Model Usage Display

English | 日本語

Keeps the remaining quota and time-until-reset of each Antigravity AI model (Gemini / Claude / GPT) visible in your status bar. Installable in VS Code, Antigravity IDE, and other VS Code-based editors.

[!IMPORTANT] Prerequisite: Antigravity IDE is required. This extension reads quota from the Antigravity Language Server running on your machine. So no matter which editor you install it into, Antigravity IDE must be installed on the same machine and running (signed in, with its Language Server up). Without it you will see the error "Antigravity Language Server not found".

Features

• Status bar: always shows the remaining percentage of your lowest model; hover for the full list
• Detail dashboard: click the status bar for every model's progress bar, remaining percentage, and time until reset
• Auto refresh: configurable polling interval (60 seconds by default)
• Color coding: warning color at 30% or below, danger color at 10% or below (thresholds are configurable)
• Threshold notifications: alerts you when a model drops below a threshold
• Reset notifications: alerts you when the quota reset time arrives

Models
──────────
Claude Opus 4.6
■■■■□□□□□□ 45%  ↻ 3h21m

Gemini 3.1 Pro
■■■■■■■□□□ 71%  ↻ 2h05m

How it works

It connects directly to the Language Server that Antigravity IDE runs internally — no extra sign-in, everything stays local:

1. Find language_server in the process list and extract --csrf_token (macOS/Linux: ps, Windows: PowerShell Get-CimInstance)
2. Enumerate that process's listening ports (macOS/Linux: lsof, Windows: netstat)
3. POST the Connect RPC /exa.language_server_pb.LanguageServerService/GetUserStatus to each port with an X-Codeium-Csrf-Token header

Because these internals are unstable, the extension displays the remainingFraction / resetTime the server returns as-is and does not try to interpret quota semantics (5-hour rolling windows, weekly caps, and so on).

Variants sharing one quota pool (for example Gemini 3.5 Flash (Low/Medium/High)) are collapsed into a single group.

Installation

From VSIX

Install Antigravity IDE first (see the prerequisite above). Then download the .vsix from Releases and, in your editor (VS Code / Antigravity IDE / any VS Code-based IDE), go to the Extensions view → … menu → Install from VSIX → pick the downloaded .vsix.

Or from the CLI (replace code with your editor's CLI name):

code --install-extension model-usage-display-x.y.z.vsix

Full paths on macOS, if the CLI is not on your PATH:

"/Applications/Visual Studio Code.app/Contents/Resources/app/bin/code" --install-extension model-usage-display-x.y.z.vsix
"/Applications/Antigravity IDE.app/Contents/Resources/app/bin/antigravity-ide" --install-extension model-usage-display-x.y.z.vsix

From source

npm install
npm run compile
npm run package   # produces model-usage-display-x.y.z.vsix

Settings

Commands

• Model Usage: Show Details — open the dashboard (same as clicking the status bar)
• Model Usage: Refresh Now — refresh manually

Supported environments

• Editors: VS Code 1.74 or later, plus VS Code-based IDEs such as Antigravity IDE. Quota comes from the Antigravity Language Server process running on the machine rather than from the editor itself, so Antigravity IDE must be running regardless of which editor you use
• OS: macOS and Windows (both verified). Linux is implemented but untested
• Display language: English (default) and Japanese, following your editor's display language

Limitations and disclaimer

• Antigravity IDE has to be running, because the extension connects to the Language Server inside it — this applies equally when the extension is installed in another editor
• It depends on Antigravity's private, undocumented internals, so an IDE update can break it without warning. This is an unofficial personal project, not affiliated with Google, and comes with no warranty (MIT license)
• All traffic goes to 127.0.0.1 (the local Language Server) only. Nothing is sent to external servers, and there is no telemetry

Privacy

The only data this extension handles is the quota information the Language Server returns: model names, remaining amounts, and reset times. It is held in memory only — never written to disk, never transmitted.

Prior art

• antigravity-usage — the connection approach (CSRF token discovery, Connect RPC)
• AntigravityQuotaWatcher
• antigravity-usage-monitor

License

MIT