About the Google Antigravity category

[Logan_Kilpatrick]

Questions, feedback, and discussion around Antigravity, an agentic development platform.

Get support, ask questions, and share what you’re working on with the community. Helpful links:

• Docs

• Changelog

[sarahL]

While the concept is remarkable, the platform’s speed is significantly lacking.

[behomonth]

quota problem experiencedIs management aware of the data quota issues? What is the plan to resolve them? We only want the promised 5-hour data quota renewal feature for Antigravity.

[ID1]

Questions, feedback, and discussion?
I have it for here…

1. (discussion) For some reason, I think that the Cursor’s guys hate you.
   But they are greedy. They lost me over their greed.
   You just can’t work properly when you think about money every step down the road.

2. (feedback) I love Antigravity…

3. (Questions) Will you be implementing the chat renaming feature?
   Other than that?

   Thank you guys for such a wonderful product.

[joelaug]

Hi team,

I’m on Google AI Pro subscription and using Anti-Gravity heavily. Last week, my quota reset went from the advertised 5-hour cycle to 5-6 days after renewal — exactly like many others are reporting here.

• Account: - Plan: Pro (renewed )

• Models affected: Gemini 3 / Opus / etc.

• Screenshot attached: [add print do contador mostrando os 6 dias]

Docs say 5 hours, but it’s clearly a silent weekly cap now. This kills workflows — can’t even test prompts without waiting forever.

Already tried chat support and email — got “we’ll investigate” but nothing changed.

Is this a bug, policy change, or can you manually reset? Please fix ASAP, tons of paying users are blocked.

Thanks!

[Igor]

I am paying €144.99/month for Antigravity Ultra, and I am completely locked out of my account because your automated security system is fundamentally broken.

Here are my active, unresolved Case IDs:

• Support Case: 3-1665000041145
• Identity Case: 8-7424000040005
• Email Thread: 3-9609000040423

The Issue: I work sequentially across 4 machines (3 PCs, 1 MacBook). I use them ONE at a time. Zero overlapping sessions. No bots, no OpenClaw, no API scrapers.

Because my hardware ID changes when I switch work computers, your bot falsely shadowbanned me. I am now hard-blocked in the IDE with this exact error:
“There was an unexpected issue setting up your account. Please try again later.”

Instead of actually looking at my backend account status, frontline support is pretending this is my fault. They sent me a completely useless, copy-paste troubleshooting checklist (reinstall, clear cache, verify age, check region).

I did every single step just to prove them wrong. Obviously, none of it worked because this is a backend account lock, not a client-side glitch.

This exact same false flag happened earlier this month, and a human engineer had to manually unban me. Now your bot has done it again.

I need URGENT Tier 2 engineer to look at my connection logs, verify my sessions do not overlap, and manually clear this backend block immediately.

[Anthony_10]

I’m not able to help but thanks for posting your experience questions and situation. Hopefully my comment will help get some attention to your issues, good luck

[Anthony_10]

Hello Igor I will tell you the same thing I told the previous poster, not able to help but thanks for posting your experience questions and situation. Hopefully my comment will help get some attention to your issues, good luck

[took]

WSL2 Crash: 11GB Core Dumps & I/O Saturation during Antigravity Server SIGSEGV (WSL Docker Backend)

The Issue:
The WSL2 instance terminates abruptly during heavy LLM Supervised Fine-Tuning (llmsft-sft) running within a Docker container (WSL2 backend).
This is not a simple container exit but a full VM termination (vmwp.exe).

Objective Evidence:

1. Massive Dump Files: Multiple 11GB dump files are generated in %TEMP%\wsl-crashes\ (e.g., wsl-crash-…-antigravity-server_…_node-6.dmp).
2. Signal 11 (SIGSEGV): Filenames consistently end in -11.dmp, indicating a Segmentation Fault in the antigravity-server (Node.js).
3. Silent Failure: No logs are recorded in dmesg or journalctl because the VM hangs/terminates during the 11GB I/O operation.
4. Launcher Inconsistency: The launcher script references the wrong WSL_EXT_ID (ms-vscode-remote.remote-wsl instead of
   google.antigravity-remote-wsl) and helper scripts are missing.

