Overview
- Cursor released Cursor 3 on Thursday, April 2, introducing an agent-first interface in its desktop app that lets developers spin up and manage multiple AI coding agents across local and cloud workspaces.
- The new Agents Window centralizes control of running tasks across repositories, and it supports parallel agents so teams can delegate feature builds, bug fixes, and reviews without juggling separate chats or terminals.
- Cursor 3 enables cloud-to-local handoff for long-running jobs, with cloud agents generating demos and screenshots for review so tasks keep progressing even when a laptop is closed.
- The release ties into a growing ecosystem, with Marketplace extensions and Model Context Protocol tools that pass structured data to agents, including examples that connect Apidog so agents read live API specs and generate schema-accurate code.
- The rollout comes as Cursor faces pressure from Anthropic’s Claude Code and OpenAI’s Codex, and trust questions linger after reports that its low-cost Composer 2 model closely tracked Moonshot AI’s open Kimi 2.5 without clear early disclosure.