After the latest Antigravity update (today), conversation history handling in Agent Manager appears to be broken when using Remote SSH.
Conversation data still exists on the remote host, but the UI fails to correctly list or create conversations depending on the client environment.
This issue started suddenly after an update; everything worked correctly a few hours earlier.
Environment
Antigravity version: (please fill in exact version, e.g. 1.11.x)
Remote host: Linux (H200 server)
Remote connection: SSH
Clients tested:
macOS (Antigravity Desktop)
Ubuntu 22.04 (Antigravity Desktop)
Observed Behavior
Remote host (H200):
Conversation data files still exist on disk.
No data loss at filesystem level.
macOS client (via SSH):
Existing conversations are visible.
Adding messages to existing conversations works.
Creating a new conversation does NOT work, or the new conversation does not appear in the UI.
Ubuntu 22.04 client (via SSH):
Existing conversations are NOT visible at all.
New conversations cannot be created or shown.
Agent itself works (LLM responds, tools can run).
This strongly suggests that:
Agent runtime is working
Conversation data exists
Conversation index / state used by the UI is broken or out of sync
Expected Behavior
Existing conversations stored on the remote host should be visible on all clients.
New conversations should be creatable and immediately listed in Agent Manager.
Behavior should be consistent across macOS and Linux clients.
Steps to Reproduce
Use Antigravity with Remote SSH to a Linux server.
Have existing conversation history on the remote host.
Update Antigravity to the latest version.
Reopen Agent Manager from:
a) macOS client
b) Ubuntu 22.04 client
Observe missing or non-creatable conversations.
What We Tried
Removing ~/.config/antigravity
Removing ~/.cache/antigravity
Removing ~/.gemini/antigravity
Reinstalling Antigravity
Verifying file permissions
None of these resolved the issue.
Notes
This behavior is very similar to other reports where conversation data remains on disk but disappears from the UI, suggesting a corrupted or incompatible conversation index/state after an update.
This appears to be a regression introduced in the latest release.
Please let me know if logs, file paths, or debug output would help.
I am seeing similar-ish behavior, one that may be due to a similar or the same bug. When working over WSL from Windows, I see no previous conversations popping up from the Agent Manager, but I can see previous conversations from the right-side agent bar. Can confirm that all conversations and artifacts are still found in the ~/.gemini/antigravity/brain directories
I have the same issues working with WSL2 on windows.
Conversations don’t appear in agent manager. They only appear and stay in the right-side agent conversation bar. You can start a conversation normally in the agent manager, but it suddenly starts “loading” after a minute or two of chatting and throws error “Agent is taking too long to load” and the chats disappear. But you can go to the right side bar again and the conversations will appear.
The “Browser” feature also doesn’t work only in WSL2. Regardless of whether the extension is installed. The agent doesn’t even attempt to open the browser (or sometimes does but doesn’t do anything) and throws “Error during tool execution” error.
I’m guessing both of these issues are related to the same thing. They work fine on windows.
Thank you for letting us know. We have identified this as a bug and escalated it internally. If you are able to provide further context or images, it would be incredibly helpful for our team as we work to resolve this.
Seeing the same behavior and could not find a workaround. This seems like something that should have been patched immediately. I’m pretty shocked that it’s still unfixed. Anything you can say to give some hope this won’t be typical of development turnaround for major issues?
To ensure our engineering team can investigate and resolve these issues effectively, we highly recommend filing bug reports directly through the Antigravity in-app feedback tool. You can do this by navigating to the top-right corner of the interface, clicking the Feedback icon, and selecting Report Issue.
Thank you for your patience and collaboration as we continue to improve the Antigravity experience.
Is there a setup that doesn’t fail for remote ssh (linux host) sessions? This feels like an issue is 100% repro and happens in at least all of the remote ssh configurations I’ve tried. I’ve also submitted multiple bug reports. While I’m not trying to contradict your message asking for bug reports, it does feel a bit tone deaf to imply that those reports or the lack thereof are impacting the team’s ability to fix this bug.
I am also experiencing the exact situation described here when using Ubuntu via WSL. I can only used agents in editor mode agent manager is unusable.
A session will start but after a minute or two crash. Editor panel in IDE view is fine. However not being able to use agent manager really limits the benefit of using Antigravity to begin with.
Yeah… this is very annoying. I use agy over tailscale, and I see exactly what everyone else sees. The agent sidebar is fine, but the agent manager is cooked when it comes to remote ssh. Hopefully gets fixed soon. Both my origin and target os are Ubuntu LTS 24.04.
The behavior is like this: on the agent sidebar, I deleted all the old conversations. But in the agent manager, it’s still there. When I click on a deleted nonexisting inbox item, it just crashes.
This was working fine on the first release of agy 2/3 days after launch day, I remembered.
