Cursor 3 Launches as an Agent Orchestration Platform — The IDE Is Now a Management Console

Cursor shipped version 3 on April 2, 2026, and it is not an IDE update. The company rebuilt its interface from scratch around a new premise: that most code will be written by AI agents, and the developer's job is to direct and review them, not write every line.

The centerpiece of Cursor 3 is the Agents Window — a unified sidebar showing all running agents across local environments, worktrees, SSH sessions, and cloud deployments. Developers can launch new agents from the interface, from Slack, from GitHub, or from Linear, and track them all in one place. Composer 2, an internally trained model optimized for agentic coding, powers the agents and shows meaningfully improved performance on long-horizon tasks compared to the previous Claude-dependent architecture.

Cloud-to-local handoff is seamless: push a task to the cloud for overnight execution, pull it back to your local environment for testing and review. A redesigned diff viewer handles staging, committing, and PR opening without context switching. Design Mode lets agents annotate browser UI elements directly, enabling precise front-end iteration loops.

The launch puts Cursor in direct competition with Claude Code, GitHub Copilot Agent Mode, and Google's Antigravity IDE — all of which are converging on the same agent-first paradigm. Cursor's differentiation is the unified workspace that retains the full VS Code experience as a fallback, ensuring zero forced migration risk for existing users.

Pricing now spans six tiers: Hobby (free), Pro ($20/month, 500 fast premium requests), Pro+ ($60/month), Ultra ($200/month), Teams ($40/user/month), and Enterprise (custom). The Pro+ and Ultra tiers exist primarily for developers running agent fleets continuously — the use case Cursor is explicitly designing toward.