I have some feedback on the Remote Control feature. This falls into 2 categories.
1. Spinning up local agents
I’ve been playing with Remote Control via web and now via iPhone app and one thing that I really lack is the ability to spin up new local agents remotely. I expected to see all my Cursor Desktop workspaces / projects in web/app, as well as recent chat history (ideally), be able to select a project and just start an agent remotely (or continue previous session, just like I’d normally do on desktop). Currently it seems I could only spin up cloud agents, or explicitly handoff local agents to Remote Control.
I understand that explicitly initiating Remote Control handoff might be a security thing. But I think a good solution might be a local password or a key that user sets up in their Cursor Desktop, and it “locks up” the Desktop sessions after eg. 1h of inactivity. So Cursor Desktop is always there ready to go, but it would only let the user connect if the user unlocks it with their security key. This key would never be stored on Cursor servers - the user would enter it on the web / iPhone app (eg. via saveable password field or even a Passkey). This way even if Cursor’s infra gets compromised, the blast radius will be minimal, while providing users the ability to effectively work with their local projects remotely.
2. Git requirement
It seems that Remote Control option doesn’t even come up if the project root is not a git repo (or a workspace, but that is even more cumbersome). Why would Remote Control require a git-based repo? Eg. I often use Cursor in non-dev-project folders, eg. for files / documents organization, video editing, and other uses that do not require versioning.
Hope this makes sense, as I would really like to start using my local Cursor remotely in its entirety.