Antigravity rate limits: quotas, resets and AI credits explained

Antigravity rate limits depend on your Google AI plan: without Pro or Ultra you get a quota refreshed weekly, Google AI Pro gets a quota refreshed every five hours until a weekly limit is reached, and Google AI Ultra gets the highest quota, refreshed every five hours. Google publishes no numbers, so check what's left in the App's View Usage panel or with /usage in Antigravity CLI; when it runs out, you wait for the refresh or, on Pro and Ultra, spend AI credits.

This page covers how the limits behave and what to do about them, from Google's plans page, models page, changelog and the May 2026 plans post, checked on 28 October 2026. For prices and what each plan includes, see the Google Antigravity guide. For errors that aren't about quota, see Antigravity troubleshooting.

How do Antigravity rate limits work on each plan?

Google's plans page describes the quota for each tier in words, not numbers:

Four rules from Google's docs explain how quota gets used:

• Capacity sets the baseline. Baseline rate limits "are primarily determined by available capacity and exist to prevent abuse," and Google says usage limits are subject to change.
• Harder tasks cost more. Limits track the work the agent does, so a complex task uses more quota than a simple one.
• Gemini models share one pool, priced by token. Since Google's 19 May 2026 post, Gemini Flash and Pro draw from "a single rate limit, drawn down according to API pricing." Google's example: if Flash is 8x cheaper than Pro, you can spend X tokens on Pro, 8X on Flash, or a mix.
• Claude and GPT models have their own pool. The same post says "Because of capacity limitations for non-Gemini models, they remain on a separate fixed rate limit." The model selector shows two groups, Gemini Models and Claude and GPT models, each with its own weekly and five-hour limit.

The same post set the ratio between paid tiers: the lower Google AI Ultra tier gets 5X the tokens of Google AI Pro, and the top Ultra tier gets 20X. For current prices, see the Google Antigravity guide.

Business and Enterprise accounts work differently. Google's App changelog (v2.13.0) removed the usage panel for Enterprise and Business accounts, "which do not have per-model quotas." Those plans run under Google Cloud terms in Gemini Enterprise.

When does your Antigravity quota reset?

• Google AI Ultra: every five hours, within the highest weekly limits.
• Google AI Pro: every five hours, until the weekly limit is reached. Google's wording means that once you hit the weekly cap, five-hour refreshes don't bring access back until the weekly window resets.
• Everyone else: once a week.

You don't have to guess the time. The model selector shows the reset time as a compact duration, with the full text in a tooltip (fixed in App v2.14.0). In the CLI, /usage shows your quota reset times; as of v1.2.15, agy -p "/usage" prints them in your local time zone in a terminal, while piped and --json output still use UTC.

Community reports, not official: a long-running Google AI Developers Forum thread, started in January 2026, collects Pro users reporting multi-day waits instead of five-hour resets. Many posts predate the 19 May plan changes. If you see a long wait today, check whether the weekly limit, not the five-hour window, is the one you've hit.

How do you check how much quota you have left?

In the Antigravity App, open the model selector and select View Usage. It shows Weekly Limit Remaining and Five Hour Limit Remaining for each model group. The plans page also says "You can view baseline quota usage across models on the Settings page."

In Antigravity CLI, three commands help:

/usage     Model Quotas panel: usage and remaining quota per model
/quota     Alias for /usage
/credits   Your AI credit balance and usage

/usage refreshes your quota from the backend each time you open it. The CLI's status line can also show your remaining credits, for example AI Credits: 42, and highlights it when credits run low. The Antigravity CLI guide covers the rest of the agy commands.

What do "individual quota reached" and 429 errors mean?

