Antigravity CLI

Use this skill when you need a coding agent from Google for code generation, refactoring, review, debugging, or any programming task. Antigravity CLI (agy) is Google's official terminal-based AI coding assistant that provides autonomous code editing, multi-step reasoning, and file manipulation. Trigger when user mentions 'coding agent', 'code generation', 'refactor code', 'debug code', 'review code', or needs autonomous programming assistance.

Akmal Dirgantara@akdira

Install

openclaw skills install @akdira/antigravity-cli

Antigravity CLI Skill

Use Google's Antigravity CLI (agy) as an autonomous coding agent for code generation, refactoring, debugging, and multi-step programming tasks.

When to Use

User asks to generate, refactor, debug, or review code
User needs autonomous code editing across multiple files
User wants a Google-based coding agent (alternative to Claude Code)
User mentions "coding agent", "code generation", "refactor", "debug", "review code"
Task requires multi-step reasoning about code structure
NOT for: Simple file reads, grep searches, or basic edits (use direct tools instead)

Prerequisites

Requirement	Detail
Binary	`agy` must be installed at `/root/.local/bin/agy` or in PATH
Authentication	OAuth login required (one-time interactive setup)
Network	Internet access to Google AI endpoints
Billing	FREE — uses Google AI Studio account limits (no subscription cost)

Installation

Linux / macOS (Recommended)

bash

curl -fsSL https://antigravity.google/cli/install.sh | bash

This installs agy to ~/.local/bin/agy and adds it to PATH.

Windows (PowerShell)

powershell

irm https://antigravity.google/cli/install.ps1 | iex

Windows (CMD)

cmd

curl -fsSL https://antigravity.google/cli/install.cmd -o install.cmd && install.cmd && del install.cmd

Verify Installation

bash

agy --version
# Expected output: 1.1.21 (or newer)

Authentication

First-time setup requires interactive OAuth login:

bash

# Run interactive mode (requires TTY)
agy

# Follow the prompts:
# 1. Select "Google OAuth"
# 2. Open the provided URL in your browser
# 3. Sign in with your Google account
# 4. Copy the authorization code
# 5. Paste it back into the terminal

After authentication, credentials are cached at ~/.gemini/antigravity-oauth-token and headless mode works without interaction.

Headless environments (SSH, CI/CD): If you can't run interactive mode, you can use GOOGLE_API_KEY environment variable:

bash

export GOOGLE_API_KEY="your-api-key-here"
agy -p "test prompt"

Usage Patterns

1. Headless Mode (Non-Interactive)

Best for: Scripts, automation, sub-agent delegation

bash

# Single prompt, text output (default)
agy -p "Explain what this function does"

# Capture output to variable
result=$(agy -p "Generate a Python function to calculate fibonacci")

# Redirect to file
agy -p "Write unit tests for main.py" > tests.md

2. Interactive Mode (TTY Required)

Best for: Development sessions, iterative coding

bash

# Launch interactive TUI
cd /path/to/project
agy

# Then type prompts directly in the TUI

3. Model Selection

bash

# Use specific model
agy -p "Refactor this code" --model gemini-2.5-pro

# List available models
agy models

Available models (as of 2026):

gemini-2.5-pro — Most capable, best for complex tasks
gemini-2.5-flash — Faster, good for simple tasks

4. File Context

bash

# Add directory to workspace context
agy -p "Review all Python files" --add-dir /path/to/project

# Multiple directories
agy -p "Compare implementations" --add-dir ./frontend --add-dir ./backend

5. Effort Level

bash

# Low effort — quick answers, less thorough
agy -p "Quick syntax check" --effort low

# Medium effort — balanced (default)
agy -p "Refactor this function" --effort medium

# High effort — deep analysis, comprehensive
agy -p "Full code review with security audit" --effort high

6. Output Formats

bash

# Text (default) — human-readable
agy -p "Explain this code"

# JSON — structured output with metadata
agy -p "List all functions in this file" --output-format json

# Stream JSON — real-time events (for monitoring)
agy -p "Analyze this project" --output-format stream-json

JSON output includes:

json

{
  "conversation_id": "uuid",
  "status": "SUCCESS",
  "response": "The agent's response text",
  "duration_seconds": 7.16,
  "num_turns": 1,
  "usage": {
    "input_tokens": 10415,
    "output_tokens": 657,
    "total_tokens": 11072
  }
}

7. Structured Output (JSON Schema)

bash

agy -p "Parse version v2.14.3 into major, minor, patch" \
  --output-format json \
  --json-schema '{"type":"object","properties":{"major":{"type":"integer"},"minor":{"type":"integer"},"patch":{"type":"integer"}},"required":["major","minor","patch"]}'

8. Permission Modes

bash

# Auto-approve all tool actions (USE WITH CAUTION)
agy -p "Refactor entire project" --dangerously-skip-permissions

# Plan mode — only suggest, don't execute
agy -p "How would you fix this bug?" --mode plan

# Accept edits — auto-apply suggested changes
agy -p "Fix all linting errors" --mode accept-edits

Examples

Example 1: Code Generation

bash

agy -p "Create a Python Flask API with endpoints for CRUD operations on a User model. Include proper error handling and validation."

