Hey, thanks for the report. This is a known issue, especially on macOS Tahoe after updating to 3.0.13. The team is aware.
Let’s try a couple things to help diagnose it:
Check the contents of cursor-socket:
ls -la "/Applications/Cursor.app/Contents/Resources/app/extensions/cursor-socket/"
If you see a dist/ folder instead of out/, create a symlink:
cd "/Applications/Cursor.app/Contents/Resources/app/extensions/cursor-socket"
ln -s dist out
Restart Cursor.
If that doesn’t help, try launching from Terminal:
cursor --disable-extensions
Check if you have MCP servers set up: cat ~/.cursor/mcp.json. If the file exists, paste the contents here.
Try opening a new agent window instead of the current one. For some users this helps as a temporary workaround.
If none of that works, a reliable option is to roll back to 2.6.22. A few users with the same Tahoe bug confirmed 2.6.22 is stable: Cursor last update freeze on macOS
Hey, if even 2.6.22 didn’t help, it’s most likely not the version. It’s your local state. Let’s try a full cleanup and collect logs.
First I need:
Your exact macOS version. Apple menu > About This Mac. I need the line that looks like 26.x
The latest file from ~/Library/Application Support/Cursor/logs/main.log. Please attach it here
The output of cat ~/.cursor/mcp.json. If the file doesn’t exist, just say so
While you’re collecting that, try a full state cleanup. Before you do it, back everything up just in case:
# 1. Fully quit Cursor using `Cmd+Q`, and make sure there are no Cursor processes in Activity Monitor
# 2. Delete Cursor.app
# 3. Clean the state:
rm -rf ~/Library/Application\ Support/Cursor
rm -rf ~/Library/Caches/Cursor
rm -rf ~/Library/Caches/com.todesktop.230313mzl4w4u92
rm -rf ~/.cursor
# 4. Reinstall the latest from https://cursor.com/downloads
If the issue is still there after a clean install of 3.0.13, try installing 2.6.22 again on top of the already cleaned state. What usually works is clean state plus older version, not just a downgrade.
And please still send main.log. It usually shows what’s crashing exactly like SIGTERM on processes, slow DB close, and so on. That speeds up troubleshooting a lot.