Cursor vs Claude Code Compared After Cursor Joined SpaceX
Cursor vs Claude Code in October 2026. Cursor now bills Claude at API rates. Current plans, models and limits, plus our own token log priced both ways.
Pick Cursor if you want a full editor, a $0 or $8 way in, and you are happy to work mostly with Cursor's own Grok and Composer models. Pick Claude Code if you want Claude models doing the work, because Cursor no longer includes any Claude usage in its new plans. On The X Subscription, every Claude request in Cursor is paid at API list price from prepaid credit, from the first request.
That last change is the whole story, and almost no page ranking for this topic has caught it. Cursor became part of SpaceX on August 14, 2026, and new customers now buy it as part of The X Subscription. We run Claude Code all day to build the sites and apps we write about here, so we are not neutral. What we can give you is both price lists as they stood on October 19, 2026, plus two weeks of our own agent logs priced on each one.
Before anything else, the quick points.
- Cursor is an editor. Claude Code is an agent. Claude Code also runs inside Cursor as an extension, so this is not either or.
- Cursor's plans changed. New buyers get Free, Plus $8, Premium $30, Super $100 or Ultra $200, all shared with Grok and X.
- Claude is no longer included in Cursor. Claude, GPT and Gemini cost API list price on Premium and up. Plus and Free cannot use them at all.
- Claude Code starts at $20. Pro is $20 a month, or $17 billed yearly. It defaults to Opus 5.5.
- Our two weeks of Opus work would cost $9,327 at the rates Cursor charges. We paid two flat Claude plans instead.
- Old Cursor plans still work. Pro, Pro+ and Ultra keep their included Claude pool until you switch.
Inside This Post
8 sections
What Cursor and Claude Code Actually Are
Cursor is a code editor built on VS Code. You open your project, see your files, and talk to an agent in a side panel. It also has a CLI, cloud agents, a mobile app and code review. It is made by Anysphere, which Cursor's own blog says was acquired by SpaceX on August 14, 2026.
Claude Code is Anthropic's coding agent. The Claude Code overview lists the terminal, VS Code, JetBrains, a desktop app and the web. You tell it what you want and it reads files, runs commands and keeps going until the job is done or it needs you.
Most comparisons miss this. Claude Code's VS Code docs have an "Install for Cursor" button. So you can run Claude Code inside Cursor and get the editor plus the agent. The real choice is about whose models you pay for and how you pay for them.
| Cursor | Claude Code | |
|---|---|---|
| Made by | Anysphere, part of SpaceX | Anthropic |
| What it is | A full editor with an agent inside | An agent that runs in many places |
| Cheapest way in | Free, $0, limited | Pro, $20 a month |
| Included models on new plans | Grok 4.7, 4.6, 4.5 and Composer 2.5 | Opus 5.5 by default |
| Claude models | API list price from prepaid credit, Premium and up | Included in the plan |
| Instruction file | .cursor/rules or AGENTS.md | CLAUDE.md, can read AGENTS.md |
| Usage resets | One weekly pool, shared with Grok | Rolling five hour window plus weekly limits |
Models and plans come from Cursor's models and pricing docs, Cursor's rules docs, claude.com/pricing and Anthropic's model config docs, all opened October 19, 2026.
What Changed When Cursor Joined SpaceX
On August 14, 2026, Cursor posted that it had been acquired by SpaceX. The post says the deal finished a process that started in April, when Cursor partnered with SpaceXAI on model training.
The pricing followed. Cursor's pricing page now sells five individual plans, and each one is also an X plan. The fine print says subscriptions are sold through X Subscription Corp. The same page lists X perks like a premium checkmark and ad free scrolling next to the coding features.

Three changes matter if you build apps.
One weekly pool, shared with Grok. Cursor's usage limits page says every paid plan gets one pool for Cursor models. Grok chat, Grok Imagine, Grok Bot and Cursor all draw from it. It resets weekly. Cursor does not publish the pool in dollars or requests. It only says Premium is 4x Plus, Super is 15x and Ultra is 50x.
