Cursor vs Claude Code pricing: each plan, usage limits, and what Claude costs in Cursor

Cursor's plans changed: new customers get Free, Plus at $8 (Cursor's own models only), Premium at $30, Super at $100 and Ultra at $200, and none of them includes Claude. From Premium up, Claude in Cursor's agent is billed at Anthropic's API list price from the first request, on top of the plan fee. Claude Code comes with Claude Pro at $20 and Max at $100 or $200, inside five-hour and weekly limits, and its extension runs inside Cursor on that plan. Prices read 20 October 2026.

Cursor vs Claude Code pricing at $20, $100 and $200

• $20. Claude Pro, with Claude Code; $17 a month billed annually. Cursor has no current plan at $20: Pro, Pro+ and the $40 and $120 Teams seats are legacy plans that existing subscribers keep and new customers cannot buy. Its nearest are Plus at $8, which runs only Cursor's own models, and Premium at $30, the cheapest that runs Claude.
• $100. Claude Max 5×, five times Pro's usage, against Cursor Super, 15 times Plus's pool for Cursor's own models.
• $200. Claude Max 20×, twenty times Pro's usage, against Cursor Ultra, 50 times Plus.

Cursor's multiples are of its pool for its own models, Claude's of Claude Pro's usage, so the two do not compare one to one. Billed yearly, Cursor lists Premium at $300, Super at $1,000 and Ultra at $2,000; Max is monthly only. For teams, a Cursor seat is $30, $100 or $200 a month and a Claude Team standard seat $25, or $20 billed annually.

Is Cursor more expensive than Claude Code?

For Claude work, yes. The cheapest Cursor plan that runs Claude is Premium at $30, against $20 for Claude Pro, and on top of either Cursor fee every Claude token is billed at the same rates as an Anthropic API key — Opus 5.5 at $4 input and $20 output per million tokens — plus $0.25 per million on Teams and Enterprise. A Claude plan includes Claude Code use inside its limits; past them you wait, buy usage credits at API rates, or move up a plan. What Cursor's fee does include is a weekly pool for its own models, Grok and Composer.

Cursor vs Claude Code usage limits

Cursor gives each plan one weekly pool for its own models, shared between Cursor and Grok and refreshed weekly; it states the pool's size only as multiples of Plus. Past it you buy top-ups of prepaid credit, wait for the reset, or upgrade. Third-party models never draw on the pool: Free and Plus do not run them, and Premium and above pay for them from prepaid credit.

Claude has a five-hour session limit and a weekly limit, shared by Claude Code, claude.ai and Claude Desktop; Max also has a separate weekly limit for Fable. Anthropic publishes no counts, only that Max 5× and 20× give five and twenty times Pro. The details: Claude Code usage limits.

What Claude costs inside Cursor

Cursor's prices for Claude match Anthropic's list, and no current plan includes any. So Claude in Cursor costs what it costs on an API key, plus the $0.25 per million token rate on Teams and Enterprise.

What that comes to for heavy agent work can only be estimated here: Caprock cannot read Cursor's agent, so the estimate is my own Claude Code calls at Anthropic's list prices. Cursor lists Opus 5 and 4.8, which most of these calls ran, at those same rates. A median Opus 5 call came to $0.11, with 169K tokens of context almost all from cache; a median active day to $153; 56 active days, 18 July to 2 October 2026, to $13,922.

Cursor's agent builds its own context, so its calls will not match these one for one, and Cursor lists one cache-write rate where most of my writes were the more expensive one-hour kind. This is not a measured Cursor bill. On a $200 Claude plan the same work is a flat fee, inside its limits. Every model's rate and a calculator: Claude API pricing.

Claude Code inside Cursor

Claude Code's VS Code extension has an "Install for Cursor" link, and Anthropic's help centre says a Pro or Max plan "also covers Claude Code in supported IDEs, including VS Code, Cursor and other VS Code forks". So you can run both: Cursor as the editor, with its own models for quick edits, and Claude Code in a panel on a Claude plan for the long agent work. Each part then goes to its own bill.

Which one to pick on cost

If it is mostly Claude, most of the day, take a Claude plan and run Claude Code in the terminal or inside Cursor: in Cursor's own agent every Claude token is extra. If you mostly run Grok or Composer, or want several makers' models in one editor and will pay the third-party ones per token, Cursor's tiers buy the pool for its own models. A team compares a Cursor seat from $30, plus Claude at API price and the token rate, with Claude Team seats from $20 billed annually; teams that use both pay for both.

Codex vs Claude Code and OpenCode vs Claude Code compare the other two agents, measured on one machine.

See your Claude side

Caprock reads Claude Code's transcripts from disk, whether or not it started the session, and prices every call — per session, per project, per model — so you can see what your Claude work would cost at API rates against your plan. The figure is a list-price estimate, as Claude Code's own /usage is, beside the plan's dollars and the share of the weekly limit the work used. It does not read Cursor's own agent.

Free, open source; nothing leaves your machine. Install Caprock and it reads back every Claude Code session still on disk.