Cursor Launches Composer Coding Model with New Multi-Agent Interface
SAN FRANCISCO, CA – Cursor, a developer tool company, today released version 2.0 of its platform, introducing a new coding model called Composer alongside a multi-agent interface designed to streamline software development. The update aims to provide developers with more control and flexibility in utilizing AI for coding tasks, allowing for parallel execution and comparison of different models.
The release comes as developers increasingly explore AI-assisted coding tools, but frequently enough face challenges with cost and performance compared to established large language models. Cursor’s approach addresses these concerns by offering Composer, trained on interactive development challenges rather than static datasets, and enabling users to run multiple AI agents concurrently-leveraging git worktrees or remote machines-to identify the optimal solution. This multi-agent functionality allows developers to compare Composer’s output against models like Anthropic‘s Claude, mitigating the risk of relying on an unproven system.
cursor trained Composer on a range of agentic tasks, focusing on interactive development challenges. while the model’s performance remains to be seen in broader adoption, the company hopes it will match the accuracy and adherence to best practices of leading AI coding assistants.
Early feedback from developers indicates that while Composer isn’t ineffective,its cost is a barrier given the perceived capability gap with larger models. However, the multi-agent interface provides a pathway for developers to evaluate Composer’s performance directly within their workflows.
Version 2.0 also includes a range of other new features and fixes, detailed in the full changelog. The update positions Cursor to capitalize on the growing demand for AI-powered development tools, offering a unique approach to model selection and task execution.