Claude moved out of the plan. The X Subscription overview says Claude and GPT are not included. You can still use them in Cursor on Premium and up, billed at API list price from the first use and paid from prepaid credit. On Plus, the docs say only Cursor's own models are available.
Old plans are frozen in place. Cursor's legacy plans page says Pro at $20, Pro+ at $60 and Ultra at $200 keep working, with a separate Other Models pool for Claude and GPT. It also says the new plans give at least twice the usage on Cursor models but drop that included third party allowance. Switching is optional and permanent. If you are on old Pro and you mostly use Claude, read that page twice before you click Unify.
Every Plan, Side by Side
All prices are US list prices per month before tax, as of October 2026.
| Price | Cursor (The X Subscription) | Claude Code (Claude plans) |
|---|---|---|
| $0 | Free, limited Agent and Chat | Not included |
| $8 | Plus, Cursor models only, fixed settings | No plan at this price |
| $17 to $20 | No plan at this price | Pro, $20, or $17 billed yearly |
| $20 to $25 a seat | No match | Team Standard seat, $25, or $20 billed yearly |
| $30 | Premium, 4x Plus, adds Claude at API price | No plan at this price |
| $30 to $200 a seat | Teams, Premium, Super or Ultra seats | Team Premium seat, $125, or $100 billed yearly |
| $100 | Super, 15x Plus | Max, from $100, 5x Pro usage |
| $200 | Ultra, 50x Plus | Max 20x tier |
| Claude usage | Extra, at API list price | Included in every paid plan |
Cursor plans come from cursor.com/pricing and its models and pricing docs. Cursor also lists yearly prices of $80 for Plus, $300 for Premium, $1,000 for Super and $2,000 for Ultra. Claude plans come from claude.com/pricing. We break the Claude side down further in Claude Code pricing explained.
Two things jump out.
Cursor has the cheaper door. If you want to feel what an AI editor is like, Cursor Free costs nothing and Plus costs $8. Claude Code has no free plan.
Claude Code has the cheaper Claude. Every paid Claude plan includes Claude Code, and Anthropic's model docs say Pro, Max, Team and Enterprise all default to Opus 5.5. In Cursor's new plans, the same Opus 5.5 is a metered extra.
What Claude Costs Inside Cursor
Cursor publishes the per token rates it charges for third party models. The Claude rows sit next to Cursor's own models below, per million tokens, from Cursor's docs.

| Model in Cursor | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| Claude Opus 5.5 | $4 | $5 | $0.20 | $20 |
| Claude Sonnet 5.5 | $2 | $2.50 | $0.10 | $10 |
| Claude Haiku 5.5 | $0.10 | $0.125 | $0.01 | $0.50 |
| Grok 4.7, past the pool | $2 | not listed | $0.50 | $6 |
| Composer 2.5, past the pool | $0.50 | not listed | $0.20 | $2.50 |
Those Claude rates match Anthropic's own API list. Cursor is not marking Claude up on individual plans. On Teams and Enterprise it is a different story. The docs add a Cursor Token Rate of $0.25 per million tokens on every third party request, on top of the model's price.
Our Own Usage, Priced in Cursor
Nobody else on this topic has this. We pulled our own Claude Code logs from one Mac for October 5 to October 18, 2026. That is 14 days, 3,198 sessions and 136,888 model calls. About 93 percent of those calls ran on Opus 5.5.
The mix was lopsided. 28.0 billion tokens were cache reads, 555 million were cache writes and 47 million were output. Cache reads were 98 percent of everything going in. That is normal for an agent. It rereads your project on every step.
We priced that exact mix on Cursor's published rates. This is a worked example, not a bill. Cursor's agent would not use the same number of tokens for the same work, and the Grok and Composer numbers ignore the weekly pool, which Cursor does not size in dollars.

