Cursor, an AI-powered coding assistant and development environment, has introduced a new feature called "Projects" in beta. This tool is designed to help developers manage complex, long-term programming tasks by using a coordinator agent that plans tasks and assigns them to sub-agents working in parallel. Projects are hosted on cloud servers and can launch a local agent on a user's computer when specific tests need to be run locally. These projects maintain a shared context across all their agents and can integrate with tools like Slack, calendars, or GitHub to respond to events automatically. The deployment of Projects has recently begun. Cursor is developed by Anysphere, Inc., a company based in San Francisco and a subsidiary of SpaceXAI. Anysphere was founded in 2022 and became fully owned by SpaceXAI in August 2026. Cursor itself was acquired by SpaceXAI in June 2026. The tool gained popularity in 2023 after its launch and saw widespread attention in 2024, though some critics argue its capabilities are overstated. Cursor is based on Visual Studio Code and integrates large language models like GPT and Claude to assist with code generation, autocompletion, and rewriting. A solo developer who has been using Cursor since its launch shared his experience while building a SaaS application. He described the public debate around AI coding as polarized—some calling it revolutionary, others dismissing it as unreliable. However, he found the reality to be somewhere in the middle. He noted that creating a ".cursorrules" file, which provides specific instructions for how Cursor should generate code for a project, was a key change for him. He also found that Cursor performed better when it could reference existing code, comparing the tool to a fast but somewhat inexperienced junior developer who needed guidance. Cursor’s new Projects feature is intended to help developers tackle larger tasks, such as building a feature, migrating a system, or creating an entire application. These projects maintain context over long periods, delegate tasks to multiple sub-agents, and handle recurring tasks automatically. The coordinator agent oversees the work, planning tasks and assigning them to sub-agents that execute the code, then returning the results for verification. This setup allows for continuous progress without constant user input. As the coordinator learns more about the project, it becomes more effective at managing tasks and improving outcomes. Projects are still in beta and will be gradually rolled out to all users.