Compare

Cursor vs Claude Code.

An AI editor with a CLI against a terminal agent, now sold very differently. Cursor comes inside the X Subscription, with Grok and Composer models in the plan and Claude billed per token; Claude Code runs on a Claude plan. Every figure below was read on 16 November 2026 from the page linked under Sources. Starkomand, which runs both CLIs, is ours; the comparison does not depend on it.

v1.0.8 · Mac and Windows · free for 3 days · no account

The short version.

  • What changedCursor is now sold as part of the X Subscription. Its plans include Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5; Claude, GPT and Gemini models are extra.
  • Claude in CursorOn Premium ($30 a month) and up, Claude models cost Anthropic’s API list price from the first request, paid from prepaid credit. Plus ($8) and Free have no Claude at all.
  • Claude in Claude CodeA Claude plan meters the same models against five-hour and weekly windows instead of per token. Pro, at $20 a month, includes Claude Code.
  • BenchmarksClaude Code with Claude Fable 5 has 83.8% on Terminal-Bench 2.1; Cursor CLI with Grok 4.5 has 79.3%. The runs use different models, and neither vendor’s newest has a run yet.
  • If you already pay for CursorPro, Pro+ and Ultra keep working as legacy plans, with their two monthly usage pools.

Side by side.

Cursor next to Claude Code, from both vendors’ public pages on 16 November 2026.

Criterion CursorAnysphere, part of SpaceXAI Claude CodeAnthropic
Made by Anysphere, a SpaceX subsidiary since August 2026 Anthropic
Where it runs The Cursor editor, its CLI, cloud agents, the web and mobile The terminal, VS Code, JetBrains, a desktop app and the web
Models in the plan Grok 4.7, Grok 4.6, Grok 4.5 and Composer 2.5One pool that refreshes weekly, shared with the Grok apps. Fable 5.2, Opus 5.5, Sonnet 5.5 and Haiku 5.5Fable is included on Max and paid with usage credits on Pro.
Claude models At API list price from prepaid credit, on Premium and upNot available on Free or Plus. Metered by the plan’s usage windows
GPT models GPT-5.6 Sol, Terra and Luna at API list priceNo GPT-6 model is on Cursor’s list. None
Terminal-Bench 2.1, best published run 79.3% with Grok 4.5 83.8% with Claude Fable 5
Context window 256K for Grok 4.7 on Plus; a 500K variant costs more 1M tokens on every current model, at standard API pricing
Starting permissions in the CLI Allow and deny lists in cli-config.json; --force runs commands without asking Auto mode: a second model, the classifier, reviews actions instead of youSince v2.1.283. Manual mode asks before most edits and commands.
Project instructions .cursor/rules and AGENTS.md CLAUDE.md
Cheapest plan with the agent Free, with limited agent requests; Plus at $8/monthPlus covers Cursor’s own models only. Claude Pro, $20/monthClaude’s Free plan does not include Claude Code.
Top individual plan Ultra, $200/month50 times Plus usage. Claude Ultra, $200/monthNo five-hour limit for verified accounts.

Same on both. MCP servers, hooks, skills, subagents and a plan mode before edits.

Every figure on this page was read on 16 November 2026 from the page listed under Sources, and both products change monthly; product names belong to their owners. Spotted a change? Tell us.

Benchmarks.

Terminal-Bench measures agents completing real command-line tasks. Version 2.1 fixes 28 of the 89 tasks in 2.0, and no task in 2.1 is left unsolved.

Harness Model Score Run date
Claude Code Claude Fable 5 83.8% 9 Jun 2026
Cursor CLI Grok 4.5 79.3% 8 Jul 2026
Claude Code Claude Opus 4.8 78.9% 28 May 2026
Claude Code Claude Sonnet 5 74.6% 30 Jun 2026

Selected rows from the Terminal-Bench 2.1 leaderboard.

Different models, not just different CLIs
Cursor’s run uses Grok 4.5, a model its plans include; Claude Code’s uses Claude Fable 5. No run puts one model in both, so the gap mixes model and harness.
The newest models are missing
Grok 4.7, Composer 2.5, Fable 5.2 and Opus 5.5 have no runs on the board yet.

No benchmark measures your repository. The better test is one real task run in both: note the wall-clock time, what it drew from your allowance and whether the diff shipped. It is a sample of one, so repeat it after the next model release.

Prices.

Cursor sells a weekly usage pool inside the X Subscription; Anthropic sells Claude Code inside its chat plans.

Plan Price What it includes
Cursor Free $0 Limited Agent and Chat usage; no third-party models
Cursor Plus $8/month Grok 4.7, 4.6, 4.5 and Composer 2.5 at fixed settings; no Claude, GPT or Gemini
Cursor Premium $30/month Four times Plus usage, cloud agents, and Claude, GPT and Gemini at API list price from prepaid credit
Cursor Super $100/month 15 times Plus usage
Cursor Ultra $200/month 50 times Plus usage
Claude Pro $20/month, or $17/month billed annually Claude Code; Fable paid with usage credits
Claude Max $100/month Five times Pro’s usage; Fable included, up to half the weekly limit
Claude Ultra $200/month No five-hour limit for verified accounts; ten times Pro’s weekly usage

Cursor Teams seats cost $30, $100 or $200 a month and add a Cursor Token Rate of $0.25 per million tokens on eligible third-party models. Paid annually, Cursor’s individual tiers cost $80, $300, $1,000 or $2,000 a year.

