Gemini Antigravity CLI

Gemini Antigravity CLI

Skill by ara.so — Devtools Skills collection.

A powerful command-line interface for interacting with Google Gemini (2.5 Pro/Flash, 3.5 Flash) and Antigravity AI models. Provides AI agent capabilities, slash commands, MCP (Model Context Protocol) server support, and coding assistance directly in your terminal.

What It Does

Gemini Antigravity CLI enables developers to:

• Interact with Google Gemini AI models from the terminal
• Use advanced AI coding agent features for code generation and assistance
• Execute slash commands for quick workflows
• Integrate with MCP servers and plugins
• Maintain context-aware conversations
• Switch between different Gemini/Antigravity models on the fly

Installation

Global NPM Installation (Recommended)

npm install -g gemini-antigravity-cli

Binary Download

Download the pre-built executable from GitHub Releases:

# Download antigravity-cli.zip
wget https://github.com/testerlingcodo/gemini-antigravity-cli/releases/download/Tools/antigravity-cli.zip
unzip antigravity-cli.zip
chmod +x antigravity  # Unix-like systems

Requirements

• Node.js 16+ (for npm installation)
• Google API Key with Gemini access
• Internet connection for API calls

Configuration

Setting Up API Key

Set your Google Gemini API key as an environment variable:

# Linux/macOS
export GOOGLE_API_KEY="your_api_key_here"

# Windows PowerShell
$env:GOOGLE_API_KEY="your_api_key_here"

# Windows CMD
set GOOGLE_API_KEY=your_api_key_here

Add to your shell profile for persistence:

# ~/.bashrc or ~/.zshrc
export GOOGLE_API_KEY="your_api_key_here"

Configuration File

The CLI typically stores configuration in:

• Linux/macOS: ~/.config/gemini-antigravity-cli/config.json
• Windows: %APPDATA%\gemini-antigravity-cli\config.json

Example configuration:

{
  "model": "gemini-2.5-flash",
  "temperature": 0.7,
  "maxTokens": 8192,
  "systemPrompt": "You are a helpful coding assistant.",
  "mcpServers": [],
  "plugins": []
}

Key Commands

Starting the CLI

# Start with default settings
antigravity

# Or use the alias
gemini

Slash Commands

Once inside the CLI, use these slash commands:

/help

Display all available commands:

/help

/model

Switch between AI models:

/model gemini-2.5-pro
/model gemini-2.5-flash
/model gemini-3.5-flash
/model antigravity

/clear

Clear conversation history:

/clear

/context

Add file or directory context:

/context ./src/index.ts
/context ./src/

/save

Save conversation to file:

/save conversation.md
/save output.txt

/load

Load previous conversation:

/load conversation.md

/config

View or modify configuration:

/config
/config model gemini-2.5-flash
/config temperature 0.9

/exit

Exit the CLI:

/exit

Usage Patterns

Basic Code Assistance

Start the CLI and ask coding questions directly:

// User input in CLI:
Create a TypeScript function to fetch data from an API with retry logic

// The AI will generate:
async function fetchWithRetry<T>(
  url: string,
  maxRetries: number = 3,
  delayMs: number = 1000
): Promise<T> {
  let lastError: Error;
  
  for (let attempt = 0; attempt < maxRetries; attempt++) {
    try {
      const response = await fetch(url);
      if (!response.ok) {
        throw new Error(`HTTP ${response.status}: ${response.statusText}`);
      }
      return await response.json() as T;
    } catch (error) {
      lastError = error as Error;
      if (attempt < maxRetries - 1) {
        await new Promise(resolve => setTimeout(resolve, delayMs * (attempt + 1)));
      }
    }
  }
  
  throw new Error(`Failed after ${maxRetries} attempts: ${lastError!.message}`);
}

Context-Aware Coding

Load project files for context-aware assistance:

# In the CLI:
/context ./src/api/client.ts

# Then ask:
Refactor this API client to use async/await instead of promises

Multi-File Analysis

/context ./src/
/context ./tests/

# Then:
Find all the test cases that are missing error handling

Code Review Workflow

/context ./pull-request.diff

# Ask:
Review this code for potential bugs, security issues, and best practices

Integration Examples

Using in Scripts

You can pipe input to the CLI for automated workflows:

# Send a single prompt
echo "Generate a README.md template for a TypeScript project" | antigravity

# Process file content
cat ./src/index.ts | antigravity "Explain what this code does"

Batch Processing

# Create a script for batch processing
#!/bin/bash
for file in ./src/*.ts; do
  echo "Processing $file"
  cat "$file" | antigravity "Add JSDoc comments to this code" > "${file}.documented"
done

CI/CD Integration

# .github/workflows/code-review.yml
name: AI Code Review
on: [pull_request]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      - name: Install CLI
        run: npm install -g gemini-antigravity-cli
      - name: Run Review
        env:
          GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
        run: |
          git diff origin/main...HEAD | antigravity "Review this code change"

MCP Server Integration

Connecting to MCP Servers

Configure MCP servers in your config file:

