New Cursor 3 ditches the classic IDE layout for an "agent-first" interface built around parallel AI fleets

Key summary
Cursor 3 rebuilds the UI into an agent-first experience focused on running multiple AI agents in parallel that are intended to write most of the code.
A unified sidebar displays all agents (local or cloud) running across multiple repositories and workspaces, while the traditional IDE layout remains available.
Agents can be launched from desktop, mobile, web, Slack, GitHub, and Linear; cloud agents auto-generate demos and screenshots, can persist after laptops close, and cloud sessions can be dragged to local machines or vice versa.
New workflow features include a diff view, built-in staging/commit/PR management, an integrated browser for agents to interact with local sites, and access to Composer 2 coding model with high usage limits.
Cursor positions Version 3 alongside Anthropic's Claude Code and OpenAI's Codex, offers a plugin marketplace with hundreds of extensions, and frames software development as entering a 'third age' driven by fleets of agents.