At the API

What a Claude model costs inside Cursor against Anthropic’s own API, per million tokens of input and output.

Model In Cursor Anthropic API
Claude Fable 5.2 $8 / $40 $8 / $40
Claude Opus 5.5 $4 / $20 $4 / $20
Claude Sonnet 5.5 $2 / $10 $2 / $10
Claude Haiku 5.5 $0.10 / $0.50 From $0.10 / $0.50

The rate is the same. What differs is what the plan covers: in Cursor every Claude request draws prepaid credit at that rate, while in Claude Code the same requests count against usage windows, and Pro costs $20 a month however much fits inside them.

So if Claude is the model you want, a Claude plan is the cheaper way to run it, and Claude in Cursor suits occasional use. If Grok or Composer suit your work, Cursor’s pool covers them with no per-token bill.

Usage limits.

Cursor refreshes one pool weekly; Claude plans have five-hour and weekly windows.

Criterion CursorAnysphere, part of SpaceXAI Claude CodeAnthropic
Window One pool that refreshes weekly A five-hour session plus a weekly limit on Pro and Max; Ultra has the weekly limit only
Shared with Grok Chat, Imagine, Voice and Grok Bot Claude chat and the IDE extensions
When it runs out Wait for the reset, keep going on prepaid credit at the per-model rates, or move up a tier Wait for the reset, or move up a plan

Cursor’s docs say requests are never downgraded in quality or speed when the pool runs out. Claude in Cursor never draws on the pool at all: it is paid from credit from the first request.

Which one.

Pick Cursor if

you want an editor with the agent built in, Grok or Composer models without a per-token bill, or one subscription that also covers Grok and X.

Pick Claude Code if

you want Claude models on a flat plan rather than per token, a 1M-token context on every model, or a terminal agent that a classifier keeps moving.

Pick both if

you edit in Cursor and want Claude for long agent runs. Cursor Plus at $8 and Claude Pro at $20 give you the editor with Grok, and Claude on its own allowance.

MoreCursor in StarkomandClaude Code in StarkomandCodex vs Claude CodeOpenCode vs Claude CodeGitHub Copilot vs Claude CodeCodex vs CursorGemini CLI vs Claude CodeRun multiple Claude Code agents in parallelWhat is an ADE?Agent orchestrationcmux for Windows

Running both by hand.

Both run in a terminal. Keep one set of rules in AGENTS.md, which Cursor reads, and make CLAUDE.md the single line @AGENTS.md, which Claude Code expands as an import.

# Cursor CLI: installs the agent command (cursor-agent works too)
curl https://cursor.com/install -fsS | bash

# CLAUDE.md, the whole file
@AGENTS.md

# a worktree each when both write code
git worktree add ../app-cursor -b cursor-task

# then, in a terminal each
claude
cd ../app-cursor && agent

Give the two agents different jobs, and when both write code, give each its own git worktree so neither overwrites the other’s changes.

Running both in Starkomand.

Starkomand is ours, so this is the one section about it. It runs on Apple silicon Macs and on Windows 10 and 11, and needs no Starkomand account: Cursor’s CLI and Claude Code each run in a real terminal window on one canvas, signed in with your own plans or keys.

Plan in one, build in the other
Open Agent actions in a window, choose Hand off… and pick the other agent. The brief carries the objective and acceptance criteria, and the new window lands wired to the one that wrote it.
See both allowances
Settings → Usage meters puts a ring in the rail for your Cursor plan and one for Claude, each with its reset.
Read across windows
An agent can read another window’s transcript by name, so the plan does not need retyping.
Redirect by voice
Hold Option and speak to an agent by name.

We have not measured whether this makes anyone faster. What it changes is that each agent stays on its own plan, in view, and work moves between them without retyping.

Questions

Is Cursor better than Claude Code?

Not on the published benchmark: Claude Code with Claude Fable 5 has 83.8% on Terminal-Bench 2.1 against 79.3% for Cursor CLI with Grok 4.5. The runs use different models and neither vendor’s newest has a run yet, so the better test is one real task in each.

Can I use Claude in Cursor?

Yes, on Premium ($30 a month) and higher. Claude models are billed at Anthropic’s API list price from prepaid credit, from the first request, and do not draw on the weekly pool. Free and Plus have no third-party models.

Is Claude cheaper in Cursor or in Claude Code?

In Claude Code, for regular use. Cursor charges Anthropic’s API rate per token, while a Claude plan meters the same models against usage windows for a flat $20, $100 or $200 a month.

What happened to Cursor Pro at $20 a month?

Pro, Pro+ and Ultra continue for existing customers as legacy plans with two monthly usage pools. New subscribers choose from the X Subscription tiers: Free, Plus at $8, Premium at $30, Super at $100 and Ultra at $200.

Does Cursor have a CLI like Claude Code?

Yes. Cursor CLI installs an agent command with the editor’s Agent, Plan and Ask modes, plus a print mode for scripts and CI.

Can I use both in one project?

Yes. Keep the shared rules in AGENTS.md, which Cursor reads, make CLAUDE.md import it, and give each agent its own git worktree when both write code.

Do I need two subscriptions?

Yes. Cursor bills through the X Subscription; Claude Code runs on a Claude plan or an Anthropic API key.

Try it against your own setup.

Every feature, every provider, free for 3 days. Your CLIs are already installed; the canvas takes a minute.

v1.0.8 · Mac and Windows · no account