They mean you hit a limit, and recent releases changed how Antigravity reacts:

• Quota reached. CLI v1.3.3 (10 October 2026) changed it so "reaching your individual quota is reported right away instead of being retried." Earlier builds spent time retrying first.
• Credits can't cover the request. CLI v1.2.15 replaced a long retry that ended in the generic "Agent execution terminated due to error." with a direct message: "Your AI credits balance is too low to continue."
• Short rate limits (429). CLI v1.2.13 waits the delay the server asks for, and "stops right away instead of retrying when the delay is longer than 30 seconds or the quota is a daily or billing cap."
• 429 right after launch. App v2.11.0 "Fixed cold-start token endpoint routing to prevent HTTP 429 resource exhaustion errors on launch." If you see a 429 the moment you open the app, update.
• "Out of credits" on Enterprise. App v2.11.0 also fixed Enterprise users signed in with Application Default Credentials wrongly seeing an "Out of credits" error.

What to do when you hit one:

1. Open View Usage or /usage to see which group ran out and when it resets.
2. If only one group is empty, switch models: Gemini and Claude or GPT models are tracked separately.
3. Wait for the reset, or turn on AI credit overages if you're on Pro or Ultra (next section).
4. If the panel shows quota left but requests still fail, update and see Antigravity troubleshooting.

How do AI credits and the AI Credit Overages setting work?

AI credits are the overage mechanism for Google AI Pro and Ultra. The 19 May post says credits now serve "solely as an overage mechanism," after Google removed them from the base plans and raised base-plan Gemini entitlements to compensate.

How they work, from the plans page:

• Pro and Ultra users can use purchased AI credits, or one-time promotional credits, above the baseline quota.
• Credits are consumed at standard Gemini Enterprise consumption pricing.
• The AI Credit Overages setting decides when they're spent. Never means credits are never used automatically and you wait for the quota to refresh. Always means credits are used once the baseline quota is exhausted, and you switch back to the baseline when it refreshes.

In Antigravity CLI, the matching setting is useG1Credits in settings.json (labeled Use AI Credits in Google's settings docs), which is off by default and isn't shown for business accounts. You can toggle it in /config.

One gotcha: App v2.23.0 fixed the overages setting appearing off after a reload, which hid the credits row. If your setting looks wrong on an older build, update.

No bring-your-own-key. Google's plans page says BYOK or bring-your-own-endpoint "for additional rate limits" is not supported. The CLI can sign in with a Gemini API key instead of your plan, but then the Gemini API's own quotas and billing apply. CLI v1.2.12 made those sessions stop at once when the API reports an exhausted daily quota or a spend cap.

Do Antigravity and Antigravity CLI share one quota?

Google's docs don't say either way. The plans page ties quota to your Google AI plan, not to a surface, and lists "scheduled tasks and the CLI" among the product features every plan includes. It doesn't publish a separate CLI quota or a breakdown by surface, so check /usage in the CLI and View Usage in the App rather than assuming one pool or two.

Two related rules from the FAQ matter for heavy users. Headless mode (agy -p) is meant for direct scripting and CI, and wrapping it inside third-party agents or automating account rotation isn't supported on consumer plans. Using third-party tools to access Antigravity with your login violates the terms and can lead to suspension, so spreading work across tools to stretch a quota carries real risk.

How can you make your quota last longer?

• Use Flash for routine work. Because Gemini usage is drawn down at API prices, cheaper models stretch the shared pool further.
• Keep tasks focused. Google says complex tasks consume more quota, so break big jobs into smaller, well-scoped requests.
• Compact long CLI sessions. CLI v1.3.5 added /compact for on-demand context reduction, which shrinks what each later request carries.
• Use the other model group. If your Claude and GPT pool is empty, Gemini models still have their own limit, and the reverse.
• Decide on overages before you need them. Set AI Credit Overages to Always for deadline work and back to Never when you'd rather wait.
• Watch the meter. Glance at View Usage or the CLI status line before starting a long agent run.

Reusable instructions also save tokens, because you don't retype context. Kipt keeps prompts and Agent Skills as Markdown in a private GitHub repo you own, and Antigravity can load skills, so a tight, reusable skill replaces a long pasted preamble in every conversation.