Agent terminated error - is Antigravity stable on Google AI Ultra for Business accounts?

Google Antigravity
1

Hello,

I am using Google Antigravity on a Google AI Ultra for Business (Workspace) account and I frequently encounter the following error:

“Agent terminated due to an error”

This happens across different tasks - sometimes mid-way through a complex agentic run, sometimes early on. It forces me to restart and lose context. The issue feels more frequent than it should be for a paid enterprise subscription.

My questions for the team:

  1. Is Google Antigravity considered fully stable and production-ready for AI Ultra for Business accounts, or is it still in an experimental/preview phase for enterprise/Workspace users specifically?
  2. Are there known triggers for the “agent terminated” error (e.g., MCP server conflicts, model context limits, quota exhaustion)?
  3. What is the recommended mitigation? I’ve tried resetting onboarding via the Command Palette, starting fresh conversations, and switching models - but the issue persists.

Environment:

  • Account type: Google AI Ultra for Business (Workspace)
  • The IDE header shows “Preview” alongside “Antigravity” - is this a signal that enterprise support is not yet at GA?

Any clarity from the team or community would be greatly appreciated. Happy to share logs if directed.

Thank you.

2

Update: This issue is now FIXED in the latest Antigravity version (1.22.2)

Wanted to follow up on my original report. The persistent “Agent terminated due to an error” issue that plagued Opus (and other models) for roughly the last month has been resolved with the latest Antigravity IDE update (v1.22.2).

Key observations:

  • The issue was consistent across multiple models but was especially severe with Claude Opus - it would terminate mid-task repeatedly, sometimes within seconds of starting
  • The bug persisted for approximately a month (early March through early April 2026)
  • It was NOT a model-side issue - if it were, a simple IDE update would not have resolved it. The fix came purely from the Antigravity client update, confirming this was an IDE/platform bug
  • Since updating to v1.22.2, I have had zero agent termination errors across extended, complex agentic sessions

Request to the team: Please document what caused this regression so it doesn’t get reintroduced in future releases. A month-long stability regression in the core agent execution loop is a serious issue that significantly impacts usability and trust in the platform.

Thanks to the team for getting this fixed. The tool is working well again.