Subscription: Google AI Pro × 5 accounts
Platform: Antigravity IDE (Windows 11)
App version: Latest as of March 2026
Region: Italy
Issue Summary
I am a professional developer running 5 separate Google AI Pro accounts to sustain a continuous development workflow. All 5 accounts hit a simultaneous 167-hour lockout on Claude Sonnet 4.6, Claude Opus 4.6, GPT-OSS 120B, Gemini 3.1 Pro High, and Gemini 3.1 Pro Low.
This is not a standard quota exhaustion. The lockout hit all 5 accounts at the exact same time, despite each account having independent quota allocations.
Observed Behavior
- All 5 Pro accounts show ~167h refresh timer for Claude, GPT-OSS and Gemini 3.1 Pro (High and Low) simultaneously
- Gemini 3 Flash on the same accounts shows the correct ~5h refresh (Pro tier working as expected)
- The lockout is server-side: logging out, clearing
%USERPROFILE%\.antigravityand re-authenticating does NOT resolve it - I spun up a virtual machine on a completely separate network/IP: the problem persists identically, confirming the lockout is tied to the Google accounts themselves — not to the local machine or IP address
- The Antigravity UI quota screen shows available quota for all accounts — but generation still returns
RESOURCE_EXHAUSTED 429 - This confirms the discrepancy between the quota display system (
fetchAvailableModels) and the generation backend (generateContent) reported in other threads
Root Cause Hypothesis
Based on my investigation and cross-referencing reports from this forum (Bug AG-859, thread #126125, #123195), I believe this is a combination of:
-
IP-level rate limiting: Multiple authenticated sessions originating from the same IP address trigger Google’s anomaly detection. The system applies a shared weekly lockout at the IP level — not per-account — causing all 5 accounts to be locked simultaneously even though no single account exhausted its individual quota.
-
Background quota drain: As reported in thread #125399, Antigravity IDE silently consumes Claude quota even when completely idle and no prompts are being sent. With 5 IDE instances open, this passive drain is multiplied ×5, accelerating quota exhaustion dramatically.
-
Metadata Entitlement Sync Failure: Once the weekly cap triggers, the backend erroneously applies the Free-tier fixed epoch reset instead of the Pro rolling 5-hour window.
Impact
This is a complete business continuity failure. I rely on Antigravity as my primary development environment. With all 5 accounts locked simultaneously for 7 days, and the issue confirmed even on a VM with a different IP, I have no fallback within the Google AI ecosystem. I am being forced to migrate my entire active project to alternative IDEs mid-development, causing significant productivity loss.
Requested Actions
- Immediate manual quota reset for all 5 affected accounts
- Clarification on whether IP-level rate limiting is intentional policy or a bug
- Fix for background idle quota consumption — the IDE should not drain quota when no prompts are active
- Separation of the quota display system from the generation backend so they reflect the same state
- Official communication on the status of Bug AG-859 / Metadata Entitlement Sync Failure
Related Reports
- [BUG] Antigravity IDE - Critical Quota Error (7-day lockout): thread #114724
MAJOR ALARM - Pro Paid Users Locked for 167 Hours: thread #123195- Antigravity IDE consuming Claude quota automatically while idle: thread #125399
- The “Pro” Tier Lockout - Post-Mortem: thread #126323



