Antigravity's usage quota is very restrictive

Hi everyone,

I want to open a discussion regarding the current quota system implemented in the new Antigravity CLI (agy v1.0.0 / 2.0) specifically for users under the Google AI Pro subscription.

The current behavior makes the CLI practically unusable for real-world, daily software development, and it feels like a massive regression compared to what we had before.

The Comparison:

  • With Gemini CLI: As a Pro user, I used to have a reliable, predictable quota of 1,500 requests per day. This allowed me to handle intense coding sessions, explore codebases, and iterate without constantly worrying about getting blocked.

  • With Antigravity CLI (agy): The new “Compute Effort” unifications and 5-hour rolling windows are completely unpredictable. In a recent real-world scenario working on a standard repository, the CLI reached the individual quota after processing roughly 1,000 lines of code. The parallel sub-agent architecture drains the token pool exponentially due to redundant context injections.

The LATAM Context & Aggressive Monetization: Paying for a Google AI Pro subscription represents a monthly investment for developers living and working in LATAM. Despite paying for a premium plan, the system forces an unsustainable financial model: if the multi-agent architecture burns through your baseline quota in a single session, you are completely locked out of your workflow.

Your only options are to manually buy additional AI Credits (on top of the Pro subscription) or wait for a massive 7-day reset cycle due to the weekly cap. Forcing users to purchase overages just to finish a standard daily coding session—or face a one-week penalty box—is an unacceptable user experience for paying subscribers.

The Frustration with the Pro Plan: It is incredibly frustrating to hit a hard wall (Individual quota reached. Contact your administrator to enable overages) after a few minutes of terminal interaction. To make matters worse, there is a persistent permission bug in the CLI that blocks the use of emergency backup AI Credits when agents call external tools or MCP servers (like Linear), leaving individual developers completely stranded even if they want to pay for overages.

The Competitive Context: If you look at the current landscape, Claude Code just significantly increased its limits for Pro users and handles context caching much more efficiently, making the $20/month value proposition from Anthropic drastically superior for terminal-based development.

As developers, we need predictability and a high-throughput ceiling. Right now, agy feels like an expensive toy that breaks down under any standard backend or app development workflow.

Is the product team aware of how restrictive this new quota system is for individual Pro subscribers? Are there any plans to fix the credit-override bugs in the CLI or re-evaluate the baseline capacity before the official Gemini CLI deprecation on June 18th?

Thanks,
Juan

1 Like