A side project starts in Grok Bot now. Claude Code vs Cursor, for AI coding in 2026, is Grok Bot for that side project, or Claude Code for the careful work. Cursor Ultra vs Claude Code Max is the price comparison, from the vendor pages on 2 November 2026.
Grok Bot is independent and isolated. It sets the work up, runs it, and tells me when it is done. It feels fast and snappy.
Claude Code can do that job. It feels more involved. I go back to it when Grok Bot runs out of usage, or when the task is too complex to send off.
This is how it is for me, because Claude Code's results were better when I moved, and Grok Bot tells me when a side project is done. It might not work for you.
Claude Max is $100 a month and includes Claude Code at 5x Pro usage.
Cursor Ultra is $200 a month. On a new Cursor subscription, that $200 covers Grok and Composer. Claude on that plan is billed at API price, from prepaid credit.
The $200 Claude plan is Ultra, at 10x Pro usage per week, and it includes Claude Code. Max 20x was this post's August name for that $200 Claude plan.
I pay for Claude Max. Work is on Claude Teams. I am on Cursor Super.
Prices below are from Claude pricing, Cursor pricing, and Cursor models and pricing, read on 2 November 2026.
Cursor vs Claude Code in 2026: Claude Code does the careful pass, Grok Bot takes the side project
The job row is my judgement. The rest is the product, as of 2 November 2026.
| Claude Code | Cursor | |
|---|---|---|
| What it is | Coding agent in the terminal and in an IDE extension | Editor with Agent and Tab |
| The job I give it | The careful pass, and the fallback when usage runs out or the task is too complex | A side project, through Grok Bot: set up, run, tell me when it is done |
| Included models | Claude, on a paid Claude plan | Grok 4.7, Grok 4.6, Grok 4.5, Composer 2.5 |
| Claude as the model | Included with Pro, Max, and Ultra | API price from Premium up. A monthly pool on legacy Pro, Pro+, and Ultra |
| CLI | Yes, including CI | Yes. The weekly pool covers it |
The same CLI runs in a terminal, in CI, and on a remote server, which the platforms page lists.
On the Anthropic API, Claude Code's model aliases resolve to Opus 5.5, Sonnet 5.5, and Haiku 5.5, and fable resolves to Fable 5.1. That is the model configuration page, read on 2 November 2026.
The default model on Pro, Max, and Ultra is Opus 5.5, unless an organization default replaces it. Fable is a model you select.
Cursor's included pool is those Grok and Composer models, shared with Grok, and it refreshes weekly. Claude, GPT, and Gemini are third-party models. Premium and above can use them at API list price, paid from prepaid credit. Plus and Free stay on Grok and Composer.
Cursor can run Claude. Claude Code, on a Claude plan, runs Claude. The overview describes Claude Code as the agent that reads the repo, edits files, and runs commands. Cursor lists Agent on every current plan, on the pricing page.
I moved to Claude Code because the results were better
This is my judgement.
Cursor came first. It was the IDE, for hand-coding and small inline, hands-on edits. I never really used the newer agent window.
I moved to Claude Code for almost everything. The results were so much better. Cursor's results felt lacking to me.
Claude Code installs in Cursor, on a separate subscription
You can use Claude Code in Cursor. Anthropic's VS Code page has an install link for Cursor, read on 2 November 2026.
The extension opens a Claude Code panel in the editor, with diffs and plan review. It bundles a CLI for that panel. The claude command in the terminal is a second install. The extension's CLI stays inside the panel.
You sign the extension in with a paid Claude plan, or a Console account. Cursor bills the Cursor plan. The two invoices stay separate.
The model picker is Cursor calling a Claude model, billed on the Cursor plan, at the prices further down. My own Claude Code runs in Orca.
I run Claude Code in Orca, and Grok Bot takes the side projects
This is my practice.
Claude Code runs inside Orca. Before that, it ran in Ghostty.
Orca is great at splitting worktrees and at keeping that work isolated, and it is the orchestrator across parallel agents. You need a really streamlined setup. The worktree side of that is in Claude Code worktrees.
Grok Bot eats usage because it does so many more things.
