Cursor agent chat consistently hangs on "Planning next moves" "Taking longer than expected," - all agents now fully unresponsive. Console shows repeated "No first token received" warnings

Where does the bug appear (feature/product)?

Cursor IDE

Describe the Bug

Issue: Cursor agent/chat consistently hangs on “Planning next moves” / “Taking longer than expected,” now fully unresponsive on all chats. UI part ok I can access settings etc but not update or about cursor for example. Console shows repeated “No first token received” warnings.

Environment: [Mac OS 26 + Newest version]

Troubleshooting already completed:

Force-quit and killed lingering background processes (pkill -f “Cursor”)
Cleared Cursor’s Cache, CachedData, and Network Persistent State folders
Checked for interfering proxies/VPNs and confirmed firewall isn’t blocking Cursor
Deleted the codebase indexing database (chunks.db) for the affected workspace
Reinstalled Cursor fresh
Disabled HTTP/2 via cursor.general.disableHttp2 in settings.json (UI toggle wasn’t saving) — diagnostics now pass fully, issue persists
Ran Settings > Network > Run Diagnostics: all checks pass
Checked Activity Monitor during the hang: CPU usage low/idle (~3.6%), ruling out a stuck loop or heavy indexing process
Opened Developer Tools during a hang: console shows repeated “No first token received” warnings plus unrelated “Trusted type expected, but string received” errors
Confirmed via status.cursor.com that no outage is currently reported
App becomes unresponsive enough that closing the project/folder requires a force-quit
Cant even open new project.
Software is unusable right now.

Steps to Reproduce

Open up any chat, Type a request.

Expected Behavior

I should get a response

Operating System

MacOS

Version Information

Using 3.15.6

For AI issues: which model did you use?

All

For AI issues: add Request ID with privacy disabled

Request ID: 019fce19-1fd7-7c08-98d9-5f387d173056

Does this stop you from using Cursor

Yes - Cursor is unusable

I’m also experiencing this issue too though on macos 27 beta 5

Hey, thanks for the detailed report. You’ve done some solid troubleshooting.

Up front, this isn’t a network issue. Since Network > Run Diagnostics passes and HTTP/2 is already disabled, there’s no need to dig further into proxy or VPN. Based on the symptoms, this looks like a client-side hang inside the app itself, not a connection problem. This shouldn’t happen, and it’s not something in your setup.

A few things to try in order:

  1. Roll back to 3.14.x. You said it used to work, and 3.15.6 lines up with when it broke. Rolling back is the fastest way to unblock yourself right now. You can download an older build from downloads/changelog.
  2. Instead of pkill -f "Cursor", fully quit with Cmd+Q, then restart. pkill can leave orphaned helper processes, and the next launch can start in a weird state. Ideally also reboot your Mac.
  3. If chat hangs again, run Cmd+Shift+P > Developer: Restart Extension Host, then retry the request.

To narrow this down, can you help with two things:

  • If you leave the frozen chat for about 2 minutes, do you see any error dialog, like “Agent Execution Timed Out” or “Reload Window”?
  • Please attach main.log from ~/Library/Application Support/Cursor/logs/<session>/main.log from a session where the hang happens.

This looks similar to an issue we’re tracking on newer macOS versions. Let me know if rolling back to 3.14.x helps, and send the logs so we can dig in.

@wingatej, you’re on macOS 27 beta 5, so it might be a slightly different case. Please reply with whether it still reproduces after rolling back to the previous version, and share the Request ID from your side.

I’m facing the same issue over here

Hi
Thanks for getting back, below are some answers:
1 - Rolled back to 3.14 and made, no difference. If i change workspace I cant even get it to load up it spins “taking a little while…” sat for longer than 10mins. No other notice.
2 - Did a full system restart.
3 - Couldn’t find this typed in “developer: restart and only option i have is " New agent:…”
4 - Second bullets: As mentioned above, just spins with same message or in a chat it will say taking longer than expected… nothing more.
5 - Just to check, with information sensitivity on a public form should i be sending them in a private manner in some way to you? If so how to do on the forum screen safely?

Thanks for the detailed write-up and for trying all the steps.

First, on your question about logs, important: don’t post them in a public thread if you’re worried about sensitive info. main.log usually doesn’t include source code, but it can include file paths and workspace names. It’s safer to send it privately. Let me know and I’ll DM you so you can share it there.

Now to the main point. The fact that rolling back to 3.14.x and doing a full Mac restart didn’t help is a useful signal, and it changes the picture. That means it’s not just the version. Since even loading the workspace hangs taking a little while for 10+ minutes, it looks like it’s not the chat itself, but something one layer below getting stuck, and we’ve seen this on macOS 26.

A few things that will help narrow it down:

  • Try launching Cursor on an empty folder: File > New Window, open a clean empty directory, then send a simple request in chat. If it hangs there too, it’s not specific to your workspace.
  • About Developer: Restart Extension Host: if it’s missing from the Command Palette and you only see New agent, that alone suggests the window or extension host is already stuck. Instead, type Reload Window in Cmd+Shift+P and see if the Command Palette comes back to life after that.
  • Next time you hit the freeze, share the main.log from ~/Library/Application Support/Cursor/logs/<session>/main.log for that session privately, as above.

This looks like a class of issues we’re tracking on newer macOS versions. Once you’ve grabbed main.log and checked item 1, let me know and we’ll dig in further.

@wingatej, @Fardeen_Zubair: you’re on different macOS versions, so you might have a slightly different case. Please reply with your macOS version and your Request ID with Privacy Mode turned off, and we’ll handle yours separately.

У меня с 12:00 такая проблема “Reconnecting” только на Windows

Shortly after posting the last reply, I restarted the mac. Which seem to have fixed the issue as when i reopened cursor macos was asking for local network permission, even though it had been left open overnight and no major updates had been done.

Strangely I have updated mac to the latest OS26 and it seems to have begun working again. still older cursor. Weirdly cursor was working initially and no updates to computer. It then randomly stopped. Im now not really clear what it was causing it to be honest.

Great news that it’s working for both of you.

@kursor_user, @wingatej: What helped you full Mac restart or OS update, and then macOS asked again for Local Network permission on the next startup fits the pattern. The issue was client-side and tied to how newer macOS versions behave, not your network or settings. So proxy or VPN really wasn’t related.

If the freeze comes back, don’t rush to pkill. First do a full Cmd+Q, and if it happens again, grab main.log from ~/Library/Application Support/Cursor/logs/<session>/main.log for that session. It’s better to share it privately since it may include paths and workspace names. Tell me and I’ll DM you. With the log we can dig deeper if it shows up again.

@Fardeen_Zubair: To confirm if it’s the same case, please share your macOS version and the Request ID with Privacy Mode turned off.

@Aleksandr_Ryumshin: “Reconnecting” on Windows is most likely a separate issue and not related to this macOS case. Please start a new thread and include the Request ID with Privacy Mode turned off and your Cursor version so we can look at it properly without mixing it into this thread.