Quota monitor & cache manager for Google Antigravity IDE. Real-time Gemini / Claude / GPT usage tracking, AI quota dashboard, usage analytics, auto-accept agent mode, and AI commit message generator.
Installation
Launch VS Code Quick Open (Ctrl+P), paste the following command, and press enter.
Real-time AI quota monitor & cache manager for Google Antigravity IDE — track Gemini, Claude, and GPT usage, visualize consumption trends, and manage conversation cache, all in one sidebar panel.
Antigravity Panel helps you stay on top of your AI model usage in Google Antigravity IDE. Get real-time quota monitoring, usage trends, and cache management—all within an integrated sidebar panel.
✨ Features at a Glance
🎯 Quota Monitoring - Real-time status with visual thresholds
📊 Usage Analytics - Interactive charts and history tracking
🧹 Cache Management - Manage AI conversation history and files
🎨 Native Integration - UI components adapted to IDE themes
🌍 Localization - Support for 15 languages including runtime notifications
🛠️ Diagnostics - Built-in connection check and error reporting
🤖 Hands-free Mode - Auto-accept agent commands for heavy workflows
✍️ AI Commit - Generate commit messages via Local LLM or Claude
⚙️ Quick Config Access - One-click editing for Rules, MCP, and Allowlist
🔄 Service Recovery - Restart, Reset, and Reload tools for Antigravity IDE stability
📸 Screenshots
Real-time quota monitoring, usage trends, and cache management in one place
🚀 Key Features
📊 Smart Quota Monitoring
See your AI usage at a glance
Visual quota display grouped by provider-defined quota pools; model view keeps individual model identities
Status bar shows remaining quota with emoji indicators (🟢🟡🔴), cache size, and a warning state when the quota connection fails
Hover tooltip showing all model quotas and reset times
Configurable warning (≤40%) and critical (≤20%) thresholds
📈 Usage Trends & Analytics
Understand your consumption patterns
Interactive bar charts showing reported quota changes over time (10-120 minutes), adaptively grouped into at most about 24 readable intervals; the card stays hidden until the first positive change is recorded
14-day history tracking with persistent storage; the latest 24 hours keep raw samples, while older data is downsampled to 5-minute intervals
Color-coded visualization by quota pool, with interval and per-pool details on hover
🔥 Usage Rate: Average consumption speed in percentage points per hour (pp/h)
⏱️ Runway Prediction: Estimated time until quota exhaustion
💳 Token Credits Tracking
Monitor your AI usage credits
Prompt Credits: Used for conversation input and result generation (reasoning)
Flow Credits: Used for search, modification, and command execution (operations)
Static Prompt/Flow rows are hidden by default; Google One AI subscription credit remains visible and the rows can be restored with tfa.dashboard.showCreditsCard
🗂️ Cache Management
Keep your workspace clean
Brain Tasks: Browse and delete AI conversation caches
See task size, file count, and creation date
Preview images, markdown, and code files
One-click deletion with automatic cleanup
Code Context: Manage code analysis caches per project
Smart Cleanup: Automatically closes related editor tabs
🤖 Auto-Accept (Hands-free Mode)
Streamline your workflow
Automatically accepts Agent-suggested terminal commands and file edits
Uses registered IDE commands first, with a panel-scoped CDP fallback for controls that are unavailable through the extension API
Runtime command discovery: accept commands are matched against what the running IDE actually registers (IDs differ between Antigravity 1.x and 2.x), so the command strategy keeps working across IDE upgrades
Checks on the configured interval; each CDP pass locates and scans the current Agent Panel once without leaving page-side observers or timers, while a short DOM-node timestamp prevents immediate repeat clicks
CDP interactions leave destructive-looking action cards for manual review. Limitation: the command-API path cannot inspect pending command text and is therefore not covered by this check. Auto-Accept remains a trust-the-agent feature; keep it off for untrusted or prompt-injection-prone tasks
Toggle on/off via the sidebar "Rocket" switch
[!IMPORTANT]
CDP Fallback Setup: For the CDP fallback to work, Antigravity must be launched with --remote-debugging-port=9222. This is only needed when the command API is unavailable due to webview sandboxing.
