Cursor debuts Composer and multi-agent suite in Cursor 2.0

Cursor has launched two major developments: its first coding model, Composer, and a multi-agent interface as part of Cursor 2.0. Composer is designed for developers who need rapid, agent-driven code generation and editing. According to the company, the model delivers responses up to four times faster than similar models, with most tasks completed in under 30 seconds. Composer’s training incorporated codebase-wide semantic search, which supports more accurate navigation and understanding of large and complex codebases. Early users have noted an increase in trust for multi-step coding tasks, especially in scenarios requiring quick iteration and broad project context.

Cursor 2.0’s new interface focuses on agent-centric workflows, allowing multiple agents to operate in parallel, managed via git worktrees or remote machines. This setup is intended for professional and advanced software teams who regularly juggle complex projects and value the ability to compare outputs from different models. The update also brings tools to streamline code review and test cycles, including a native browser tool that enables agents to automatically test and refine their code until a correct solution is reached. Cursor 2.0 is publicly available for download and supports major operating systems.

Cursor, the company behind these releases, has built its reputation on developing AI-driven coding tools that prioritize productivity for software engineers. With this new release, it aims to address workflow challenges seen in modern development teams, particularly the need for rapid iteration, reliable code review, and efficient automated testing.

Source