor, watch third-party model spend, since that's where surprise bills come from. My AI Token Calculator helps estimate API-rate costs.

Workflow: how each one feels to use

Cursor: control in the editor

Cursor wins on control. You stay in a familiar VS Code-style editor, the agent proposes changes and you accept or reject diffs. For UI tweaks, quick fixes and code you must fully understand before shipping to a client, that matters. Typical Cursor jobs: adjusting a landing page layout, fixing a form validation bug, renaming a component across a few files of a Next.js or React codebase. Premium adds cloud agents and Bugbot code reviews, so it's no longer editor-only.

Claude Code: delegate the whole task

With Claude Code, you describe the outcome and review the result. Per Anthropic's Claude Code docs, it reads your codebase, edits files, runs commands, works with git (commits, branches, pull requests), connects to tools over MCP, and reads a CLAUDE.md (or an existing AGENTS.md) for project rules. It runs in the terminal, VS Code and Cursor, JetBrains, a desktop app and the web, and can run scheduled tasks in the cloud. It suits jobs like "write tests for the auth module, run them and fix failures" far more than tiny edits.

Codex: ChatGPT's agent, strong in the cloud

Codex works locally in the desktop app, CLI and IDE extension, and also runs cloud tasks you can start and check back on. On Plus and above you get GitHub @codex delegation, automatic PR reviews, and Slack and Linear integrations. If your team lives in GitHub pull requests, that's a real advantage. With an API key you can run it in CI, but without the cloud features.

A good habit with any agent: give it one clear, scoped task, ask for a plan first, and review the final diff like a junior developer's pull request. Run your tests before you merge.

Claude Code vs Cursor

The Claude Code vs Cursor choice is about who drives. In Cursor, you drive and the AI assists; in Claude Code, the agent drives and you review. Price-wise, Claude Pro ($20) includes Claude Code with Claude models in the plan, while Cursor Premium ($30) includes Grok models in the pool but bills Claude models at API list price. So if you specifically want Claude models for coding, Claude Code is usually cheaper. If you want the best editing experience with Grok models included, Cursor wins.

Claude Code vs Codex

Both are agents first. Codex is the better pick if you already pay for ChatGPT, want cloud tasks, or want @codex reviews on GitHub. Claude Code fits better if you work mostly in a local terminal or IDE and already use Claude for writing and chat. At $20 both cost the same (Claude Pro is $17/month on annual billing; ChatGPT Plus is monthly only). At the top end, ChatGPT Pro offers a $500 tier with Ultrafast; Claude tops out at Ultra $200.

Cursor vs Codex

Pick Cursor if you want an AI editor where you approve each diff. Pick Codex if you'd rather describe a task and let an agent run it in the app, CLI or cloud. Many developers use one of each: Cursor for everyday edits, Codex (or Claude Code) for bigger, well-defined tasks.

Where GitHub Copilot and Devin Desktop fit

GitHub Copilot is the budget alternative if you don't want to switch editors. Per GitHub's plans page, Pro is $10/month, Pro+ $39 and Max $100, all paid in GitHub AI Credits. It works in VS Code, Visual Studio, JetBrains, Xcode, Neovim, Zed and more, and Pro+ and Max can delegate tasks to third-party agents like Claude and Codex (preview). See Copilot alternatives and Claude vs Copilot.

Windsurf is now Devin Desktop. Its pricing page lists Free, Pro at $20/month (with frontier models from OpenAI, Claude, Gemini and SpaceXAI in the quota) and a new Max at $200/month. It's the closest editor-style rival to Cursor — see Windsurf vs Cursor and Cursor alternatives.

Which one should you choose?

• Non-technical or early-stage founder building an MVP: Claude Code (Claude Pro, $20) or Codex (ChatGPT Plus, $20). Agents do more of the heavy lifting, and you review results.
• Developer doing client work who wants control: Cursor Premium ($30).
• Already paying for ChatGPT Plus: use Codex first before buying anything else.
• Already paying for Claude Pro: Claude Code is included — try it before adding another tool.
• Heavy daily agent user: start at $100 (Claude Max, ChatGPT Pro 100 or Cursor Super), and move to $200 only if you hit limits most days.
• Want Claude models in an editor without API bills: Claude Code inside VS Code or Cursor on a Claude plan, rather than Claude models billed through Cursor.
• Team on GitHub that wants low friction: Copilot, adding Claude Code or Codex for bigger tasks.

You don't have to pick just one. A practical combo is an editor tool (Cursor or Copilot) for everyday edits plus an agent (Claude Code or Codex) for bigger tasks. Start with one paid plan and add a second only when you hit real limits.

If you mostly want an assistant for writing and research rather than coding, read ChatGPT vs Claude or Grok vs ChatGPT. Prefer coding in the browser with no local setup? See Replit alternatives.

Need help choosing or setting up an AI coding workflow? Get in touch.

FAQ

My recommendation

If I had to pick one tool for a founder who doesn't want to think about it, I'd go with Claude Pro at $20/month, since it bundles Claude Code with chat. ChatGPT users should try Codex first, because it's already in Plus. Developers who want control should pick Cursor Premium. Heavy users should start at the $100 tier and only move to $200 when limits bite. Test an entry tier on a real project for a week first.

Prices and plans above were rechecked on cursor.com/pricing, docs.x.ai, claude.com/pricing, developers.openai.com/codex/pricing and OpenAI's help centre on November 7, 2026. Always confirm on those pages before you buy.

More AI tool comparisons on the blog.