Cloud agent stuck at "Running start script": noVNC keeps running in the foreground

Support Bug Reports
, , ,
1

Where does the bug appear (feature/product)?

Background Agent (GitHub, Slack, Web, Linear)

Describe the Bug

I used a cloud agent last week for the first time where I just tested, if it could push to my repository, which worked. Today I wanted to use it, but the cloud agent is stuck on “Running start script” at:

Sat Jun 13 09:27:16 AM UTC 2026] AnyOS desktop ready. Connect via noVNC on port 26058.
WebSocket server settings:
  - Listen on :26058
  - Web server. Web root: /usr/local/novnc/noVNC-1.2.0
  - No SSL/TLS support (no cert file)
  - proxying from :26058 to localhost:5901


Navigate to this URL:
    http://cursor:26058/vnc.html?host=cursor&port=26058

Press Ctrl-C to exit

I used my local cursor instance and copied the logs into it and it said:

Your logs show desktop init actually finished (AnyOS desktop initialization complete), but then noVNC keeps running in the foreground (Press Ctrl-C to exit). Cursor appears to wait for that script to exit, so the UI stays on “Running desktop init script”.

If it still hangs after that, copy the Request ID from the stuck chat (chat menu → Copy Request ID) and report it to Cursor — the noVNC foreground block looks like a platform bug on their side.

So I am reporting this here and hope to get help.

Steps to Reproduce

I just ran a cloud agent.

Expected Behavior

The cloud agent actually runs.

Operating System

Linux

Version Information

Version: 3.7.19
VS Code Extension API: 1.105.1
Commit: 80c653c2c3528e65016a0d304b54486084b470b0
Date: 2026-06-07T01:42:56.126Z
Layout: glass
Build Type: Stable
Release Track: Default
Electron: 39.8.10
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
xterm.js: 6.1.0-beta.256
OS: Linux x64 7.0.9-arch2-1

For AI issues: which model did you use?

Composer 2.5

For AI issues: add Request ID with privacy disabled

Request ID: bc-7462b6a3-8c42-44e6-a1d3-dee175cc5373

Does this stop you from using Cursor

No - Cursor works, but with this issue

5

It did its work now, over an hour after starting it. I tried it again and it is again stuck on a new cloud agent task, so I guess it waits for a timeout until it does its work? Is this a bug or a configuration issue on my side?

6

Hey, thanks for the detailed report with logs and the Request ID, that really helped.

This isn’t an issue with your repo or setup. The pod does get created, but on our side the agent’s internal service gets stuck starting, which is why you see “Running start script” forever. We can see this in your specific agent logs, and it’s not just you. It looks like a server-side regression that started today.

We’re already looking into it. I can’t give an exact ETA for a fix yet. You can try canceling and starting the agent again. There’s a chance you’ll land on a recovered pod, but if it gets stuck again, it’s the same server-side issue, not anything on your end.

I’ll post an update in the thread as soon as I have one.

7

I’ve had the same issue multiple times and am having it even now

9

We are seeing the same issue on a Cursor Teams Premium account.

Cloud Agent repeatedly hangs at:

AnyOS desktop ready. Connect via noVNC on port 26058.
Press Ctrl-C to exit

We have:

  • Active Teams Premium subscription

  • Paid invoices

  • 54.7% included usage consumed (45.3% remaining)

Additional errors in our logs:

Failed to create recording directory /opt/cursor/artifacts/
Failed to create recording directory /opt/cursor/recording-staging/

The environment remains on “Running start script” for over an hour.

Has Cursor identified the root cause yet, or is there a workaround?

10

Seeing the same issue. Please update the cursor status page since cloud agents are down.

11

I am having this exacxt same issue since about 7 hours ago

12

I’m having the same issue for 7+ hours now. Please fix it and update status page

13

Hey, thanks everyone for the confirmations. Just to make it clear for anyone joining the thread: this is a known issue on our side, not in your repos or settings. The pod gets created, but the agent internal service gets stuck during startup, which is why you see the endless “Running start script”. It started today, looks like a server-side regression, and it’s affecting more than one user.

We’re already looking into it. I can’t share an exact ETA for a fix yet. As a workaround, cancel + retry sometimes lands on a recovered pod and the request completes like for @makneidinger, it started working again after about an hour. If it gets stuck again, it’s the same server-side issue. You don’t need to change anything on your side.

I’ll post an update here as soon as we have one.

14

Any reason this isn’t posted to status.cursor.com? That would help to know when a good time to resume would be. Thanks!

15

Hey, quick update on this thread. This is still a server-side regression that started on June 13. The pod gets created, but the agent’s internal service hangs during startup, so you see an infinite “Running start script”. This isn’t related to your repos or settings, and it affects more than one user.

@TeamsAsia: the Failed to create recording directory errors are coming from the same issue (pod startup instability). You don’t have a separate problem.

Same workaround as before: cancel + retry sometimes lands on a healthy pod and the request goes through (for one reporter it started working after about an hour). If it gets stuck again, it’s the same server-side issue. You don’t need to change anything on your side.

Quick question: as of June 16, does this still reproduce for you, or have your latest runs started working? That’ll help us understand the current status. I’ll post another update here as soon as I have one.

16

Thank you for the update @deanrie.

I just tried it and it worked for me. So either I got lucky or it is fixed.

17

Good news. If the run completed normally without getting stuck for a long time on “Running start script”, it most likely means things have stabilized on our side, and it was not just a lucky pod.

I’m not marking the issue as fully closed yet. I want to make sure it’s stable and not a one-off recovery. If any of the next runs gets stuck on “Running start script” again, reply here and include a new Request ID from the stuck agent. It’ll be the same server-side issue, so you don’t need to change anything on your side.

Thanks for the update. For everyone else in the thread, if your cloud agents are starting normally right now, let us know. It’ll help us confirm the regression is gone.

cc @TomO @Olo @Karim_Mena @Dylanmfe @neals

19

Thread update: this was a server-side issue that caused cloud agents to get stuck on “Running start script”. We’ve fixed it on our side, so everything should work normally now and you don’t need to change anything.

If you hit the same stuck state again on a fresh run, reply in the thread with a new Request ID and we’ll take another look.

cc @Dylanmfe @TeamsAsia @TomO @Karim_Mena @Olo @makneidinger