Chat, Agent, and all other features work fine, only Tab is affected.
Cursor support (ticket T-C49516) identified a subscription sync issue on their end and resynced my account, which fixed it temporarily. However it broke again shortly after.
Issue 2: Codebase indexing permanently stuck on “loading…”
The codebase index never finishes — stuck on “loading…” for all projects, including local ones with no SSH involved.
Getting [not_found] on every Tab request while Chat and Agent still work usually means an entitlement desync on the Tab service side. Your subscription is active in the database, but the Tab backend can’t see it. The previous resync on ticket T-C49516 fixed this, but it looks like it got out of sync again. The Indexing being stuck on “loading…” is likely a side effect of the same auth state. Anything that depends on auth isn’t authenticating correctly.
The client steps you already tried like wiping %APPDATA%\Cursor, reinstalling, turning off HTTP/2, signing out and back in won’t help here. This needs a backend fix.
Best next step: reply in the existing email thread for T-C49516 or email [email protected] and reference that ticket, and mention the issue is back. The team can run another resync, and since this is the second time, ask them to also check why the entitlement keeps dropping. Tab and indexing should come back together after the resync.
Let me know if anything is still broken after the resync and we’ll dig deeper.
I was able to fix the issue (I believe?) for about 2 minutes by logging out then back in, but the issue started happening again afterwards.
So far I haven’t had tab completions stop working in the last ~30 minutes after setting my release channel to default, then installing the update, then changing back to nightly and installing that.
It seems to have gone back to not working. I’m not sure exactly what may have caused it, but my Cursor window wasn’t focused for ~30 minutes or so while reading and it no longer works.
It worked for about 20s after reloading cursor from the command palette, but it’s gone back to not working.
I have the same issue. After restarting Cursor, it will work once, and then won’t work again. Same logs, and codebase indexing is also stuck on “Loading…”
Same issue. Also getting the illegal state error on quit again. I’m on the nightly channel, I believe it was also present in stable 3.3 though.
Also doesn’t happen straight away, it works for a minute then errors out. Been also getting some npm task detection errors - it’s a pnpm VueJS monorepo, not my usual stack. Not sure if it happens on other stacks, might be good to know
Version: 3.4.1
Commit: 2a298dd06944a9b9ea541d28225b779fcbcc6200
Date: 2026-05-08T16:05:07.818Z
Layout: editor
Browser: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Cursor/3.4.1 Chrome/142.0.7444.265 Electron/39.8.1 Safari/537.36
I have the same issue since yesterday. Network diagnostics shows everything ok, and Cursor Tab output keep printing “Error in streaming cpp (connect error) [not_found] Error” which is same as op.
I’m using stable version. I have tried change my network, delete ~/Library/Application Support/Cursor, quit and relogin, but didn’t fix it.
Hey, thanks everyone for the reports in this thread. I can see this isn’t just one user issue like it looked from the first post. Everyone’s symptoms match: Cursor Tab drops 20 to 60 seconds after a restart or relogin with [not_found] on the cpp endpoint, while Chat and Agent still work. Indexing gets stuck for the same reason too, shared auth state.
I moved the scope from an individual subscription resync to a broader issue. The team is looking into it. No ETA yet.
To help us debug faster, if any new reporter can share:
The Request ID from a failing Cursor Tab request right after it breaks right corner of chat > Copy Request ID, or from logs Output > Cursor Tab line requestId ...
The exact time with timezone for one of the [not_found] events
Your Cursor version and OS if it’s not already in your post
This lets us match requests on the backend across multiple accounts and confirm it’s one shared root cause, not a few different issues with the same symptom.
I’m having the same issue with the Cursor Tab feature. I have to log out and log back in for the Cursor Tab to work properly. But even then, the error reappears after a while. My internet connection is fine and the chat and agent features are working normally.
This was a bug in 3.4.x that caused Tab to crash with [not_found] 20 to 60 seconds after a restart or re-login, and indexing would get stuck on “loading…”. The symptoms in this thread all match, same root cause.
The fix is already in the stable release starting from 3.4.17. The current stable is 3.4.20. Please update to the latest version, Tab and indexing should work again.
If you still see [not_found] after updating to 3.4.17 or later, reply here with the Request ID from the failing request and the exact time with your timezone, and we’ll take a look.