Cursor Ultra vs Claude Code Max: Ultra is $200, Max is $100
Claude Code vs Cursor pricing below is what the vendor pages showed on 2 November 2026.
| Plan | Price | What you get |
|---|---|---|
| Claude Free | $0 | No Claude Code |
| Claude Pro | $20/mo, or $17/mo annual | Claude Code included. Five-hour and weekly limits |
| Claude Max | $100/mo | Claude Code at 5x Pro, every five hours and every week |
| Claude Ultra | $200/mo | Claude Code at 10x Pro weekly. No five-hour limit on a verified account |
| Cursor Free | $0 | Limited Agent and Chat. No third-party models |
| Cursor Plus | $8/mo | Grok and Composer only |
| Cursor Premium | $30/mo | 4x Plus. Claude at API price, from prepaid credit |
| Cursor Super | $100/mo | 15x Plus. Same third-party pricing |
| Cursor Ultra | $200/mo | 50x Plus of Grok and Composer. Claude still at API price |
| Cursor Pro, legacy | $20/mo | Existing subscribers. A monthly pool for third-party models |
| Cursor Ultra, legacy | $200/mo | Existing subscribers. A larger monthly pool for third-party models |
Claude Max includes Claude Code. Cursor Ultra, for a new subscription, includes Cursor's own models, and Claude on that plan is API price on top of the $200. The Claude plan at $200 is Claude Ultra, and that one includes Claude Code.
Max 20x was the name this post used in August for a $200 Claude plan. The pricing page on 2 November 2026 lists Max at $100 and Ultra at $200.
This is my usage, not the price card. Claude Max is really hard for me to hit. At work, I don't think the Ultra plan ever hit its cap before we moved to a Teams plan. On Cursor Super I am hitting the caps. I am just starting out, which is my guess for why. I like Cursor's usage view.
Annual Claude Pro is $200 billed up front, which the pricing page shows as $17 a month. Annual prices on the current Cursor plans are $80 for Plus, $300 for Premium, $1,000 for Super, and $2,000 for Ultra.
Cursor Teams seats are $30, $100, or $200 a month, matching Premium, Super, and Ultra. Claude Enterprise is $20 per seat per month plus usage at API rates.
On Pro, Fable is paid with usage credits. On Max and Ultra, Fable is included up to half the weekly limit. Claude Code on a paid plan shares that plan's limits with the rest of Claude.
At a limit, the plan help page says you wait for the reset, move to a higher plan, or turn on usage credits at API rates. Max includes $100 a month of API credits, and Ultra includes $200. Signed in with the Claude plan, Claude Code still draws the subscription limits.
The 7 October 2026 update says those credits cover the API, and they cover the Agent SDK and claude -p when those run with an API key from the linked Console organization. The June 15, 2026 split is still paused. Agent SDK, claude -p, and third-party app usage still draw the subscription limits when you are signed in with the plan. The announced mechanics are in Claude Code pricing in 2026.
Cursor's included pool refreshes weekly. When it runs out, you wait, add prepaid credit, or move up a tier. Legacy Pro, Pro+, and Ultra keep two monthly pools, and the third-party pool is drawn down at each model's API price. On-demand usage on those legacy plans is off until you turn it on, and you can set a spend limit. A switch from a legacy plan onto the new plans is optional, and Cursor says the switch is permanent.
An earlier version of this post gave Cursor Pro about 500 premium requests, and treated Auto as outside the pool. The current docs describe usage pools, and the 500-request figure is not on them. Cost mode draws from the pool, on the models and pricing page.
A benchmark here scores the model that was selected
Cursor's docs publish model prices. I have not found an official SWE-bench number for Cursor the product. A score with Cursor's name on it is the score of the model that was selected.
This update quotes no SWE-bench percentage for Opus 5.5, Sonnet 5.5, or Fable 5.1. Comparison posts sometimes cite a blind-test win rate and a token multiple. I looked for the author and the method. I found neither.
I keep Claude Code for the careful pass, and Cursor for the job I send off
This is my judgement. This site is the one piece of work I can put next to both.
Pi built it, then Cursor did. It came out unoptimised, and it performed poorly. I had Claude Code do an in-depth pass.
Later, Cursor refactored the whole site. It did a good job, and the site was much faster. It used a lot of usage. My read is that Claude Code would have used less usage, and it would have cost me more mental effort.
Pick Claude Code if you want the careful pass. That is where the results were better for me, and Max is really hard for me to hit.
Pick Grok Bot, in Cursor, if you want a side project set up, run, and reported back. It eats usage.
Pick the Cursor IDE if the day is hand-coding and small inline edits. That is the Cursor I started on.
Cursor vs Claude Code vs Codex: Codex bills a different account
Codex is OpenAI's coding agent. It has a CLI, an IDE extension, and Codex Cloud.