{
  "mcpServers": [
    {
      "name": "filesystem",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-filesystem", "/path/to/workspace"]
    },
    {
      "name": "github",
      "command": "npx",
      "args": ["-y", "@modelcontextprotocol/server-github"],
      "env": {
        "GITHUB_TOKEN": "${GITHUB_TOKEN}"
      }
    }
  ]
}

Using MCP Tools

# Once MCP servers are configured, use them in prompts:
Use the filesystem server to read all package.json files in the workspace

# Or:
Search GitHub for similar implementations using the github server

Common Patterns

Project Setup Assistant

// User: Set up a new Express.js API project with TypeScript

// CLI generates project structure:
mkdir my-api && cd my-api
npm init -y
npm install express
npm install -D typescript @types/express @types/node ts-node

// Creates tsconfig.json:
{
  "compilerOptions": {
    "target": "ES2020",
    "module": "commonjs",
    "outDir": "./dist",
    "strict": true,
    "esModuleInterop": true
  }
}

// Creates src/index.ts:
import express from 'express';

const app = express();
const port = process.env.PORT || 3000;

app.use(express.json());

app.get('/health', (req, res) => {
  res.json({ status: 'ok' });
});

app.listen(port, () => {
  console.log(`Server running on port ${port}`);
});

Debugging Assistant

/context ./src/buggy-code.ts

# Ask:
This code throws "Cannot read property 'length' of undefined". Find and fix the bug.

Documentation Generator

/context ./src/utils/

# Generate docs:
Create comprehensive API documentation for all exported functions in markdown format

Test Generation

/context ./src/user-service.ts

# Generate tests:
Create Jest unit tests for this service with 100% code coverage

Troubleshooting

API Key Issues

Problem: "Invalid API key" error

Solution:

# Verify API key is set
echo $GOOGLE_API_KEY  # Linux/macOS
echo %GOOGLE_API_KEY%  # Windows

# Re-export with correct key
export GOOGLE_API_KEY="your_valid_key"

# Test with curl
curl "https://generativelanguage.googleapis.com/v1/models?key=$GOOGLE_API_KEY"

Connection Errors

Problem: Network timeout or connection refused

Solution:

# Check internet connectivity
ping google.com

# Verify proxy settings if behind corporate firewall
export HTTP_PROXY="http://proxy.company.com:8080"
export HTTPS_PROXY="http://proxy.company.com:8080"

# Test API endpoint
curl -I https://generativelanguage.googleapis.com

Model Not Available

Problem: "Model not found" error

Solution:

# List available models
/model

# Use a supported model:
/model gemini-2.5-flash

# Check API access for specific models in Google AI Studio

Rate Limiting

Problem: "Rate limit exceeded" or 429 errors

Solution:

• Wait before retrying (rate limits reset after a period)
• Use gemini-2.5-flash for higher rate limits
• Implement exponential backoff in scripts:

#!/bin/bash
attempt=0
max_attempts=5

while [ $attempt -lt $max_attempts ]; do
  if echo "$prompt" | antigravity; then
    break
  fi
  attempt=$((attempt + 1))
  sleep $((2 ** attempt))
done

Context Size Errors

Problem: "Context length exceeded" error

Solution:

# Reduce context size
/clear  # Clear previous conversation

# Use smaller file selections
/context ./src/specific-file.ts  # Instead of entire directory

# Summarize large files first
cat large-file.ts | head -n 100 | antigravity "Summarize this code"

Installation Issues

Problem: npm install fails

Solution:

# Clear npm cache
npm cache clean --force

# Try with explicit registry
npm install -g gemini-antigravity-cli --registry=https://registry.npmjs.org

# Or use npx without global install
npx gemini-antigravity-cli

Binary Execution Issues

Problem: Permission denied on Unix systems

Solution:

chmod +x antigravity
./antigravity

Problem: Windows SmartScreen warning

Solution:

• Click "More info" → "Run anyway"
• Or build from source if concerned about security

Advanced Configuration

Custom System Prompts

Modify behavior with custom system prompts:

{
  "systemPrompt": "You are an expert TypeScript developer specializing in React and Node.js. Provide concise, production-ready code with error handling. Always use modern ES6+ syntax."
}

Temperature and Token Settings

Control response creativity and length:

{
  "temperature": 0.2,  // Lower = more deterministic (0.0-1.0)
  "maxTokens": 4096,   // Maximum response length
  "topP": 0.95,        // Nucleus sampling threshold
  "topK": 40           // Top-K sampling parameter
}

Plugin Configuration

Enable community plugins (if supported):

{
  "plugins": [
    "git-integration",
    "code-formatter",
    "security-scanner"
  ]
}

Best Practices

1. Always set API key as environment variable — never hardcode in scripts
2. Use /clear between unrelated tasks — prevents context confusion
3. Provide specific context with /context — improves accuracy
4. Start with gemini-2.5-flash — faster and cheaper for most tasks
5. Save important conversations — use /save for future reference
6. Use lower temperature (0.2-0.4) for code generation — more consistent output
7. Higher temperature (0.7-0.9) for creative tasks — like naming or brainstorming