I am using the latest version of Cursor 3.1, and a few updates before I started getting this bug where I couldnt work with Source control or agents when I open a project in a new Window.
Steps to Reproduce
I open Cursor IDE, with in a new Window local repo that I worked on previously.
When I go to Source Control, it shows No source control providers registered.
The agents are also not working and a popup suggestion appears to reload the Window but it rarely works.
Have you tried clearing the editor history by running Cmd/Ctrl+Shift+P? This resets cached state that may be causing the Source Control providers to fail to load.
Hey, this is a known issue with the extension host on macOS. It doesnāt initialize when you open a new window, so both source control and agents stop working. A few users have been hitting this since 3.0.x: Cursor last update freeze on macOS
A couple questions to narrow it down:
What exact macOS version are you on? Apple menu > About This Mac
Do you have any MCP servers set up? Cursor Settings > MCP
Also, try this:
Launch Cursor from Terminal with cursor --disable-extensions to see if extensions are the cause
If it still happens, try clearing the cache. Delete ~/Library/Application Support/Cursor/Cache and restart
Also check Console.app for Cursor crash reports. If you find any, paste them here
The team is aware of this issue. No timeline yet, but your report helps us prioritize.
Yes, Iāve been getting this issue for weeks. Itās really frustrating. I also found deleting the cache worked for a few days but the issue reappeared.
My colleague and I are both plagued by this issue for weeks now, including today. Iām on Tahoe 26.4.1 and Cursor 3.2.11. It is intermittent and a few tries at restarting Cursor can often get Cursor working. I havenāt tried deleting cache or running without extensions but will try that next time. Just registering that on current OS and current Cursor this still occurs. Note that both of us often have 5+ windows open.
Hey @gregfunky@Alex_Fox@Suvesh, thanks for following up about the regression. We know this symptom and the extension host pattern is on our radar, but we really want to nail down why itās showing up again on 3.2.x, especially on Tahoe and with lots of windows.
To catch it in logs, next time it happens please donāt restart Cursor right away. Do this first:
Cmd+Shift+P ā Developer: Export Logs
From the exported zip/folder, we need the Window, Main, and Extension Host sections. If you use Remote SSH, include that too.
Also check Console.app and search for Cursor for the same time window. If thereās a crash report, please attach it.
In the meantime, a few workarounds besides clearing ~/Library/Application Support/Cursor/Cache:
Run from Terminal: cursor --disable-extensions
If the issue doesnāt reproduce in this mode, an extension is likely the cause.
In Activity Monitor, check the Cursor Helper (Plugin) and extension host processes. Sometimes they get orphaned and stick around from old windows. Killing all Cursor processes before restarting can help.
@gregfunky, about 5+ windows: this might be related to orphaned extension host processes piling up. If the issue doesnāt come back for at least a while after fully killing all Cursor processes via Activity Monitor, not just Cmd+Q, thatās a strong signal. Please mention that in your reply.
I canāt share an ETA for a fix yet, but the logs will really help us match your case to the right issue.
Ive also been able to fix the issue by doing this:
I open Cursor and the issue is present. I close all my opened files and opened Agent Chats. I go to New Window and pick the same project that I wanted to work on. It usually resolves in about 10s max.
Iāve been having this same issue, seems to be getting worse. No repos come up, canāt go to settings in the editor window, no files load.
@deanrie I donāt get Developer: Export Logs as an option. I have it happening regularly on 2 MBPs, after restarting the editor window a few times it usually resolves, but this morning it seems to be fully stuck. Iāll clear cache for one of my setups and keep the other in itās current state in case thereās anything else I can provide.
I found this in Console.app crash reports, but doesnāt happen on every load so may be unrelated:
I just encountered this in one window (window5 out of 6 windows) but not any other. InterestinglyCmd+Shift+P ā Developer: Export Logs was not available in window5 but was available elsewhere. I opened the logs under Application Support and think I have the logs for main and window5 (the one that was acting up). Since they seem to lack PII I will upload here.
Also of interest, when I tried to open extension logs for that window and it failed with the following error (this did work for other windows where it wasnāt failing) and I can confirm there is no exthost folder under window5 logs directory.
Error:
Command āDeveloper: Open Extension Logs Folderā resulted in an error
Unable to resolve nonexistent file ā/Users/gfunk/Library/Application Support/Cursor/logs/20260429T140113/window5/exthostā
The renderer.log has lots of more but these stood out:
2026-04-29 14:01:46.015 [warning] [WorktreeManager] Timed out waiting for git context provider, skipping worktree discovery
2026-04-29 14:03:45.414 [error] Error fetching user pricing info: No Connect transport provider registered.
2026-04-29 14:03:45.416 [error] Error fetching user privacy mode: No Connect transport provider registered.
2026-04-29 14:03:45.416 [error] No Connect transport provider registered.: Error: No Connect transport provider registered.
at LBc.transport (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:110724)
at async nD.createSingleServer (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:112251) No Connect transport provider registered.
2026-04-29 14:03:45.950 [error] No Connect transport provider registered.: Error: No Connect transport provider registered.
at LBc.transport (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:110724)
at async nD.createSingleServer (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:112251)
at async nD.get (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:111457)
at async ITm.keepCppModelStateUpdated (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:46410:16293)
at async ITm.loadCppConfigIncludingHandlingProAccess (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:46420:1259) No Connect transport provider registered.
