Cursor debuts Composer and multi-agent suite in Cursor 2.0

Cursor 2.0 bundles two separate releases: Composer, the company’s first in-house coding model, and a multi-agent interface built around the model. Composer is pitched on speed over raw capability, with most tasks completing in under 30 seconds and training that leaned on codebase-wide semantic search for navigating large repositories. The interface change is aimed at teams running several agents in parallel, coordinated through git worktrees or remote machines, with a native browser tool letting agents test and refine their own code until it passes.

Fichamento

#

• Composer is Cursor’s first coding model, built for agent-driven generation and editing, claimed to respond up to four times faster than comparable models with most tasks under 30 seconds.
• Its training incorporated codebase-wide semantic search, aimed at more accurate navigation of large, complex codebases; early users reportedly trusted it more for multi-step tasks requiring broad project context.
• The Cursor 2.0 interface is built around parallel agent workflows, managed through git worktrees or remote machines, targeted at teams that juggle multiple projects and want to compare outputs across different models.
• A native browser tool lets agents automatically test and refine their own code until they reach a correct solution, without a human manually re-running tests between iterations.
• Cursor 2.0 is publicly available and supports the major desktop operating systems.