I have the same version of Cursor running on both pc’s, and only want to use each machine one at a time.
I set up a MKLINK to move the .cursor and the Cursor folders from my user to a different location on the same machine then set up a sync job to sync these two folders plus the project folder to my NAS. All works fine and my source Cursor keeps everything running as before.
On the target machine, I did the same MKLINK with the .cursor and Cursor folders and then sync’d from my NAS to the target machine.
Everything is still working and on the target machine I can see the project but not any of the conversations.
Has anyone got any pointers why this might be? I home there is no ‘machine level’ checking going on that is stopping this from working.
Hey thanks for the question. There is no machine level check that blocks this. The reason is how history gets stored.
Chat and composer history lives locally on each machine and ties to the MD5 hash of the absolute path to the project folder. Cursor recalculates this hash from the path where the project opens. If the target machine opens the project with even a slightly different absolute path like a different Windows username in C:\Users\... or a different drive letter or location the hash changes. Cursor then looks in a different workspaceStorage subfolder and finds nothing. The chats are not lost they are just orphaned under the first machine’s hash.
Two things to make this work.
Open the project with the exact same absolute path on both machines. Keep the drive letter and user path identical.
state.vscdb is a live SQLite database. Copy it only when Cursor is fully closed on both machines and include the -wal and -shm files too. Copying while Cursor runs can cause truncation or corruption which shows up as missing conversations.
This DIY sync is not officially supported and stays fragile. A more reliable way is to keep important context in files inside the repo like .cursor/rules and AGENTS.md. They travel with the project.