2026-04-29 14:03:46.142 [error] No Connect transport provider registered.: Error: No Connect transport provider registered.
at LBc.transport (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:110724)
at async nD.createSingleServer (vscode-file://vscode-app/Applications/Cursor.app/Contents/Resources/app/out/vs/workbench/workbench.desktop.main.js:29322:112251) No Connect transport provider registered.
Would like to iterate that Anastasijaās response worked, but I did get the same problem again 10 minutes later. Closing all code windows to have it work again is a bit frustrating as I tend to have multiple coding sessions open. Please prioritise this
Iāve got the same problem! Iāve spent hours and loads of tokens trying to fix the issue with the 5.5 xHigh.
Symptoms:
Source Control panel did not detect the repository.
git status worked normally in terminal.
Git binary was valid, for example /opt/homebrew/bin/git.
A clean Cursor profile worked, but my normal profile did not.
Cursor logs did not contain window*/exthost/vscode.git/Git.log.
Renderer logs showed lines like:
Started local extension host with pid ā¦
No search provider registered for scheme: file, waiting
The important clue was that the local extension host was starting but never fully bootstrapping. Because of that, built-in extensions like vscode.git never activated. So the real problem was not Git itself. Git was only a symptom.
In my case the cause was corrupted or bad per-workspace Cursor state:
~/Library/Application Support/Cursor/User/workspaceStorage//state.vscdb
That SQLite DB stores local UI/workspace state for a specific folder. If it gets corrupted or contains bad state, Cursor can hang before the extension host initializes. Then Git, file search, and other extension-based features may not work.
Fix
Close Cursor completely first.
Then find the workspace folder: for d in ā$HOME/Library/Application Support/Cursor/User/workspaceStorageā/; do if [ -f ā$d/workspace.jsonā ]; then echo ā== $d ==ā cat ā$d/workspace.jsonā fi done
Look for the entry matching your project folder.
Then back up and remove that workspace DB: WORKSPACE_ID=āā BASE=ā$HOME/Library/Application Support/Cursor/User/workspaceStorage/$WORKSPACE_IDā BACKUP=ā$HOME/Cursor-workspace-state-backup-$(date +%Y%m%dT%H%M%S)ā mkdir -p ā$BACKUPā mv ā$BASEā/state.vscdb ā$BACKUPā/
Reopen Cursor normally.
Cursor will recreate a fresh state.vscdb. This does not touch your repository, .git, global Cursor settings, extensions, or chats. It can reset workspace-local UI state such as open tabs, search history, terminal buffer, and some layout state.
If many workspaces are affected, you can back up all workspace-local state DBs: BACKUP=ā$HOME/Cursor-all-workspace-state-backup-$(date +%Y%m%dT%H%M%S)ā mkdir -p ā$BACKUPā for f in ā$HOME/Library/Application Support/Cursor/User/workspaceStorageā//state.vscdb; do rel=ā${f#$HOME/Library/Application Support/Cursor/User/workspaceStorage/}ā mkdir -p ā$BACKUP/$(dirname ā$relā)ā mv ā$fā ā$BACKUP/$relā done
After reopening Cursor, check logs for successful Git activation:
ExtensionService#_doActivateExtension vscode.git
Extension activated success: vscode.git
Using git āā¦gitā¦ā
Opened repository: ā¦
In my case this fixed the issue without resetting the full Cursor profile and without touching the repo.
Thanks to everyone who sent logs and steps, it really helps.
@gregfunky: got your logs and dug in. The timeline is pretty clear:
At 14:01:15 all 4 local extension host processes started normally pids 46142, 46157-46159
30 seconds later we see [WorktreeManager] Timed out waiting for git context provider which means the main exthost pid 46142 started, but vscode.git never activated
Starting at 14:03:45 there is a repeating No Connect transport provider registered every 5 minutes while polling pricing info, privacy mode, and cpp config and it keeps going for almost 2 hours nonstop
The missing exthost/ folder in window5/logs/ while other windows have it is the same signal. The process is alive, but bootstrap never finished. This matches the IDE-1923 pattern of a wedged extension host. I added these logs to the existing issue.
@Joshua_Truscott: about logs when Developer: Export Logs is not available because the window is stuck, you can grab them manually. Fully quit Cursor, then go to:
We need main.log and renderer.log for the broken window, plus exthost/ if it exists. Please attach them here. About the zsh crash, based on the stack segfault in getlinknode ā deletefilelist via zhandler/zexit, it looks like zsh crashed on the child side fork pre-exec during exit cleanup. That is a separate shell process, not the extension host itself. Thanks for sharing it, but it is probably not related to the main issue.
@Omar_Shoura@Nizar_Khalili: while we keep investigating, please try the workaround that @ivnmood posted above, deleting the per-workspace state.vscdb for the specific project. Our guess is a corrupted workspaceStorage SQLite blocks exthost bootstrap, which matches what we see in gregfunkyās logs. If it helps, let us know. That will help confirm the root cause.
I canāt give an ETA for a fix yet, but we now have a solid timeline and a stack trace to work with. If your workaround lasts longer than Anastasijaās recipe, or if the issue comes back, please post an update too.