Cursor has launched version 3.0, a complete redesign that transforms the popular AI coding editor from a souped-up IDE into a multi-agent orchestration platform. Developed under the codename "Glass," the update is Cursor's direct answer to Anthropic's Claude Code and OpenAI's Codex.

From Editor to Agent Workspace

The centerpiece is the new Agents Window, a dedicated interface built from scratch around agent workflows. Developers can now spin up multiple AI coding agents in parallel - running across local machines, Git worktrees, cloud sandboxes, and remote SSH environments simultaneously.

A new /best-of-n command lets developers run the same task across multiple AI models in parallel and pick the best output. The /worktree command creates isolated Git branches instantly, keeping experimental agent work separate from the main codebase.

Cloud-Local Handoff

Cursor 3 introduces seamless transitions between local and cloud agents. Developers can move a session to the cloud when closing their laptop, or pull cloud work back to local for hands-on editing. Cloud agents now produce screenshots and demos of their work for human review.

The Competitive Landscape

The $29.3 billion startup faces increasing pressure from the very AI labs whose models it relies on. "A lot of the product that got Cursor here is not as important going forward anymore," said Jonas Nelle, Cursor's head of engineering, to WIRED. The new interface retains full IDE capabilities while adding agent-first workflows, letting developers choose their preferred mode.

Cursor 3 is available now as an in-app update across desktop, web, and mobile.