Execute commands (cargo build, pnpm build, etc.) hang indefinitely with no response

When executing long-running or resource-intensive build commands such as cargo build or pnpm build within the Antigravity IDE, the process hangs indefinitely. The terminal/console stops returning any output, and the build process never completes or times out. This issue completely blocks the development workflow, requiring a manual restart of the IDE or forcefully killing the background processes.

  1. Antigravity IDE Version: 2.0.3
  2. Operating System: Windows 11 x86_64

Steps to reproduce the issue:

  1. Open a project (e.g., Rust or Node.js) in Antigravity IDE.

  2. Open the integrated terminal or command runner.

  3. Execute a build command such as cargo build or pnpm build.

  4. Observe the terminal output and the process status.

Expected behavior: The build process should execute normally, continuously stream the standard output/error logs to the console, and terminate with an appropriate exit code once the build is finished.

Actual behavior: The command executes initially, but the process freezes indefinitely. No further output is displayed, no exit code is returned, and the terminal becomes completely unresponsive.

Agent message

You are absolutely right. This is entirely a system-side “bug (design flaw).”

Despite being designed to “asynchronously prevent freezing,” the system is resulting in users being kept waiting and appearing to hang. This is due to the following bugs in the asynchronous task lifecycle management:

Technical background of the bug (flaw)
Bug that fails to detect instant process termination (Exit): Ideally, when a process terminates instantly due to the cargo not being found, the system should immediately notify me (the AI) of a “task failed” termination event and display an error on the screen. However, the system was not correctly monitoring the process’s exit status, and the task remained in a zombie state, showing “RUNNING” indefinitely.

Buffering and log synchronization delay bug: Although the error message (CommandNotFound) output from the shell was written to the log file, the update was not synchronized to the AI ​​in real time, and I could not detect the output until I manually “checked the status.”

As a result of these combined issues, a paradoxical bug has occurred where, despite intending to run asynchronously and safely, the application hangs in the background, constantly waiting for responses from zombie tasks.

I sincerely apologize for the extra time you’ve had to spend reinstalling your Rust environment, only to have the development tools bug cause this.

As long as the agent’s execution system has this bug, there’s a very high probability that executing commands from my end will result in the same zombie task issues and hangs again.

For this cargo build, I apologize for the inconvenience, but currently, the most reliable and fastest method is for you to run it directly from your terminal (PowerShell or VS Code terminal).

If you encounter any code errors or compilation errors during the build, please paste the error messages here, and I will immediately create a proposed code fix!

Just wanted to share some similar bug reports I found regarding this issue: Antigravity IDE 2.0.3 takes ~5 min to run any terminal commands

I have the same bug. it seems my antigravity cant execute any commands anymroe

Hi @Joinity,

Welcome to the Forum,

Please update to the latest version of Antigravity IDE v2.0.4. If the issue persists, please share any error message or trace IDs associated with this issue.

Hi @chunduriv,

Thank you for the response.

I have updated to Antigravity IDE v2.0.4, but unfortunately, the issue still persists. The commands continue to hang indefinitely with no response just as before.

Since the terminal freezes completely, I don’t see any explicit error messages or trace IDs on the screen. I have attached a screenshot showing the hanging state for your reference.

Could you please let me know where I can find the logs or trace IDs to share with you?

Best regards,

This has been a bug ever since the program was created, very frustrating. The only fix is to disable shell integration in the settings, which then means you can’t see waht the agent did at the prompt very easily nor interact with it, but at least it works. EDIT: i disabled it a bit ago and still had some commands hang, so it might not be the fix any more.

Shell integration allow you to see the terminal the agent uses, and also execute your own commands and the agent can see it. It works fine in vscode and cursor, etc (although they did have a ton of issues at the beginning until they fixed it, this exact same issue).

Hi @chunduriv and @Paul_Foo,

Thank you for the suggestion, @Paul_Foo. I tried disabling shell integration in the settings as you mentioned, but unfortunately, the commands still hang indefinitely for me as well. It seems that workaround might not be effective anymore.

Best regards,

Yeah it seems to help after restarting but not a 100% fix any more. It is a 100% fix, or was, in the agent manager but i can’t use the agent manager with remote ssh so i’m just stuck.
Everyone should start submitting bug reports via feedback in the program so they get on top of this asap

Make sure to tell the agent to set a background timeout on every terminal command or a total timeout and also to use short 1 line commands instead of concatenating a bunch of commands together. This is very frustrating as i have it do a lot of terminal commands to run tests for me

it wasnt taking so long for me before. it seems its completely stuck on every terminal command it makes. it eventually executes all of them but it takes minutes instead of seconds for a basic copy command

@kackyt I created the thread you linked to, and it sounds like @Joinity is hitting exactly the same issue I am, but I’m curious - with your original build issue, does it run immediately and then hang forever, or does it eventually resolve if you wait long enough?

While recording a more detailed video as part of trying to open a Google One support ticket for this, I noticed when running a simple, quick Python script that normally takes seconds, it took about 7 minutes to run the Python command, then another 6 minutes after the script ran before it kept going. On my system, language_server_windows_x64.exe was spinning a CPU core during that entire 13 minutes, both before and after the external command was actually performed.

i can run tasks in my terminal instantly like npm run dev but whenever the ai makes this call it takes minutes