AI programming editor Cursor released version 2.4, and the core updates revolve around "Subagents" and "Image Generation". The new mechanism allows the master agent to split independent subtasks into multiple sub-agents for parallel processing, each using an independent context, and configurable prompts, tool permissions, and models to improve the overall speed of complex, time-consuming workflows and the contextual focus of the main conversation.
This version also adds the ability to generate images directly from the agent: users can use text descriptions or upload reference images to guide the underlying image model to generate content; The generated results are returned as an inline preview and saved to the project's assets/ directory by default, targeting scenarios such as UI prototypes, product assets, and architecture diagrams. The Enterprise Edition also introduces Cursor Blame, which adds AI attribution to the traditional git blame, distinguishes tab autocompletion, agent runs by model, and manual editing, and links each line of code to the corresponding session summary for easy review and traceability.
In addition, the interactive Q&A tool used by Plan and Debug mode has been extended to initiate clarifying questions in any conversation; While waiting for a response, the agent can continue reading the file, modify code, or run commands, and automatically include the next steps when the answer arrives. Improvements and bug fixes are listed in summary form in the changelog, and the details are still subject to the official page.
FAQs
Q: What are the functions of Subagents in Cursor 2.4?
A: Sub-agents are independent agents that can run in parallel, handling different parts of the main task in their own contexts, accelerating complex planning and deep codebase exploration.
Q: Where does Cursor 2.4's image generation save images?
A: Images are returned as an inline preview and saved to the project's assets/ folder by default for direct reference in the UI and product assets.
Q: Which users can use Cursor Blame and what problems can Cursor Blame solve?
A: Cursor Blame is geared towards Enterprise and enhances git blame with AI attribution, distinguishing between AI and human sources, and linking to session summaries that generate code to assist in code review.
Q: How is the Agent's clarification question different from before?
A: Agents can now ask clarification questions in any conversation and continue to perform parallel read, edit, and command tasks while waiting for a response.
