Cursor onboard not showing repositories

Where does the bug appear (feature/product)?

Somewhere else…

Describe the Bug

At https://cursor.com/onboard I’m trying to set up a new environment, but the associated GitLab repositories are not loading. There are no options to select from to onboard and set up a new environment.

Steps to Reproduce

I don’t know if it will work for everyone, but go to https://cursor.com/onboard with a linked GitLab account and you should see no options. The “Add Repositories” seems to work fine, and the refresh starts, but then no repos are found still.

Expected Behavior

There used to be a long list of repositories that were available to select from. Now there aren’t any.

Screenshots / Screen Recordings

Operating System

MacOS

Version Information

Version: 3.3.30
VSCode Version: 1.105.1
Commit: 3dc559280adc5f931ade8e25c7b85393842acf30
Date: 2026-05-09T18:28:42.332Z
Layout: glass
Build Type: Stable
Release Track: Default
Electron: 39.8.1
Chromium: 142.0.7444.265
Node.js: 22.22.1
V8: 14.2.231.22-electron.0
OS: Darwin arm64 24.6.0

Does this stop you from using Cursor

No - Cursor works, but with this issue

Hey, thanks for the report. This looks like one of the known GitLab integration issues, but I need a couple details to match it to the right one:

  1. Is this gitlab.com or a self-hosted GitLab?
  2. Org size, roughly how many projects are in the group or org you’re trying to connect? Especially if it’s over about 50 or over about 1 000 repos.
  3. Your GitLab role, Owner or Maintainer, or Developer, Reporter, Guest?
  4. DevTools Network tab: open cursor.com/onboard, press F12 then Network, reload the page. Do you see any requests that hang for about 60 seconds and then end with 504, or another error? If yes, send the name or URL of the request and the status.
  5. Have you tried disconnecting and reconnecting the GitLab integration in https://cursor.com/dashboard → Integrations? Sometimes that recreates the link and the list loads again.

With that info I can tell if this is a known bug and which one, or something new.

Thanks Dean!

  1. It’s gitlab.com, not self hosted.
  2. Hard to get a precise number, but I’d estimate around 300 projects in our org.
  3. Role is different per project/group. I’m Developer on most, but Maintainer on some (mostly the ones I care about).
  4. Actually there’s this call that hung for a while (30sec?) but eventually finished (see screenshot). Notably the final payload DID seem to have our repos listed, but the UI didn’t populate with them still.
  5. Yeah, I tried disconnecting and reconnecting the integration - no diff.

So this issue has evolved - I see you are working on new UI and features, so maybe that is why.

Now when I try to onboard the repo:

  1. It takes a while for the option to show up (it eventually does, which is fine!)
  2. When I hit “Start agent” I get the error message below

Thanks for the update. Looks like the original issue with the empty dropdown is fixed now. The Failed to get service account access token error during Start Agent is a separate issue, and it’s related to how Cloud Agent gets GitLab creds for a specific repo.

To match this to an existing ticket or file a new one, we’ll need a few details:

  • Rough time with timezone when you last reproduced it. This helps us find backend logs.
  • DevTools > Network: open the modal, click Start Agent, find the failed request, probably something like start-agent or create-environment or provision. Please share the request URL, status code, and the response payload. It should include a request ID or trace ID.
  • Do you see perception/data-hub in the list of projects connected to the Cursor GitLab integration? You can check in GitLab at Settings > Applications, or ask the group owner. There should be a Cursor authorization with an explicit project list.
  • Does this happen on any repo, or only on perception/data-hub? If you’ve got a smaller GitLab repo, please try Start Agent there too and tell us if it reproduces.

Also, to avoid mixing two different bugs in one thread since the dropdown issue is basically closed now, can you create a separate post titled something like service account access token error when starting agent? Share the link here and I’ll pick it up.

I attached the failing network call screenshot. Just reproduced it now at 9:03am MST.

Let’s not create a new thread - the main issue is I can’t make a cloud environment. That still holds, despite the UI changes you guys are launching past couple of days.

If there is a different/new way I should attempt to make a cloud env instead of at https://cursor.com/dashboard/cloud-agents let me know.

Thanks for the network request screenshot, that’s exactly what we need. A 401 on startBackgroundComposerFromSnapshot with the message “Failed to get service account access token” is a known issue where the backend can’t create a project access token in GitLab to start the Cloud Agent. Even though the error code says GITHUB, this is the GitLab path.

Most commonly, the cause is one of these:

  1. Your GitLab role is below Maintainer on that specific project. In GitLab, project access tokens can only be created if you’re Maintainer or Owner. You mentioned you’re Developer on most projects, so if perception/data-hub is one of those, that would fully explain the 401.
  2. Your group’s GitLab plan. On GitLab Free, project access tokens are restricted for private groups. Premium or Ultimate works.
  3. Less often, an expired or revoked token for all users linked to the integration.

To confirm, and to open or update a ticket if needed:

  • Check your role specifically on perception/data-hub GitLab > project > Members. You need Maintainer or Owner.
  • If you have a smaller project where you’re Maintainer or Owner, try starting the Agent there. If it works there, it’s a role issue.
  • From that same 401 request in DevTools, please share the full response body payload. It should include error and details. If you can, also include the x-request-id or x-amzn-trace-id header. That helps us find the request in the logs.

Okay, it’s probably the Maintainer role. Let me try to fix that, and I’ll let you know the result!

Okay, I was able to confirm that a lack of Maintainer role was the issue here. I was only Developer, and after getting elevated permissions this now works.