3.0
CursorSummary
Cursor 3 introduces the powerful new Agents Window, enabling users to run multiple agents in parallel across various environments with improved visibility and control. This release also adds Design Mode for precise UI targeting and Agent Tabs for managing multiple chats simultaneously. Several usability improvements focus on performance, such as faster large-file diff rendering, alongside key bug fixes across input handling and workspace loading.
New Features
- Introduced the new Agents Window interface allowing users to run many agents in parallel across repos and environments (locally, worktrees, cloud, SSH).
- Enabled Design Mode within the Agents Window to annotate and target UI elements directly in the browser for more precise feedback.
- Added keyboard shortcuts for Design Mode: → ⌘+Shift+D to toggle, Shift+drag to select an area, → ⌘+L to add element to chat, and → Option+click to add element to input.
- Added Agent Tabs in the Editor to view multiple chats side-by-side or in a grid.
- Added a new command /worktree that creates a separate git worktree for isolated changes.
- Added a new command /best-of-n that runs the same task in parallel across multiple models in isolated worktrees and compares outcomes.
- Enabled MCP Apps to support structured content for richer tool outputs.
- Added directory group name to audit logs for improved human readability.
- Added a team-level Admin setting to restrict creating, editing, and deleting team secrets to Admins.
- Added an Enterprise Admin control for disabling "Made with Cursor" code attribution for the entire team.
Bug Fixes
- Fixed text area behavior for Network Access Controls so pressing Enter reliably adds a newline at the end of the input.
- Fixed hooks loading so multi-root workspaces correctly read project hook files from all workspace folders instead of only the first one.
- Fixed a markdown parsing bug where parenthesized HTTP(S) links could be misread as citations.
- Fixed todo card visibility to prevent them from disappearing after all todos complete.
- Fixed Composer queued prompts that were not resuming automatically after editing operations.
- Fixed picker behavior for disabled but selectable models by removing misleading styling and auto-enabling the model upon user selection.
- Fixed a bug where expanding/collapsing thinking blocks didn't work while streaming was still in progress.
- Fixed a bug where Shift+Enter line breaks weren't treated as multiline content, preventing the prompt input field from staying in an incorrect state.
Improvements
- Moved native worktree support from the Editor to the new Agents window with improved visibility and UX for managing multiple workspaces.
- Deprecated the previous worktree and best-of-n selection commands from the Editor.
- Removed cloud agents from the Editor interface.
- Set third-party plugin imports to default to off for Enterprises when unset, respecting explicit Admin overrides.
- Made large-file diff rendering much faster, smoother, and less memory-heavy.
- Reduced the browser automation tool surface and tightened the subagent to use browser tools only, improving focus.
- Improved browser instructions to reduce error loops and added screenshot-based coordinate clicking as a fallback for unreliable DOM interactions.
- Included plans alongside the transcript in shared chats.
- Added caching to improve Explorer subagents startup time.
- Past chat transcripts are now surfaced directly in at-mention search results.
- Added a "scroll to bottom" button in the agent panel that appears when content overflows.
- The tab bar can now span the full available width in maximized chat layouts.
- Consolidated the Early Access release track behind Nightly.