Codex Cloud runs tasks on OpenAI-managed computers, and a task can keep going while the computer is asleep. Codex itself is included across ChatGPT plans, including Free and Go. Codex Cloud is for eligible Plus, Pro, Business, Enterprise, Healthcare, and Education accounts. Free and Go do not include Codex Cloud.
That split is from OpenAI's ChatGPT plan article and the Codex Cloud article, read on 2 November 2026.
My week, as I described it above, is Claude Code in Orca and Cursor through Grok Bot.
A ticket comes in, and a pull request opens
This is my judgement.
The pipeline I want is a ticket or a bug coming in, and a pull request opening automatically. I can picture that in Cursor. Cursor is heading toward shooting more and more work off. In Claude Code, setting that up would be annoying.
Frequently asked questions
- What is Cursor Ultra vs Claude Code Max?
- On 2 November 2026, Claude Max is $100 a month and includes Claude Code at 5x Pro usage, every five hours and every week, while a new Cursor Ultra is $200 a month for Grok and Composer at 50x Plus, with Claude at API price from prepaid credit. The $200 Claude plan is Ultra, at 10x weekly usage, and it includes Claude Code: https://claude.com/pricing and https://cursor.com/docs/models-and-pricing
- What is Claude Max 20x vs Cursor Ultra?
- The Claude pricing page on 2 November 2026 lists Max at $100 and Ultra at $200. Max 20x was this post's August name for that $200 plan: Claude Ultra includes Claude Code at 10x Pro usage per week, with no five-hour limit on a verified account, and a legacy Cursor Ultra at $200 still has a monthly pool for third-party models: https://claude.com/pricing and https://cursor.com/docs/account/pricing/legacy-cursor-plans
- Is Cursor better than Claude Code for coding?
- On my work, Claude Code is the careful pass, because that is where the results were better for me, and Grok Bot is what I send a side project when I want to be told it is done. Cursor can run Claude as a model, at API price from Premium up: https://cursor.com/docs/models-and-pricing
- Does Claude Code run inside Cursor?
- You can install Claude Code in Cursor. Anthropic publishes an install link, the extension runs in the editor, the claude terminal command is a separate install, and the two subscriptions stay separate: https://code.claude.com/docs/en/vs-code
- Does Cursor use Claude?
- Yes. Cursor lists Claude models from Anthropic, along with GPT and Gemini. On Premium, Super, and Ultra they bill at API list price from prepaid credit, and on legacy Pro, Pro+, and Ultra they draw a monthly third-party pool: https://cursor.com/docs/models-and-pricing
- Do I need Claude Code if I already use Cursor?
- I kept both. Grok Bot is my first stop for a side project, and I go back to Claude Code when the usage runs out or the task is too complex to hand off. That is my split, and it might not work for you.
- Is Claude Code replacing Cursor?
- Claude Code is Anthropic's coding agent, and Cursor is the editor with its own agent. I run Claude Code in Orca, I use Grok Bot from Cursor for side projects, and each subscription stays on its own bill.
- How much do Claude Code and Cursor cost in 2026?
- On 2 November 2026, Claude Pro is $20 a month and includes Claude Code, Max is $100, and Claude Ultra is $200. A new Cursor plan is Plus at $8, Premium at $30, Super at $100, or Ultra at $200, Claude inside Cursor is extra from Premium up, and legacy Cursor Pro at $20 remains for existing subscribers: https://claude.com/pricing and https://cursor.com/pricing
- Is Claude Code expensive?
- Pro includes Claude Code at $20 a month, or $17 a month if you pay $200 for the year, and Max at $100 and Ultra at $200 raise the included usage. Past the limit, usage credits continue at API rates when they are turned on: https://claude.com/pricing
- Are Claude Code and Cursor free?
- Claude Code starts on Pro at $20 a month. Cursor Free includes limited Agent and Chat, and Claude inside Cursor starts on Premium: https://claude.com/pricing and https://cursor.com/pricing
- Does Cursor publish a SWE-bench score?
- I have not found an official SWE-bench number for Cursor the product. A benchmark number belongs to whichever model was run, and this update quotes no Claude leaderboard number.
- Cursor vs Claude Code vs Codex: what is the difference?
- Claude Code bills a Claude plan, Cursor bills a Cursor plan with Claude as an optional model from Premium up, and Codex is OpenAI's agent with a CLI, an IDE extension, and Codex Cloud on OpenAI-managed computers: https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan