Agent terminal deadlock - stuck on "Running" for all commands

Subject: Agent terminal deadlock - stuck on “Running” for all commands

Problem: The Agent terminal tool is completely deadlocked. Every command initiated by the Agent (even a simple dir) stays in the “Running” state indefinitely. No output is ever displayed in the chat interface.

Key Evidence:

  1. Manual execution works: I can manually type and run the commands in the same terminal window, and they work perfectly.

  2. Infrastructure is fine: My SSH tunnels (ports 5432 and 3000) are active and the remote server is receiving requests.

  3. Agent failure: The issue is specifically with the Agent’s ability to trigger and read from the terminal. It is not a network or code issue, but a UI/Agent bridge failure.

Steps to reproduce:

  • Open any terminal task.

  • Let the Agent try to run any command.

  • Result: Perpetual “Running” spinner with zero output.

Environment: Windows, SSH Tunneling to Ubuntu remote server.

I seem to have narrowed the issue and have a temporary fix.
Go to Antigravity Agent terminal commands fail to execute due to missing "Enter" character for the report.

image

1 Like

My agent is completely ignoring this, commands keep freezing.

I tried running the command manually and simultaneously asking the agent to run it. It worked, but the agent added some extraneous “c” characters when I entered the command, which broke the command. In short, Antigravity simply works terribly with the terminal, and the worst part is that no one is addressing this issue or responding to my posts.

Thank you for the report. This has been escalated to our engineering team for investigation. To provide our developers with the necessary diagnostic data, we encourage you to use the in-app feedback tool (found via the Feedback icon in the top-right corner > Report Issue). This is the most efficient way to ensure a prompt resolution.

1 Like

I sent a bug report as you said through feedback, but since Antigravity doesn’t notify the user in any way, I didn’t even understand whether my report was actually sent or not.

Thanks for this .. happening to me too today since the 1.16.5 upgrade … inconsistently …

I’ve instructed my agent to add the character to commands for now.

I’m still having this issue … it does not appear to be limited to \n newlines … started with 1.16.5.

In some cases I’m seeing commands that fail to actually submit to system, in other cases I’m seeing commands are successfully executed but the tool handler fails to return state.

Its against company policy to submit diagnostic data that may contain controlled data so I can’t use the feedback form.

Bro, I don’t think anyone cares what they write here. My microwave gets updates more often than Google Antigravity.

2 Likes

I’m having the same issue, I found that downgrading to the last 1.15 version fixed it for me temporarily. If you are running an apt based Linux distro run sudo apt-get install antigravity=1.15.8-1769233008

2 Likes

@Andrew_Pyka I thought about downgrading, but I noticed that the agent appears to “wait” on backend computing resources in 1.16 rather than error/stop …. so there’s not a clear win.

Hoping for a clean update soon, though I will probably stop riding bleeding edge now that I can’t trust the release engineering. A canary/stable branch philosophy is needed.

….. real changelogs would be nice too, the web blog is garbage.

ps … I suspect its a Linux only issue … as I don’t see masses of Mac/Windows people complaining.