Logical Cause:
When the Docker container (llmsft-sft) consumes high memory resources, the antigravity-server (Node.js) on the WSL host encounters a
SIGSEGV—potentially exacerbated by the extension ID mismatch. The server then attempts to write an 11GB core dump from the guest to the
Windows host. This massive disk I/O saturates the virtual interface, causing the Hyper-V worker to hang and eventually terminates the entire
WSL instance.

[Winsera_Techology]

Hi, I am google AI pro user and i found everytime i run some task in editor and agent manager, when i try to open a tasks in agent manager that runs in editor, the apps suddenly crashed, the task progress stopped and the app need to relaod. this is a poor experience, please fix it

[Dineshwaran_P]

Dear Google AI Support Team,

I am writing to report critical issues with my Google AI Pro subscription and Antigravity IDE that are completely breaking my workflow. I am facing a severe quota management bug, persistent authentication loops, random application crashes, and context failures, all of which are preventing me from using the premium service I am paying for.

Account Details:

• Subscription: Google AI Pro (Active through July)

• OS Environment: Ubuntu

Problem Description: I am experiencing a massive 140-hour quota lockout in Antigravity, which directly contradicts the advertised 5-hour refresh cycle for Pro subscribers. Additionally, the application’s basic login functionality is failing, the IDE frequently crashes or loses context, and it fails to consistently read repository configurations. The service worked fine for the first two days, but from the third day onward, these frustrating bugs have made my Pro subscription feel like a huge mistake.

What I Was Doing When Locked: I opened Antigravity and the UI falsely displayed my quota as 100% available. I submitted a single, small input to the Gemini Pro model. Instantly, the system hit the total quota limit and locked me out, displaying a 140-hour wait time to refresh. I even tested this on a friend’s Pro account to verify—it also showed 100% available, accepted one small prompt, and instantly locked them out for 140 hours.

Expected Behavior: As a Google AI Pro subscriber, I should experience:

• Accurate quota UI that displays my true limits, not a false 100%.

• A maximum wait time of 5 hours for my quota to refresh.

• A seamless, successful login on the first attempt when opening the IDE.

• Stable generation without sudden crashes or context corruption.

• Automatic and consistent utilization of custom skills and guidelines configured in the repository’s .agent file.

Actual Behavior:

• Severe Quota Lockout: A 140-hour (nearly 6 days) lockout period triggered immediately after a small prompt.

• Authentication Loop: I am required to go through a 3-attempt login loop just to get into the app (Open → Login fails → Close → Reopen → Login fails → Close → Reopen → Login successful on the 3rd try).

• Sudden Errors & Forced Restarts: The IDE frequently stops mid-generation with a generic “error occurs” message. I cannot simply continue; I am forced to completely close and reopen Antigravity to resume work.

• Context Collapse & False Outputs: Sometimes, if an error occurs and I am able to click “continue,” the agent’s previous context, thinking, and data completely collapse. This causes the system to crash momentarily, recover, and then generate completely false outputs based on broken context.

• Inconsistent .agent Parsing: I have configured custom skills and guidelines in the .agent file. While the agent occasionally picks these up automatically, most of the time it ignores them entirely. I have to constantly manually instruct the agent to use these established skills, defeating the purpose of repo-level configuration.

Troubleshooting Steps Already Taken:

• Verified my Google AI Pro subscription is fully active.

• Tested and replicated the exact same false-100% quota bug on a second Pro account.

• Attempted to restart the IDE to bypass sudden errors, which only sometimes works but often results in corrupted context.

Impact: This issue is incredibly frustrating and makes the Pro subscription unusable for its intended purpose. I upgraded specifically for the 5-hour refresh cycles, but going from an advertised 5 hours to a 114-240 hour lockout over a simple input is unacceptable. Furthermore, struggling to simply log into the application, constantly having to manually remind the agent of its configured skills, and losing my generation context to sudden crashes completely destroys my development flow and wastes even more time.

