Cursor 2.1 stops guessing. When you ask it to plan, it now asks you questions first — and gives you an interactive UI to answer them before it writes a single line.
The 2.1 changelog is short on detail but names three headline features:
- Improved Plan Mode — Cursor responds with clarifying questions before generating a plan. An interactive UI surfaces inline so you can answer without leaving the editor. The result: plans that match what you actually meant, not what the model assumed.
- AI Code Review in Editor — Code review powered by AI, built directly into the editing surface. No separate tool, no copy-paste into a chat window.
- Instant Grep — Fast search across your codebase, integrated into the Cursor workflow.
The plan-mode change is the most interesting. Every AI coding tool struggles with ambiguous prompts — the model picks an interpretation, writes 200 lines, and you discover three minutes later it went the wrong direction. Cursor’s fix is structural: force a clarification step with a real UI, not a chat back-and-forth. It’s closer to how a senior engineer would handle a vague spec — ask two pointed questions before touching the code.
AI code review inside the editor puts Cursor in direct competition with GitHub Copilot’s pull-request reviews, but at a different point in the workflow. Copilot reviews after you push. Cursor reviews while you’re still editing. Whether that’s better depends on your team’s process, but catching issues before a commit beats catching them after.
NOTE
The changelog entry is unusually sparse — no version-specific model details, no pricing changes, no API updates. Expect Cursor to fill in docs over the coming days.
Available now in Cursor 2.1. Update from the app or grab it at cursor.com/changelog/2-1.