Cursor

What Cursor does

Cursor is a code editor built on the Visual Studio Code foundation with AI woven through it rather than added as a panel. It offers inline completion that predicts multi-line edits, a chat sidebar aware of your codebase, and an agent mode that plans and applies changes across many files while running terminal commands. Rules files let a repository specify conventions the assistant must follow, indexing gives the model project-wide context, and multiple underlying models can be selected per task. Because it is a fork of a familiar editor, extensions, keybindings and settings mostly carry over. It suits developers who want AI assistance without leaving a graphical editor, and teams standardising on shared rules. The limitations are pricing and drift: usage-based billing on top of a subscription can surprise heavy users, and agent edits across many files sometimes need reverting wholesale when the approach turns out wrong.

How to use it

Download Cursor from cursor.com and import your existing Visual Studio Code settings during setup, which takes a minute and makes the editor feel familiar immediately. Open a project and let it finish indexing before expecting good answers, as chat responses are noticeably worse mid-index. Day to day you use three things: tab completion for the next edit, the inline prompt for a change in the current file, and the agent for anything crossing file boundaries. Add a rules file to the repository describing stack, conventions and forbidden patterns, then reference specific files in chat with the mention syntax instead of hoping the index surfaces them. Commit before running the agent so you can discard a bad run cleanly. Billing is the usual stumbling block: plans include a request allowance and heavier models draw from it faster, so check which model is selected before a long session, and watch for usage-based charges once the included allowance is spent.

Best for

• daily development
• multi-file edits
• codebase questions
• rapid prototyping
• editor-based workflows

Strengths and limitations