The problem: My app's editing branch has drifted out of sync with the published version and I can't recover it. Every editor session — even a brand-new browser tab — loads the same state with the "A new version of this app has been published — please integrate changes" banner at the bottom. When I click Integrate changes, it hands off to the AI builder and the merge just fails.
The AI builder reports that the git merge is failing at the infrastructure level — no conflict output is produced, it fails before generating anything to resolve, and it has no reset, fetch, or checkout available to work around it. So it's not a conflict I can hand-resolve.
What I've already tried / ruled out:
• Clicking "Integrate changes" → fails every time
• Opening fresh editor tabs → still loads the same broken branch
• Version History → only shows diffs, no "restore this version" button
• App "…" menu → Share/Rename/Duplicate/Download ZIP, but no reset or restore option
• Since it's Retool-managed git, there's nothing to clone/revert locally
What I need: a way to hard-reset the editing branch back to the currently published version. I don't need anything on the branch — totally fine discarding all unpublished changes. The live published app is working fine and should stay untouched; it's only the editing sandbox that's broken.
Questions for the community/staff:
1. Is there any self-service way to reset or discard a broken editing branch on a Retool-managed-git app that I'm missing?
2. If not, is this something only Retool support can do server-side?
Any pointers appreciated — this is time-sensitive. Thanks!
On a Cloud instance using the new app builder with Retool-managed source control, an editing branch is stuck out of sync with the published version — the AI-assisted 'Integrate changes' / pull from main fails, and 'Reset branch' doesn't restore a clean, in-sync state.
AI Response
This pattern has two documented causes. Recent 'Integrate changes' / git-pull-main failures in the new app builder were traced by Retool staff to Vite port collisions in the sandbox, which was fixed in 4.17.0 and completed its Cloud rollout around July 2, 2026 — so first confirm the branch is being edited in a freshly initialized in-editor thread and retry the integrate/pull, since older threads can stay pinned to a stale origin/main. Separately, note that 'Reset branch' does not apply merge commits into the editor; it resets the local app to a baseline snapshot, which is why previously merged changes can surface as removals and make the branch look permanently out of sync. To recover a clean state, avoid committing those 'removal' diffs and instead recreate the feature branch fresh from main, then re-apply the outstanding work — Retool-managed source control is designed for short-lived, mostly linear branches, so long-running or externally-merged branches on the same app are a known limitation.
Sources
Source control sync broken for one app (can't pull main / AI build sandbox seeds empty)
A new-app-builder Cloud report of the same 'Integrate changes'/pull-main failure that Retool staff attributed to a Vite port-collision bug fixed in 4.17.0 (rolled out to Cloud by early July 2026). Reset branch to merge changes from main
A solved topic where Retool Support explains why 'Reset branch' leaves a branch looking out of sync and recommends recreating the branch from main rather than committing the phantom removal diffs.
The Community Team is testing out a new automation. Let us know if it's helpful (or not) by leaving a , , or . Or by marking this post as the "Solution"! Let us know if you have any feedback here.
Thanks for the detailed writeup, it really helped. Our engineering team is actively looking into this now. I know it's time-sensitive, so I'll follow up here very soon with next steps. Hang tight!
My published version was clean at the time so I had it start fresh from that last publish and delete anything built from that point forward and rebuilt the modules/progress
, 
, or 
. Or by marking this post as the "Solution"! Let us know if you have any feedback 
Our engineering team is actively looking into this now. I know it's time-sensitive, so I'll follow up here very soon with next steps. Hang tight!