| Priced as | Cost for our two weeks |
|---|---|
| Two Claude Max 20x plans (what we pay, $400 a month) | About $184 |
| Cursor, Claude Sonnet 5.5 at list price | $4,664 |
| Cursor, Composer 2.5 past the weekly pool | $6,005 |
| Cursor, Claude Opus 5.5 at list price | $9,327 |
| Cursor, Grok 4.7 past the pool, cache writes counted as input | $15,416 |
| Cursor Teams, Opus 5.5 plus the $0.25 token rate on every token | $16,490 |
Two lessons came out of this.
Cache reads decide the bill. Grok 4.7 has cheaper output than Opus 5.5, $6 against $20. But its cache read rate is $0.50, which is 2.5 times Opus. On a workload that is 98 percent cache reads, that one line flips the result.
Heavy Claude users should not pay per token anywhere. At our level of use, the flat Claude plans were worth about 50 times their price compared with Opus at list rates. That holds whether the meter is Anthropic's API or Cursor's credit.
Our tool log tells the same story. 52 percent of 39,602 tool calls in those two weeks were shell commands. The agent spends its day running and checking things, and every step rereads the context.
How Each One Keeps You Safe
An agent that runs shell commands on your laptop needs limits. The two tools set them differently.
Cursor asks first. Cursor's agent security docs say terminal commands need your approval by default. You can loosen that with Run Modes, which range from an allowlist to an Auto-review classifier. Cursor calls those best effort guardrails, not a hard security wall. Every MCP tool call also needs approval unless you allowlist it. The Cursor CLI adds a /sandbox switch for command runs.
Claude Code starts in auto mode. Anthropic's permission modes docs say auto mode is the starting mode for terminal and VS Code sessions on version 2.1.283 or later. A second model reviews risky actions instead of you. Claude Code also has an OS level Bash sandbox you turn on with /sandbox.
So Cursor puts you in the loop by default and Claude Code puts a reviewer model there. If you are new, keep Cursor's approvals on, or turn on the sandbox in Claude Code. Either way, never hand an agent a production database key. We wrote up what that looks like in the view that ignored RLS.
Which One Should You Use?
This is how we would pick for people building apps, not for people paid to write code.
| You are | Pick | Why |
|---|---|---|
| Curious, not ready to pay | Cursor Free or Plus | The only $0 or $8 way in |
| Want an editor and are fine with Grok models | Cursor Premium | Agent, cloud agents and code review for $30 |
| Want Claude doing the work | Claude Code on Pro, then Max | Claude is included, not metered |
| Want an editor and Claude | Cursor Free plus Claude Code inside it | The Claude Code extension installs in Cursor |
| On old Cursor Pro, mostly using Claude | Stay put for now | Legacy Pro still includes a Claude pool |
| Running long agent sessions all day | Claude Max | Our log shows per token costs about 50 times more |
Can you use both? Yes, and it is easy. Put your project rules in AGENTS.md. Cursor's rules docs accept it, and Claude Code can read it in place of a CLAUDE.md. Then install the Claude Code extension in Cursor and use each for what it does best.
If you are still choosing your first tool, start with choosing your first AI coding tool. Setting up the Claude side takes about five minutes with our Claude Code install guide. Our resource pages on Cursor and Claude Code cover what each one is good at.
Bad Advice We Keep Seeing
Most of the advice on this topic is fine on the basics. All of it missed the biggest change of the year.
They still sell the old Cursor plans. The most linked page we checked, from Builder.io, says it was last updated in February 2026. It lists Cursor Pro at $20 with credits that buy Claude Sonnet requests. New buyers cannot get that plan now. Qodo's page from May 31 lists Pro, Teams at $40 and Ultra, with no word on The X Subscription.
They name old models. Builder.io compares Opus 4.6 with GPT-5.3. Tembo's page from September 2025 still calls Claude 3.7 Sonnet the best coding model. Today Claude Code defaults to Opus 5.5 and Cursor's own lineup is Grok 4.7 and Composer 2.5.
They treat multi model access as free. Several pages praise Cursor for letting you swap between Claude, GPT and Gemini. That is still true. But on the new plans every one of those swaps is metered at API price, and on Plus you cannot make it at all.
Still Wondering?