Request:

• Please investigate the synchronization bug causing the 140-hour lockouts and the false 100% UI display.

• Reset my quota immediately and ensure my account correctly follows the 5-hour refresh schedule.

• Investigate and patch the caching issue causing the 3-attempt login loop.

• Address the stability issues causing sudden “error occurs” halts and the subsequent context collapse.

• Fix the .agent configuration parsing so the agent reliably and automatically utilizes repository-level skills without requiring manual prompting every time.

[Dev_Maldaner17]

I’m having similar quota issues. I had just renewed my subscription, and after running an analysis in the app, it completely depleted the quota for Sonnet/Opus and then for Gemini 3.1 PRO (something that would normally take days to run out, as it wasn’t that complex).
I hope they fix these bugs soon.

[sj00]

Is there a project specific local scope for MCP configuration?
(Maybe something like mcp_config.json within .agents dir)

[illuminato_salvatore]

You are not alone. This forum is full of topics, plenty of complaints about quota not being 5-hours refreshed as advertised, Google and its silence are making a lot of noise. Just cancel subscription here and switch to a different provider for next month. Hopefully they will fix the issue where paying customers get treated as paying customers

[Quoc_Vo]

Same issue 8 Mar 2026

[lyh]

I am writing to report a critical issue regarding my AI Pro subscription quota and a strange UI discrepancy I’ve noticed between two of my Pro accounts.

1. Abnormal Quota Cooldown (Bug): On my primary account the refresh timer for high-end models (Claude 4.6 Thinking, GPT-OSS 120B, etc.) has suddenly jumped to 124 hours and 41 minutes. As an AI Pro subscriber, the refresh cycle should typically be much shorter (around 5 hours). This has effectively locked me out of using the models I paid for.

2. UI Discrepancy: I own two AI Pro accounts. One shows the Model Quota as a continuous progress bar, while the other shows a segmented (grid-style) bar with a manual “Refresh” button in the top right corner.

• Why is there such a difference between two identical subscription tiers?

• I would prefer the UI with the manual refresh button as it seems more responsive.

Environment:

• Platform: Antigravity Desktop Client

• Subscription: AI Pro

• Account: `

I have already submitted a bug report via the internal “Provide Feedback” tool (with logs attached), but I wanted to post here to see if others are experiencing the same “120-hour lock” or if this is a known sync issue.

Could a moderator or staff member please look into my account status and reset the quota cooldown?

Attached is the screenshot of the abnormal 124-hour cooldown.

[lyh]

I am experiencing a severe issue with my model quotas. I have barely used the application today, yet my quotas for almost all premium models have been completely exhausted out of nowhere.

As shown in the attached screenshot:

• The quotas for Gemini 3.1 Pro (High/Low), Claude Sonnet/Opus 4.6, and GPT-OSS 120B are entirely depleted.

• More importantly, they are showing an abnormal refresh time of over 75 hours, completely ignoring the standard refresh cycle.

• Only Gemini 3 Flash is showing a normal refresh time of roughly 5 hours.

Since my “AI Credit Overages” is disabled and I haven’t done any heavy token-consuming tasks recently, this seems to be a system bug incorrectly calculating usage or locking the account.

Could you please investigate this quota sync issue? It is currently blocking my workflow. Thank you!

[Idris_Musluman]

Like everyone else, I’m experiencing data limit issues. It must be a general error or bug. My data limit reset itself without any usage, and I saw 4-6 day restrictions. I suspect it’s due to yesterday’s update. They should provide a general announcement.

[Fei_Lu]

Hello Support Team / Community,

Specifically:

• Quotas are appearing as “exhausted” even when I haven’t used them since the last refresh.

• The refresh countdowns are showing unusually long times (e.g., 6 days or 7 days) instead of the expected 5-hour rolling window.

I have attached a screenshot showing my current status where multiple models are locked out for nearly a week. I am currently on a Pro subscription and expect more consistent access.

Is there a known issue with the quota tracking system or are there new policies I should be aware of?

Thank you for your assistance.

[McD]

i have the same issue, since i went to sleep my quotas been drained…