[Bug] Antigravity v2.28.2: Continuous crash loop and reconnection error in Standalone App and VS Code

I am facing a critical and recurring crash loop error in the Antigravity ecosystem (version v2.28.2). The issue occurs simultaneously in both the Standalone Application / App and the IDE / VS Code Extension on Windows 11.

Symptoms:

  1. In the Standalone Application / App: The program closes and restarts on its own continuously right upon opening or while sending messages in chat, showing cancellation errors such as Canceled project structure overview and backend process failure.

  2. In the IDE / VS Code Extension: The interface enters an infinite loop and gets stuck on the loading screen: “Reconnecting to Antigravity (attempt 1 of 3)…”.

  3. Logs / Communication Errors: In the interface and console, local connection errors net::ERR_FAILED occur, along with forced process termination with code process_exit_nonzero in the shared daemon/Language Server.

Steps already taken without success:

  • Full clean reinstall from the official installer (v2.28.2 x64).

  • Deletion of cache and global state folders in %APPDATA%, %LOCALAPPDATA% (Antigravity folder), and %USERPROFILE%\.antigravity.

  • Force-killing all hanging processes (taskkill /F /IM node.exe /IM code.exe /IM antigravity*).

Since the crash persists across both versions (App and IDE) even after a completely fresh install, it is clear that the issue lies within the internal communication daemon (IPC/Language Server) or in the Antigravity server itself in this build.

System Specifications:

  • OS: Windows 11 (64-bit)

  • App/IDE Version: Antigravity v2.28.2 (x64)

Can anyone from the dev team or the community check if this is a known bug in version v2.28.2 and how to fix this continuous daemon crash?

This restart loop on Windows 11 in version v2.28.2 (both in the standalone app and the VS Code extension) usually happens when the local language_server_windows_x64.exe process tries to restore a corrupted local SQLite database state or an orphaned gRPC socket inside %APPDATA%\Antigravity.

To break the crash loop without having to reinstall everything from scratch, close VS Code and the standalone app, terminate any lingering language_server processes in Task Manager, and clear only the local state cache in PowerShell:

Stop-Process -Name "*antigravity*","*language_server*" -Force -ErrorAction SilentlyContinue
Remove-Item -Path "$env:APPDATA\Antigravity\User\workspaceStorage\*" -Recurse -Force
Remove-Item -Path "$env:USERPROFILE\.gemini\antigravity\daemon\*.lock" -Force -ErrorAction SilentlyContinue

When reopened, the language server rebuilds a clean index and stops crashing during reconnection.

Hi @Ryan_Nascimento,

Welcome to the Forum, and thank you for listing the cleanup steps you already ran.

Both the Antigravity App and VS Code extension (VS Code Extension) use the local agy.exe backend and store global conversation state under %USERPROFILE%\.gemini\antigravity\ (rather than %USERPROFILE%\.antigravity or %APPDATA%). If a stale entry in that summary index fails during startup recovery, agy.exe exits with process_exit_nonzero and triggers the "Reconnecting to Antigravity (attempt 1 of 3)..." loop.

Please try these steps:

  1. Update to v2.29.0: Update from v2.28.2 to Antigravity App v2.29.0 (Changelog), which fixes a background conversation database watcher loop, and update the VS Code extension.
  2. Rebuild the summary index: Close both apps, terminate agy.exe (taskkill /F /IM agy.exe), and rename %USERPROFILE%\.gemini\antigravity\conversation_summaries.db and agyhub_summaries_proto.pb (e.g., to .bak) so the backend rebuilds a clean index on launch.

If the issue persists, please reply here with the latest trace from %USERPROFILE%\.gemini\antigravity\crashes\ and the last 20–30 lines of %USERPROFILE%\.gemini\antigravity\cli.log.