- Which is better, Claude Code or Cursor?
Neither wins everything. Cursor is the better editor and the cheaper way to start. Claude Code is the better deal if you want Claude models, because Cursor's new plans bill Claude at API price. Many builders run Claude Code inside Cursor and get both.
- How much does Claude Code cost per month?
Claude Code comes with Claude Pro at $20 a month, or $17 billed yearly. Max starts at $100 a month for 5x the Pro usage. Team seats start at $25 a month, or $20 billed yearly. The free Claude plan does not include Claude Code.
- Can I use Claude in Cursor?
Yes, two ways. On Cursor Premium and up you can pick Claude models and pay API list price from prepaid credit. Or you can install the Claude Code extension in Cursor and use your Claude plan. Cursor Plus and Free do not offer Claude models.
- Did Cursor get bought by SpaceX?
Yes. Cursor's blog says SpaceX completed the acquisition on August 14, 2026. New Cursor customers now buy it as part of The X Subscription, one plan for Grok, Cursor and X.
- Is Cursor still $20 a month?
Not for new customers. The $20 Pro plan is now a legacy plan. Existing Pro users keep it until they switch. New buyers choose Plus at $8, Premium at $30, Super at $100 or Ultra at $200.
- Can I use Claude Code in Cursor?
Yes. Claude Code's VS Code docs include an Install for Cursor link. The extension adds the Claude Code panel with inline diffs and plan review, and it runs on your Claude plan, not on Cursor's credit.
- What are the usage limits on Cursor and Claude Code?
Cursor gives each paid plan one weekly pool for its own models, shared with Grok, and sizes plans as multiples of Plus. Claude plans reset on a rolling five hour window with weekly limits on top. Neither vendor publishes an exact request count.
Like this guide? Add Vyebe as a preferred source on Google and see more of our posts when you search.
The Paper Trail
- Cursor, Cursor is now a part of SpaceX, August 14, 2026, the acquisition, accessed October 2026
- Cursor, Pricing, plans, prices and the X Subscription Corp note, accessed October 2026
- Cursor docs, Models and pricing, plan table, per token rates, Cursor Token Rate and yearly prices, accessed October 2026
- Cursor docs, Legacy Cursor plans, what old Pro, Pro+ and Ultra keep, accessed October 2026
- Cursor help, What is The X Subscription, third party models and switching, accessed October 2026
- Cursor help, Usage and limits, weekly pool and resets, accessed October 2026
- Cursor docs, Rules,
.cursor/rulesandAGENTS.md, accessed October 2026 - Cursor docs, Agent security, approvals and Run Modes, accessed October 2026
- Cursor docs, CLI overview, install and sandbox controls, accessed October 2026
- Anthropic, Plans and pricing, Claude plans, seats and usage rules, accessed October 2026
- Claude Code docs, Overview, surfaces and
AGENTS.mdsupport, accessed October 2026 - Claude Code docs, VS Code, the Install for Cursor link, accessed October 2026
- Claude Code docs, Model configuration, default model per plan, accessed October 2026
- Claude Code docs, Permission modes, auto mode and the Bash sandbox, accessed October 2026
- Builder.io, Claude Code vs Cursor, last updated February 2026, the most linked ranking page we compared against, accessed October 2026
- Qodo, Claude Code vs Cursor, May 31, 2026, accessed October 2026
- Tembo, Cursor vs Claude Code, September 19, 2025, accessed October 2026
Stalled somewhere in here?
Send us the project and what's blocking you. You'll get a straight read on what's actually wrong, including if it's something you can fix yourself.
Tell us what happenedWorth Reading
Codex vs Claude Code Compared on Price, Limits and Safety
Codex vs Claude Code in October 2026. Current plans, published limits, API rates, sandbox defaults and our own 24 day usage log, so you can pick the right agent.
Getting StartedHow to Install Claude Code on Mac, Windows and Linux
Install Claude Code with one command on Mac, Windows or Linux, sign in, and fix the errors that stop most installs, from PATH to 403s and npm permissions.