Antigravity cost tracking: agy /usage, credits, session cost

David Crowe · November 7, 2026 · Updated November 7, 2026 · 12 min read

antigravity cost usage reference

Google Antigravity does not show a bill per session. The agy CLI shows how much quota is left (/usage), how many AI credits you have (/credits), and token counts for the running session. The only dollar figure is an estimated session cost in the JSON the CLI hands to a custom status line. Google calls it “estimates based on list prices, not billed amounts.”

ACP on Antigravity. One hook covers the agy CLI, the IDE and the app. Each native and MCP tool call is recorded, and by default the destructive ones (force-push, destructive SQL, curl | sh, a recursive delete outside the project) are blocked until you approve on the console or by email.

macOS · Linux · WSL

curl -fsSL https://agenticcontrolplane.com/install.sh | bash

Windows · PowerShell

irm https://agenticcontrolplane.com/install.ps1 | iex

Needs a free workspace: --local skips Antigravity. No hook runs in a session started with --dangerously-skip-permissions. Connected, the hook sends ACP each call's tool name and arguments, and up to 200 KB of its output. Free up to 5 agents. What the installer writes · Antigravity guide · What ACP sees on each agent

The rest of this page covers how Antigravity is priced, what each screen shows, what is missing, and how to keep a running cost visible. Antigravity is Google’s IDE, app and agy CLI. The CLI replaced Gemini CLI in May 2026. Everything here is as of 2026-11-07, from Google’s docs.

How Antigravity is priced

There are two ways to pay, and they work differently.

A Google AI plan (individuals). Antigravity’s plans page describes three levels:

Plan	Quota
No Pro/Ultra plan	“A meaningful quota, refreshed weekly” and a weekly rate limit
Google AI Pro	“A high, generous quota, refreshed every five hours until the weekly limit is reached”
Google AI Ultra	“The highest, most generous quota, refreshed every five hours”, plus third-party models

Prices come from Google’s plan changes post of May 19, 2026: Pro is $20 a month, and Ultra comes in a $100 tier with “5X worth of tokens” of the Pro rate limit and a $200 tier with 20X. The same post combined Gemini Flash and Gemini Pro into “a single rate limit, drawn down according to API pricing”: a cheaper model stretches the same quota further. Non-Gemini models stay on a separate fixed limit.

Quota is not a prompt count. The plans page says the limits “correlate with the amount of work done by the agent, which can differ from prompt to prompt,” and that they “are subject to modification.” Google publishes no token numbers for any tier.

AI credits (overage). Since May 2026, credits are “used solely as an overage mechanism.” Pro and Ultra users can buy credits, and they “are consumed at standard Gemini Enterprise consumption pricing.” The plans page does not list credit prices or quantities.

Google Cloud (organizations). The pricing page offers access “via API with consumption-based pricing”, Gemini Enterprise Standard and Plus plans “starting at $30 USD per seat per month”, and a pay-as-you-go plan with a $0 seat fee. The enterprise setup connects Antigravity to a Google Cloud project with an invoiced Cloud Billing account, and billing is by consumption at Agent Platform pricing rather than through a subscription.

A Gemini API key. The CLI can skip the Google sign-in and use a key, per the authentication page:

// ~/.gemini/antigravity-cli/settings.json
{ "modelProvider": "gemini" }

export GEMINI_API_KEY="your-key"
agy

Model requests then go directly to the Gemini API, and the header shows “Gemini API key” instead of an email. Setting GEMINI_API_KEY alone does nothing, and the CLI does not load .env files.

What agy shows

Three commands, from the CLI reference:

/usage    (alias /quota)  Display model quota usage.
/credits                  View remaining G1 credits and purchase links.
/context                  Open the context usage visualization panel.

/usage refreshes your model configuration and quota from the backend and opens a Models & Quota panel. Per the /usage page, it lists remaining quota for each model group, such as Gemini models and Claude and GPT models, and “the models in a group share a weekly limit and a five-hour limit.” It reports quota left, not tokens or dollars.

/credits opens a panel with your active balance, “a breakdown of credits consumed in the current billing cycle”, and links to buy credits or upgrade (/credits page). It is per billing cycle, not per session.

The built-in status line shows remaining credits on the right, for example AI Credits: 42, and highlights when the balance drops below the warning threshold (CLI AI credits).

The IDE and app. The plans page says you “can view baseline quota usage across models on the Settings page.”

The one dollar figure: the status line JSON

Whenever agent state changes, agy runs your status line command and pipes the session state to it as JSON. The status line docs list these fields among others:

cost.total_usd                          estimated spend, including subagents
cost.subagent_usd                       estimated subagent spend
cost.estimated                          flag that the figure is an estimate
context_window.total_input_tokens
context_window.total_output_tokens
context_window.used_percentage
context_window.current_usage            input, output, cache-creation, cache-read tokens
quota.<bucket>.remaining_fraction       for example gemini-weekly
quota.<bucket>.reset_in_seconds
model.id, plan_tier

The caveat in Google’s words: “These are estimates based on list prices, not billed amounts (omitted until a cost is reported).” On a plan, you are paying a subscription and drawing down quota, so the number is what the same tokens would cost at list price, not a charge.

