I’ve got Google AI pro and have been using Antigravity on a mix of Pro 3 high and low then 3.1 from last week. Up until last week I never had any issues with quotas, they rarely ran out and always refreshed every 4 hours.
This morning I go back to a project I was working on yesterday and notice my Pro quota is down to 20% and the refresh is in 4 days!!
It’s just a personal project but it seems like something has changed. Last week (and since Christmas) the quota have been good - I’ve only reach zero a couple of times and the refresh was always 4 hours max.
The frustrating thing is the lack of transparency around quotas. The plan talks about 4 hour quota refreshes, but then says it’s not guaranteed. Not sure how anybody could use this stuff commercially when it can all change on a whim.
It’s also not clear what actually uses quota, and how we can influence its use. Letting the dev make choices about planning/fast mode and high/low pro versions… that is all stuff the ai should be deciding.
Most lacking is the AI’s concept of a project. Every request it trawls a bunch of files. It doesn’t seem to have concept of the project structure, of working to an overall spec. It’s like a dev coming at a project afresh every time their asked to make a small change. That’s fine for small projects but for larger projects it’s silly.
It might be that my small project has hit some sort of complexity wall that’s causing a sudden churn of credits, but it’s really not big, and very modular.
I’m not complaining about quotas as such. It makes sense that we can’t have infinite resources. But we should have more transparency. I really like the direction that antigrav is going in, but we need certainty, consistency and confidence with the tools we’re working with.
4 Likes
I was having this issue a few weeks ago. I had used it without issues and rarely burned through more than 1x 5hr quota a day, then suddenly my quotas were disappearing faster than I could get anything done. We’re talking 6 prompts, not even major refactor. Then, multi-day lockout penaltybox and the most frustrating part is their site is terrible and I feel like info about their quota system is purposely obfuscated.
However, I FOUND what MY specific issue was. I had accidentally enabled the Stitch MCP.
Once I removed this from my setup, it went back to normal and honestly I have not been below 40% on any 5hr quota window in over 7 days now and I feel like my usage is back to what they consider “acceptable” for the consumer subscriptions.
I don’t know if this info can help anyone else but I hope that it does. I know not all problems are created by the same thing, but check to make sure you didn’t enable an MCP server that is blowing up your quotas. Goodluck homies!!! I know it’s frustrating…I was ready to break my monitors at one point…..
2 Likes
Thanks for highlighting this. I did wonder if I’d done something similar - there’s also some chatter around some of the 3rd party extensions eating credits.
This morning I found my quotas were suddenly at 100%. That was way before the quotas were due to be reset. Today, although I’ve not used it heavily, the quotas have stayed at 100% (using pro low as I’m implementing and bug fixing rather than doing anything architectural). So either the AI gods have smile upon me, or somebody’s tweak whatever was causing the issue!
I suspected some of the same things but I can’t dog on things I don’t actually know about. I thought about it extensively too, like could things that are sending heartbeats causing issues? I highly doubt the antigravity toolkit plugin is causing unseen issues because I don’t really have anything beyond that installed for plugins. Rainbow CSV and python and thats about it. Unless you meant something else by extensions (I am really tired).
What we really need is usage clarity and metric we can look at from google you know?
Hi @Splodes_Azena,
Thank you for sharing your solution! We appreciate you documenting this fix for the community!
Yes, an extension like Rainbow CSV or the python extensions were written for generic VS Code, so they don’t ‘know’ they’re running in antigrav, and are probably happily unaware of gemini, less be doing anything to consume gemini quota.
The extension that was being discussed is one I haven’t used but it sounded like it was maliciously calling gemini - or there was the suspicion that it was - and it was asking for credentials to do that. It’s similar to bad actors finding a google api key accidentally saved to source control and using it to run up big debts using it to call apis.
There’s a bunch of antigrav helper extension that people have written since antigrav was introduced - I was using Antigravity Quota (super small), recently swapped to Toolkit for Antigravity (bit prettier). But they just check quota balances, that shouldn’t count against quotas. It’s not using AI calls to do that, so shouldn’t be hitting credits.
I agree better clarity is needed. I’m guessing that the quotas are some function of number of tokens in and number of tokens out. Gemini is (needlessly) very chatty even when told (in rules) to make responses snappy, but who knows if that is impacting things.
The behaviour I saw Monday where single requests were using up 10-20% of quota was totally different to normal behaviour. Today, I’ve refactored my whole project, and added several new features, on gemini pro high, and it’s hardly touched the quota. Added to the way my quota reset on Monday night, when it was saying it wouldn’t reset for 3 days, it says to me there was a problem Google-side and it got cleared down. I’m now happy that I can puttle along again, but I’m glad I’m not using this stuff commercially. Unless it’s running locally it’s just too open to somebody throwing a kill switch.