Cursor opens multiplayer Projects, with separate agents for each teammate
Cursor announced a multiplayer version of Projects on November 10, allowing colleagues and their coding agents to work within a common project. The company describes the release as a beta rollout, extending a product previously organized around an individual directing a coordinator and its workers. Cursor announcement
The practical access requirements are narrower than the announcement’s broad rollout language. Cursor’s documentation says sharing requires a team account and version 3.26.13 or later. Invitations stay within the same Cursor team; people on personal accounts or another team cannot join. Each member retains a separate conversation, coordinator and workers, with activity billed to that member. Cursor documentation
Members share a task board and project context. When one coordinator needs information from another person’s work, it can contact that person’s coordinator rather than relying on a human to carry the question between conversations. Cursor says these exchanges are identified as agent input and treated as untrusted, rather than as instructions from the receiving agent’s user. Cursor documentation
The separation extends to personal information. The documentation says personal stores are not visible to teammates’ agents, while project context is shared. Requests to disclose personal files or connected-source data require approval. Responses to another coordinator take place in a side conversation with restrictions, including a prohibition on sharing credentials. These are vendor-described controls, not findings from an independent security assessment. Cursor documentation
Projects itself debuted in September. Its original design placed planning and delegation with a coordinator, while implementation went to workers. The project runs on cloud infrastructure, allowing work to continue after a user closes a laptop. That history matters: this week’s announcement adds collaboration among people and their agents, rather than introducing persistent agent projects for the first time. September launch
The change puts a familiar team-management problem into an AI interface. Separate developers can ask for work that overlaps, depends on unfinished code or changes a shared assumption. A common board and direct coordinator messages could reduce the manual work of keeping those requests aligned. Whether they do so reliably will depend on how the system handles conflicting priorities, stale information and permission boundaries in real projects.
For teams evaluating the beta, the useful test is therefore larger than code generation quality: can several people stay informed and retain control while their agents coordinate a shared delivery?
Sources and citations
Revision history
Published .