To keep it on screen, register a command:

// ~/.gemini/antigravity-cli/settings.json
{
  "statusLine": {
    "type": "command",
    "command": "~/.gemini/antigravity-cli/statusline.sh",
    "stack_with_default": true
  }
}

stack_with_default puts your line under the built-in one, so the credit count stays visible. A minimal script that reads the documented fields:

#!/bin/sh
# ~/.gemini/antigravity-cli/statusline.sh
jq -r '"est $" + ((.cost.total_usd // 0) | tostring)
      + "  in " + ((.context_window.total_input_tokens // 0) | tostring)
      + "  out " + ((.context_window.total_output_tokens // 0) | tostring)'

Make it executable with chmod +x. The CLI runs the script at most once every 300 ms, kills runs over 5 seconds, and disables it after 30 failures in a row. /statusline delete reverts to the default.

Where usage lives

Settings: ~/.gemini/antigravity-cli/settings.json holds statusLine, modelProvider and useG1Credits.
Conversations: saved, since --conversation resumes one by ID. The reference does not give the path.
Logs: --log-file overrides the log path. The default location is not documented.
Quota and credits: on Google’s servers. /usage and /credits fetch them.
Cloud consumption: on the enterprise route, in the Google Cloud project’s billing account.

Google documents no local usage file to parse. To keep a per-session record of the estimate, capture it in the status line script.

What is missing

A billed cost per session. The only dollar figure is a list-price estimate in a JSON payload you have to wire up.
Cost per tool call or per turn. The payload has session totals only.
Token numbers for quota. Plans are described as “generous” and as multiples of Pro, so you cannot turn remaining quota into dollars.
A per-developer team view. The docs describe Cloud billing and spend caps for organizations, but no per-user report for agy sessions.
History. Nothing keeps yesterday’s totals unless your script writes them down.

How to put a ceiling on it

On a plan, the quota is the ceiling. The setting that decides whether overage costs money is useG1Credits, off by default and not shown for business accounts (settings):

{ "useG1Credits": false }

Off, the CLI stops at the plan quota and you wait for the refresh. On, it spends your AI credits after the quota runs out. The plans page describes the same choice as the AI Credit Overages setting: Never waits for the refresh, Always uses credits and goes back to baseline once quota refreshes. For organizations, the pricing page lists spend caps across API and subscription plans under the Google Cloud route.

What these numbers leave out

Every number above is about model usage. None of it says what the agent did: which commands ran, which files changed, what was denied. Antigravity’s hooks see those tool calls. ACP uses them to record each agy tool call and check it. If someone launches agy with --dangerously-skip-permissions, Antigravity drops its hooks, ACP’s included, and nothing is recorded for that run. The Antigravity controls page has the detail, and so do the permissions reference and the full-permissions guide. The install box below sets up the recording in one line.

Frequently asked questions

How do I check usage in the Antigravity CLI (agy)?

Run /usage (alias /quota). It refreshes your quota from the backend and opens a Models & Quota panel with the remaining quota for each model group; models in a group share a weekly limit and a five-hour limit. /credits shows your AI credit balance and consumption in the current billing cycle, and /context shows how full the context window is.

Does agy show cost per session?

Only as an estimate, and only to a custom status line. The JSON the CLI pipes to a statusLine command includes a cost object with total_usd, subagent_usd and estimated. Google’s docs say these are estimates based on list prices, not billed amounts. The built-in status line shows remaining AI credits, not dollars.

Is there an agy /stats command?

Not in agy’s command reference. The documented commands for usage are /usage (quota), /credits (credit balance and usage this billing cycle) and /context (context usage). Token counts for the running session are in the status line JSON under context_window.

What happens when my Antigravity quota runs out?

On Google AI Pro or Ultra you can use purchased AI credits, consumed at standard Gemini Enterprise consumption pricing. In agy the useG1Credits setting, off by default, lets the CLI spend credits after the plan quota is exhausted. Leave it off and you wait for the quota to refresh.

Can I use a Gemini API key with agy?

Yes. Set modelProvider to gemini in ~/.gemini/antigravity-cli/settings.json and export GEMINI_API_KEY. Setting the environment variable alone has no effect, and the CLI does not read .env files. Model requests then go directly to the Gemini API.

Where to read more

Antigravity plans: quota tiers, AI credits, overage
Changes to Antigravity plans: prices and the shared Gemini rate limit
/usage command and /credits command
Status line customization: the cost, context_window and quota fields
Authenticate: Gemini API key and Google Cloud credentials
Antigravity permissions reference: allow, ask and deny rules for agy

Running Antigravity? One command adds a hook. Each tool call it sees is recorded, and a force-push, destructive SQL or curl | sh is blocked until you approve it on the console or from the email.

macOS · Linux · WSL

curl -fsSL https://agenticcontrolplane.com/install.sh | bash

Windows · PowerShell

irm https://agenticcontrolplane.com/install.ps1 | iex

What the installer writes · Antigravity guide · free up to 5 agents

Share: Twitter LinkedIn

← back to blog