pro plan , same problem
I posted a fix, but half the people says it didn’t work for them while the other half told me it did. Last time I posted a link to the thread someone yelled at me for helping, so I’ll wait to see if you want the info.
I’m able to use it nonstop no issues the past 2 days with this fix.
I have the same issue since 4 months and since 1 month it’s just too blatant and impossible for normal usage.
I’ve encountered the same issue inside Antigravity. When using Claude, you run into frequent timeouts and “high traffic” errors, but everything runs smoothly when you stick with Gemini. There are no timeouts with Gemini, and it handles long development sessions much more reliably.
The problem starts when you begin with Gemini and then switch to Claude in the same session. Token usage feels much faster, and overall stability drops significantly. Claude inside Antigravity just doesn’t seem to behave consistently for continuous development work.
At this point, it feels like the safest approach inside Antigravity is to stick with Gemini for active coding sessions. Claude seems to work much better outside of it, like in Claude Code or directly in VS Code/Cursor-style setups. Inside Antigravity, the Claude integration feels more like an API layer with extra moving parts, and it doesn’t seem as optimized as what you see in dedicated coding environments like Cursor or Windsurf.
Actually for Claude it happens even more often, or at least was happening mostly on claude.
Which is nonsense, we pay a huge amount of money to get not working IDE, while for claude users it works perfectly.
Right now Google offers wasting output limit without generations and waste of money by us.
It cannot look like that, this “Our servers are experiencing high traffic right now, please try again in a minute.”
Is the issue made by developers of antigravity and they should fix it in 2 hours max 2 days-weeks. Not months and no solution.
Greetings from Italy. I’m an Ultra subscriber and at this point I’m honestly just exhausted and frustrated. For days Google Antigravity has been basically unusable because of constant “high traffic” errors. I keep trying to access something i am paying for, and it simply doesn’t work. It’s hard to accept that a premium user ends up with the same experience as someone on the free tier. It makes the whole subscription feel pointless. What’s made this even worse is the billing situation. I tried to cancel on the day of renewal specifically because of these ongoing issues, but I was still charged for another month anyway. Then my refund request was denied without any real explanation. It didn’t feel like a proper response, just something generic to close the case. At this point, it doesn’t feel like just a technical problem anymore, it feels like paying customers are being ignored. Being charged for a service that isn’t working and then not being given a fair refund is incredibly frustrating. I’m now looking into this under Italian consumer protection laws (Codice del Consumo), because charging for something that can’t actually be used and refusing a reasonable refund is not acceptable. Google really needs to address both the reliability issues for Ultra users and the way these billing situations are handled, because right now the whole experience feels unfair.
@Thatipamula_Namratha please pass this on to the relevant team.
Can you send the link of the post you did to fix it ?
I’ve been struggling with this exact same issue for the past couple of weeks. The constant 503 Error: Our servers are experiencing high traffic right now makes it impossible to leave the agent running for long tasks, especially when trying to save credits using Gemini 3.1 Pro.
Since there isn’t an official auto-retry feature implemented in the extension yet, I wrote a lightweight Python workaround to keep things moving.
It’s a simple RPA bot that monitors the screen, detects the blue “Retry” button when the error pops up, and automatically clicks it for you after a smart delay so the loop never breaks.
You can grab the script and see the instructions here:
Antigravity-AutoBot on GitHub
Quick Start:
- Clone the repo or download the ZIP.
- Run
pip install pyautogui opencv-python pillow screeninfo - Execute
start_bot.bat(on Windows) and leave your VS Code window visible.
It’s a temporary “band-aid” fix, but it works perfectly if you want to leave your agent coding overnight. Hope this helps someone out there until we get an official patch! 
Let’s skip the pleasantries. The constant “Our servers are experiencing high traffic right now” (503) errors have crossed the line from a temporary annoyance to a critical, systemic failure of your service. And to be absolutely clear: we are experiencing this on the PAID tiers.
We are developers and founders who integrated your API and, most importantly, gave you our money. In return, we expect a reliable, enterprise-grade tool. Instead, we are getting the stability of a weekend hobby project.
Here is the harsh reality of what your lack of server capacity is doing:
-
Breaking our production: Our apps are failing, our end-users are furious, and our own businesses are taking a reputational hit.
-
Burning our money: We are paying for an API that is essentially a lottery.
What we demand immediately:
Fair Compensation: You are charging us for a service you are actively failing to provide. We expect immediate compensation in the form of massive free token grants or extended subscription periods for all affected paid accounts. It is the absolute bare minimum for the downtime and broken production you’ve caused.
The Ultimatum: The AI market is highly competitive. We have excellent alternatives. If you do not stabilize this API and compensate the developers who funded your product, we will take our billing accounts straight to your competitors.
And trust us, when we migrate, we won’t do it quietly. We will make sure the broader developer community knows exactly why building on Google’s AI infrastructure is a technical and financial liability.
Fix the servers and compensate your paid users. Or watch us leave.
its actually even happening for claude web and grok.ai…. i don’t why this high traffic errors is taking these big companies to resolve moreover for their paying customers!!!
same as Constant! can you send the link to fix this?
This problem has persisted from February until now and has not been resolved. Previously, it could still be used, but now even sending a “hello” in a brand new dialog window will prompt “Our servers are currently experiencing high traffic, please try again in a minute.”
this is quite frstrating getting an annual plans and having this issue i have two plans both the same fing thing
You can prompt the model to try again or start a new conversation if the error persists.See our for more help.
the error still persist in 7th May, free plan but idk what to do lol
I have started 10 new conversations, prompt again the same model
I have started a new project in a new folder
I have changed the models
Same problem in agent manager
All for nothing 
