Cursor

Sources checked: 2026-11-13

What Cursor does

Cursor brings an agent workflow into a code editor. The official quickstart walks through opening a project, requesting a change and reviewing its diff. This is a useful starting point if you want to read and steer the agent's work beside the affected code.

Evaluate it on a task with an observable outcome. Choose a defect or a small feature you understand well enough to review, and write down the acceptance check before starting.

Current and legacy plans

The current public price list offers Free, Plus at US$8/month, Premium at US$30, Super at US$100 and Ultra at US$200, before tax. Check the checkout for your billing period and account.

The model and pricing documentation distinguishes these plans from legacy Cursor subscriptions. Current paid plans share a weekly allowance across Cursor and Grok; third-party models on Premium and above are charged separately from prepaid credit. Legacy plans retain different usage pools and billing rules. Check which plan you actually have before applying a limit or price comparison.

Where to check limits

Open your account's usage settings and inspect the selected model and spending cap. A subscription price alone cannot predict the cost of your task. For API work outside the editor, the API pricing guide and calculator compare token workloads; they do not estimate Cursor's included allowance.

Install Cursor and open your practice project

Download the desktop installer for your operating system from Cursor, sign in and open a project folder. The quickstart shows Agent access with Cmd+I on macOS or Ctrl+I on Windows/Linux. Check your chosen model and available usage before the first request.

A useful first task

Use a practice project with Python 3 and save its initial state in Git. Create tags.py with this deliberately incomplete function:

def normalize_tags(tags):
    return [tag.lower() for tag in tags]

For [' Rust ', 'python', 'RUST', '', ' Python '], require ['rust', 'python']. The function must trim surrounding whitespace, lowercase, drop empty tags and remove duplicates in first-appearance order. Its input is a list of strings and must stay unchanged. Add check_tags.py from the shared exercise to the same folder.

Ask Agent to read the two files and explain the mismatch before editing. Then paste:

Fix normalize_tags in tags.py to meet the stated requirements.
Change only tags.py; keep check_tags.py unchanged.
Use only the Python standard library.
Run python3 check_tags.py and show the diff and check result.

Replace python3 if your Python 3 command has another name. Review the changed file and run the checker in the project terminal. Accept after it prints 4 cases passed; input unchanged and you can explain the change on another input. The Agent documentation describes its editing and tool workflow.

Keep every attempt in the comparison worksheet, including rejected changes and their review time. For a larger task, use the planning workflow documented in the quickstart before editing.

Repeat the same task in Codex or Claude Code to compare your own workflow. These steps are a suggested trial, not benchmark results. The timeline below keeps earlier announcements and pricing reports in their historical context.