Knowing which AI feature fits which job (typing, a small edit, a multi-file task) is the core skill of using Cursor well.
Cursor is a code editor, the program you write code in. It is built from the VS Code codebase (a fork), so the layout, extensions and keyboard shortcuts feel familiar if you have used VS Code. What it adds is AI at three levels of size:
- Tab: as you type, Cursor predicts your next edit, sometimes several lines or a change further down the file, and you accept it with the Tab key. Smallest and fastest.
- Inline Edit (Cmd+K on Mac, Ctrl+K on Windows/Linux): select some code, describe a change, get a diff in place.
- Agent (Cmd/Ctrl+I): a side panel where an AI agent takes a task in plain language, searches your project, edits several files, runs terminal commands and reports back.
Around these sit features for giving the AI the right information: @-mentions of files, an index of your codebase for meaning-based search, and rules files with your project's conventions. There are also Cloud Agents, which run tasks on remote machines, MCP for connecting outside tools, and a command-line version.
Cursor offers its own models plus models from other AI companies, and you choose per task. Paid plans unlock more models and usage; plans change, so check the current pricing page rather than memorizing it.
What it does not do: guarantee correct code. The models can misunderstand or invent APIs that do not exist. Your job shifts from typing every line to describing work clearly, reviewing diffs and running checks. This curriculum goes from setup and the three core features to rules, context, safety settings and team workflows, as of Cursor 3.x.
Remember
- Cursor is a VS Code-based editor with AI built in
- Three sizes of help: Tab (typing), Inline Edit (a selection), Agent (multi-file tasks)
- You still review and verify everything the AI produces
Watch out for
- Using Agent for one-line changes Tab or Inline Edit would do faster, or Tab for work that needs a plan.