Example 2: Code Review

bash

agy -p "Review the code in src/ directory. Check for:
1. Security vulnerabilities
2. Performance issues
3. Code style violations
4. Missing error handling
Provide specific file:line references for each issue." --effort high

Example 3: Refactoring

bash

cd /path/to/project
agy -p "Refactor the authentication module to use JWT tokens instead of session cookies. Update all related tests." --mode accept-edits

Example 4: Debugging

bash

agy -p "The function calculateTotal() in src/cart.js returns NaN when the cart is empty. Debug and fix this issue. Add proper null checks."

Example 5: Test Generation

bash

agy -p "Generate comprehensive unit tests for src/utils/validation.js. Cover edge cases, error conditions, and happy paths. Use Jest framework."

Example 6: Documentation

bash

agy -p "Generate JSDoc comments for all exported functions in src/api/*.js. Include @param, @returns, and @example tags."

Example 7: Multi-File Changes

bash

agy -p "Migrate the database layer from MongoDB to PostgreSQL. Update:
- All model files in src/models/
- All query functions in src/db/
- All tests in tests/
- Database config in src/config/
Keep the same API interface." --add-dir ./src --add-dir ./tests

Troubleshooting

Error: "authentication required"

Cause: Not authenticated or token expired

Solution:

bash

# Run interactive mode to authenticate
agy

# Or set API key
export GOOGLE_API_KEY="your-key"

Error: "bubbletea: error opening TTY"

Cause: Running interactive mode in non-TTY environment

Solution: Use headless mode instead:

bash

agy -p "your prompt"

Error: "command not found: agy"

Cause: Binary not in PATH

Solution:

bash

# Add to PATH
export PATH="$HOME/.local/bin:$PATH"

# Or use full path
/root/.local/bin/agy -p "prompt"

Error: "rate limit exceeded"

Cause: Hit Google AI Studio quota

Solution:

Wait for quota reset (usually hourly/daily)
Check limits at https://aistudio.google.com/
Consider upgrading to paid tier if needed

Headless mode hangs or times out

Cause: Complex task or permission prompt

Solution:

bash

# Increase timeout
agy -p "complex task" --print-timeout 10m

# Or auto-approve permissions (use with caution)
agy -p "task" --dangerously-skip-permissions

Billing & Limits

Antigravity CLI is FREE — no subscription cost.

Limits:

Uses your Google AI Studio account quota
Rate limits: RPM (requests per minute), TPM (tokens per minute), RPD (requests per day)
Check your limits: https://aistudio.google.com/
Free tier = rate limited, not paid

Important:

If your Google Cloud project has billing enabled, you'll be charged per token
For free usage, ensure billing is NOT enabled on the project
Multiple API keys in same project share the same quota

Integration with OpenClaw

Using in Sub-Agents

When spawning sub-agents for coding tasks, you can delegate to Antigravity CLI:

javascript

// In sub-agent task
sessions_spawn({
  task: `You are a coding agent. Use Antigravity CLI for implementation:
  
  Task: ${taskDescription}
  Files: ${relevantFiles}
  
  Run: agy -p "implement the task" --add-dir /path/to/project`
})

Combining with Claude Code

Both Antigravity CLI and Claude Code can be used:

bash

# Claude Code (default, subscription-based)
claude --permission-mode bypassPermissions --print "refactor code"

# Antigravity CLI (free, Google-based)
agy -p "refactor code"

When to use which:

Claude Code — Default choice, flat-rate subscription
Antigravity CLI — Alternative when Claude limit exhausted, or for Google ecosystem tasks

Configuration

Config file: ~/.gemini/antigravity-cli/settings.json

json

{
  "theme": "dark",
  "model": "gemini-2.5-pro",
  "effort": "medium"
}

View/edit config:

bash

# Interactive config editor
agy
# Then type: /config

# Or edit file directly
nano ~/.gemini/antigravity-cli/settings.json

Resources

Official Docs: https://antigravity.google/docs/
GitHub Repo: https://github.com/google-antigravity/antigravity-cli
Google AI Studio: https://aistudio.google.com/
Rate Limits: https://ai.google.dev/gemini-api/docs/rate-limits

Antigravity CLI

Install

Antigravity CLI Skill

When to Use

Prerequisites

Installation

Linux / macOS (Recommended)

Windows (PowerShell)

Windows (CMD)

Verify Installation

Authentication

Usage Patterns

1. Headless Mode (Non-Interactive)

2. Interactive Mode (TTY Required)

3. Model Selection

4. File Context

5. Effort Level

6. Output Formats

7. Structured Output (JSON Schema)

8. Permission Modes

Examples

Example 1: Code Generation

Example 2: Code Review

Example 3: Refactoring

Example 4: Debugging

Example 5: Test Generation

Example 6: Documentation

Example 7: Multi-File Changes

Troubleshooting

Error: "authentication required"

Error: "bubbletea: error opening TTY"

Error: "command not found: agy"

Error: "rate limit exceeded"

Headless mode hangs or times out

Billing & Limits

Integration with OpenClaw

Using in Sub-Agents

Combining with Claude Code

Configuration

Resources

License

Related skills