Unable to use Antigravity or Gemini

Google Antigravity
1

I am unable to use Antigravity or Gemini.

I’ve been on the Ultra plan for months, but was unable to get more than 1-2 messages per day. Resource access denied or “overloaded” error on practically every message, despite having quota.

I’ve lost the money, and I’ve downgraded to lowest plan, but I will probably cancel this one as well since it is impossible to use it.

This is not some intermittent error, or mild inconvenience. The service I paid for doesn’t work, and never worked properly.
It only worked properly during the free trial, or when I was on ultra on a private, not business (workspace) account. And even then it suffered from the same auth issues everyone else had.

`Your account lacks the required IAM permission “cloudaicompanion.companions.generateChat” on resource “projects/rising-fact-p41fc”. Please contact your administrator.`

EDIT: This happened after I downgraded from Ultra to Extended AI plan. So I assume that might be an error because Extended AI no longer has access to Antigravity API. However that would not explain why it intermittently works (rare) and doesn’t. This seems like a separate issue than what I’ve been seeing for the past months - when I was extensively using (or failing to use) Antigravity, with “overloaded” errors all the time.

I have switched to OpenCode since it had a little better hit-rate before hitting this error, than Antigravity (which throws an error almost every time), or Gemini CLI.

If this is a ban resulting from that then - well cool. It worked a small amount of time - UNLIKE ANTIGRAVITY - but still was hardly usable. I never even approached my quota limits, since it is IMPOSSIBLE to actually get a proper inference, in any tool. I assume it is a ban going by your own forums and Github issue comments.

Sometimes it still works. For a message or two. Then hits the limit.

And after downgrading I am unable to upgrade to Ultra again, EDIT: searching the web says it’s probably because EMEA need to jump thru additional hoops to get it, even if they previously had it and downgraded it. Would be much better if you just said that instead of crippling a service that 1) I have never abused 2) I still paid for despite being banned for no apparent reason.

Not that I would want to upgrade - since the service DOES NOT WORK.

2

Prior to downgrading my workspace subscription from AI Ultra, I have NEVER used the subscription outside of Antigravity. I was just hoping it will start working.

After downgrading from AI Ultra I no longer see an option to upgrade to it (tho I doubt I will ever do that), which tells me my account must have been limited.

I was never informed of any limits on me and treated the errors as your infrastructure being overloaded - as the messages STATED.

You continued to charge me for the service.

To reiterate - while paying for Ultra I have never used a different harness than Antigravity. The errors said “overloaded” despite having 100% quota available, making 80% of inference calls fail at worst times, 50/50 at good times. For months.

I have since cancelled even “extended AI access” - now the error does not appear it just does not generate at all. So I conclude this might not be a ban, since I didn’t violate any terms anyway, it is just the standard of service Antigravity provides… Please confirm. And if it was a ban - why was I not informed and why did you continue billing me for the service?

What information do you need about my account?

3

Update: I’ve now confirmed I can authenticate Gemini CLI against Gemini Code Assist Standard - sign-in works, IAM permissions resolve, inference succeeds. This rules out an account-level ban: per google-gemini/gemini-cli#20632, bans cascade across Antigravity and Code Assist with specific error patterns, and I’m seeing none of those on the working Code Assist path. So whatever produced months of “overloaded” errors during my Ultra subscription was service-side, not enforcement. I see no warnings in Antigravity either - just the same errors in 8/10 inference calls as always. By looking thru the issues posted all over this forum, since the same timeline I had - months - it seems I am not the only one having these problems, and it the common experience.

Also, since Code Assist Standard is locked to Gemini 2.5. Accessing 3.x models requires Enterprise, which has a 10-seat minimum at ~$45-54/seat/month, $500+/mo minimum for a solo developer. That costs more than all my other AI licenses, of much more capable models, combined. That’s effectively “no individual access” to your current-gen models via subscription pricing. So no, I won’t be upgrading.

The combined picture: account isn’t banned, the service I paid for was just broken, and alternatives do not exist or are not priced for individuals.

And even if such a path existed, you limit usage of any available subscription paths to your tooling, which are borderline unusable, or do not provide a GUI app like every other provider. An inference call from OpenCode is identical in payload and server cost to one from Gemini CLI - the only difference is the UX which the outdated TUI app and the VSCode extension you provide do not satisfy.

I would also add, that since doing a Google search for “Gemini Code Assist / Antigravity in OpenCode” - the AI Overview by Google states that “auth plugin is recommended for Subscription users” - perhaps you should block these plugins from authenticating with your servers at all, since you can detect it, instead of banning people who were pointed down this path from within your own ecosystem as the recommended path, and are not even aware this is not allowed. Just a side-note since I am not banned.

image
image1119×1360 268 KB