Recommended Setup (Dedicated Launcher):
Create a script to cleanly kill background instances before launching.
Open Settings (Ctrl+, / Cmd+,) in Antigravity IDE and search for tfa to customize:
📊 Quota Settings
Setting
Default
Description
Polling Interval
90s
How often to refresh quota (min: 30s)
Show Quota
✓
Display quota in status bar
Quota Style
semi-arc
Visualization style: semi-arc or classic-donut
Visualization Mode
groups
Show dashboard by groups or models
UI Scale
1.0
Global scale factor for panel elements (0.8 to 2.0)
Show User Info Card
✓
Show the user email and subscription tier in the sidebar
Show Prompt/Flow Credits
✗
Show the static Prompt/Flow rows; Google One AI remains visible
Show GPT Quota
✗
Whether to display GPT family models in the panel
History Range
90 min
Time range for usage chart (10-120 minutes)
Warning Threshold
40%
Status bar turns warning color at this level
Critical Threshold
20%
Status bar turns critical color at this level
💾 Cache Settings
Setting
Default
Description
Show Cache Size
✓
Display cache size in status bar
Check Interval
120s
How often to check cache size (30-600s)
Warning Threshold
500 MB
Status bar color warning when exceeded
Hide Empty Folders
✗
Hide empty folders in Brain and Code Tracker trees
Auto Clean
✗
Automatically clean cache when exceeded
Auto Clean Keep Count
5
Number of newest tasks to keep during auto-clean (1-50)
🔧 Advanced Settings
Setting
Default
Description
Server Host
127.0.0.1
Address of Antigravity Language Server
API Path
/exa...
gRPC-Web path for User Status
Auto-Accept
✗
Enable hands-free acceptance of Agent actions
Auto-Accept Interval
800ms
Polling interval for Auto-Accept (200-5000ms)
Debug Mode
✗
Enable verbose logging in Output panel
🤖 Commit Message Settings
Setting
Default
Description
Endpoint
http://localhost...
API URL (Ollama, Anthropic, OpenAI compatible)
Model
llama3.2
Model name (e.g. llama3.2, claude-3-haiku)
Max Diff Size
80000
Max characters of diff to send to LLM
Format
conventional
Message format (conventional or simple)
🔒 Privacy & Safety Disclaimer
Your data stays under your control.
Antigravity Panel does not collect or store analytics data. Quota, cache, diagnostics, and core panel operations communicate with local Antigravity IDE components on your machine. The optional commit message generator sends staged diffs only to the LLM endpoint you configure, which may be local or external.
Experimental Feature Notice:
The Smart Quota Monitoring feature relies on internal metrics exposed by the local Antigravity environment. This functionality is experimental and provided "as-is" to help users better understand their personal usage. It is not an official Google product and may be subject to changes in future IDE updates.
🤝 Contributing
We welcome contributions. Development, debugging, and testing must be done in Antigravity IDE with the local Antigravity Language Server available. Use Node.js 24+ and npm for local builds.
See CONTRIBUTING.md for setup, Extension Host debugging, test commands, coding standards, localization rules, packaging, and PR workflow.
If you find Antigravity Panel helpful, please give us a Star 🌟 on GitHub. It's the best way to support our work and help others discover it.
@AMDphreak - Sidebar title fix, Gemini Flash/Pro grouping, quota reset window alignment with API cycles, and Claude+GPT shared pool display.
@chonkydonkers - Display user tier available credits in status bar and sidebar.
@vincenzofabiano92 - Synchronous command registration, connection stability optimization, Italian NLS localization, and server integration test runner (v2.6.0).
🌐 Localization Policy
To ensure technical consistency and professional standard across all 15 supported languages:
UI Labels & Technical Terms: Remain in English (e.g., Rules, MCP, Auto-Accept, Reset Status).
Tooltips & Descriptions: Fully localized to provide detailed explanations in the user's native language.