Cursor review

What it is

Cursor is a fork of VS Code rebuilt around AI. Rather than bolting an assistant onto an editor, it assumes the assistant is always present: it indexes your repository, keeps context across files, and can plan and apply changes across a codebase while you review.

Where it shines

Repository awareness is the difference. Ask why an authentication flow behaves oddly and it traces the actual call path through your files rather than guessing from the file you have open. Ask it to rename a concept across a project and it finds the twenty places it appears, including the ones you forgot.

Agent mode is the headline feature in 2026. Describe a change, and it produces a plan, edits multiple files, runs the test suite, and iterates on failures — with every diff presented for approval. On well-structured codebases it lands substantial features unattended.

Tab completion deserves separate mention: it predicts your next edit anywhere in the file, including deletions and moves, so accepting suggestions repeatedly walks you through a refactor. It is the feature people miss most when they go back to another editor.

Where it falls short

Usage limits. Agent work consumes far more compute than completions, and heavy users exhaust the $20 tier well before month end — the $200 Ultra plan exists because that ceiling is real. Budget accordingly.

It also requires changing editors, which some teams and IDE-locked ecosystems will not do. And agents that can edit twenty files require discipline: review diffs properly, keep changes on branches, and never let an agent commit unreviewed.

Pricing in plain English

Free Hobby tier with a Pro trial. $20 a month for Pro, which suits most developers. $200 for Ultra. Around $40 per seat for Teams with SSO and enforced privacy mode.

Who should use it

Professional developers working in large existing codebases, and anyone doing frequent multi-file refactors or framework migrations. Stay with Copilot if your team cannot change editors.