Cursor 2.0: Composer and a multi-agent interface for faster agentic coding
Cursor released two updates that shift the UI and model stack toward agent-centric coding workflows: a new coding model called Composer, and a redesigned interface to run many agents in parallel.
Composer: a low-latency coding model
Composer is introduced as a frontier coding model focused on speed and responsiveness. The company reports Composer is roughly 4x faster than similarly intelligent models, and the model completes most turns in under 30 seconds, enabling quicker iteration during agent-driven coding sessions. Composer was trained with a set of tools that includes codebase-wide semantic search, intended to improve the model’s ability to understand and work across large repositories.
A key aim for Composer is low-latency, multi-step coding: the model’s speed is highlighted as enabling more fluid back-and-forth when agents are asked to perform complex changes across a codebase.
More on Composer training and capabilities is available at the developer writeup: https://cursor.com/blog/composer
The multi-agent interface and parallel workflows
The product update centers the interface around agents instead of files, making agents the primary unit of work while still allowing inspection of files or a return to the classic IDE when deeper edits are needed. The interface is designed to run many agents in parallel without interference, using git worktrees or remote machines to isolate concurrent attempts.
Running multiple agents against the same problem and selecting the best output is presented as an approach that can improve results on harder tasks. The new layout emphasizes quick review of agent-generated changes and easier navigation from high-level outcomes down into specific files.
Built-in review and testing tools
Cursor 2.0 addresses two emerging bottlenecks in agent-driven development: code review and automated testing of changes. The update includes features to make it simpler to review changes produced by agents and a native browser tool that lets agents test their work and iterate until they reach a correct result. These additions aim to shorten the loop between generation, verification, and refinement.
Availability
Cursor 2.0 is available for download at cursor.com/download. A full list of changes and detailed notes are in the changelog.
Original source: https://cursor.com/blog/2-0
