🚀 Claude Code Gemini Proxy

Use Google Gemini models in Claude Code via Vertex AI

This proxy server transparently translates Anthropic API calls to Google Vertex AI, allowing you to use powerful Gemini models (Flash, Pro, and Experimental) directly in Claude Code.

✨ Features

• 🔄 Transparent API Translation - Claude Code thinks it's talking to Claude, but it's actually using Gemini
• ⚡ Multiple Models - Choose between Gemini 2.0 Flash, Pro, or Experimental
• 🌊 Streaming Support - Real-time streaming responses just like Claude
• 🔐 Flexible Auth - Supports Service Account, gcloud CLI, or Application Default Credentials
• 🎨 Interactive Setup - Simple CLI wizard to configure everything
• 🌍 Multi-Region - Deploy in any GCP region

📦 Installation

npm install -g claude-code-gemini-proxy

Or install locally:

git clone https://github.com/stormdaemon/claude-code-gemini-proxy.git
cd claude-code-gemini-proxy
npm install
npm run build
npm link

🚀 Quick Start

1. Setup (One-time)

Run the interactive setup wizard:

gemini-proxy setup

You'll be prompted to:

• ✅ Choose a Gemini model (Flash / Pro / Experimental)
• ✅ Select authentication method
• ✅ Enter your GCP Project ID
• ✅ Choose a region
• ✅ Test the connection

2. Start the Proxy

gemini-proxy start

The proxy will start on http://localhost:8080 (or your configured port).

3. Configure Claude Code

In your terminal, set these environment variables:

export ANTHROPIC_BASE_URL=http://localhost:8080
export ANTHROPIC_API_KEY=dummy-key

Add them to your ~/.bashrc or ~/.zshrc to make them permanent:

echo 'export ANTHROPIC_BASE_URL=http://localhost:8080' >> ~/.zshrc
echo 'export ANTHROPIC_API_KEY=dummy-key' >> ~/.zshrc
source ~/.zshrc

4. Use Claude Code!

Now just use Claude Code normally - it will use Gemini instead:

claude "Explain this function"

🎮 CLI Commands

| Command | Description | |---------|-------------| | gemini-proxy setup | Interactive setup wizard | | gemini-proxy start | Start the proxy server | | gemini-proxy status | Show current configuration | | gemini-proxy test | Test connection to Vertex AI | | gemini-proxy reset | Clear all configuration |

🔐 Authentication

Option 1: Application Default Credentials (Easiest)

gcloud auth application-default login

Then choose "Application Default Credentials" in the setup.

Option 2: Service Account

1. Create a service account in GCP Console
2. Download the JSON key file
3. Choose "Service Account JSON file" in the setup
4. Provide the path to your key file

Option 3: gcloud CLI

If you're already authenticated with gcloud, just choose "gcloud" in the setup.

🤖 Available Models

| Model | Description | Context Window | Best For | |-------|-------------|----------------|----------| | gemini-2.5-flash | Best price-performance | 1M tokens | Large-scale processing, low-latency tasks | | gemini-2.5-pro | Advanced thinking model | 2M tokens | Complex reasoning, large datasets | | gemini-3-pro-preview | Most intelligent (preview) | 2M tokens | Multimodal understanding, agentic tasks |

🛠️ Configuration

Configuration is stored in ~/.config/gemini-proxy/config.json

Example configuration:

{
  "projectId": "my-gcp-project",
  "location": "us-central1",
  "model": "gemini-2.5-flash",
  "authMethod": "adc",
  "port": 8080
}

🌍 Supported Regions

• us-central1 (Iowa)
• us-east4 (Virginia)
• europe-west1 (Belgium)
• europe-west4 (Netherlands)
• asia-southeast1 (Singapore)

🔧 Advanced Usage

Running as a Background Service

Linux/Mac (systemd)

Create /etc/systemd/system/gemini-proxy.service:

[Unit]
Description=Gemini Proxy Server
After=network.target

[Service]
Type=simple
User=yourusername
ExecStart=/usr/local/bin/gemini-proxy start
Restart=always

[Install]
WantedBy=multi-user.target

Enable and start:

sudo systemctl enable gemini-proxy
sudo systemctl start gemini-proxy

Mac (launchd)

Create ~/Library/LaunchAgents/com.gemini-proxy.plist:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
    <key>Label</key>
    <string>com.gemini-proxy</string>
    <key>ProgramArguments</key>
    <array>
        <string>/usr/local/bin/gemini-proxy</string>
        <string>start</string>
    </array>
    <key>RunAtLoad</key>
    <true/>
    <key>KeepAlive</key>
    <true/>
</dict>
</plist>

Load:

launchctl load ~/Library/LaunchAgents/com.gemini-proxy.plist

Custom Port

# During setup, choose a different port
# Or edit ~/.config/gemini-proxy/config.json directly

Change Model

gemini-proxy reset
gemini-proxy setup
# Choose a different model

🐛 Troubleshooting

Authentication Errors

# Test your auth
gemini-proxy test

# Re-authenticate with gcloud
gcloud auth application-default login

# Verify your service account has permissions
gcloud projects get-iam-policy YOUR_PROJECT_ID

Port Already in Use

# Check what's using the port
lsof -i :8080

# Or change the port during setup
gemini-proxy reset
gemini-proxy setup

Claude Code Not Using Proxy

Make sure environment variables are set:

echo $ANTHROPIC_BASE_URL  # Should be http://localhost:8080
echo $ANTHROPIC_API_KEY   # Should be dummy-key or anything

# Re-export them
export ANTHROPIC_BASE_URL=http://localhost:8080
export ANTHROPIC_API_KEY=dummy-key

Vertex AI API Not Enabled

# Enable the Vertex AI API
gcloud services enable aiplatform.googleapis.com --project=YOUR_PROJECT_ID

📚 How It Works

┌─────────────────┐
│   Claude Code   │  Thinks it's calling Anthropic API
└────────┬────────┘
         │ POST /v1/messages
         ↓
┌─────────────────────┐
│  Gemini Proxy       │  Translates request format
│  (localhost:8080)   │  Routes to Vertex AI
└────────┬────────────┘
         │ POST /v1/projects/.../models/gemini-...:generateContent
         ↓
┌─────────────────────┐
│  Vertex AI Gemini   │  Processes with Gemini
│  (GCP)              │  Returns response
└─────────────────────┘

The proxy:

1. Accepts Anthropic API format requests
2. Translates them to Gemini API format
3. Calls Vertex AI
4. Translates responses back to Anthropic format
5. Returns to Claude Code

🔒 Security

• Service account keys are stored locally only
• No data is logged or transmitted except to Vertex AI
• All authentication uses official Google libraries
• HTTPS is used for all Vertex AI communication

🤝 Contributing

Contributions welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Test thoroughly
5. Submit a pull request

📝 License

MIT License - see LICENSE file for details

🙏 Credits

Built with:

• Fastify - Web server
• Google Cloud AI Platform - Gemini API
• Inquirer - Interactive CLI
• Chalk - Terminal styling

🆘 Support

• 🐛 Report a bug
• 💡 Request a feature
• 💬 Ask a question

⭐ Star History

If this project helps you, consider giving it a star! ⭐

Made with ❤️ by Droid AI

🆕 Changelog (v2.0.0 - December 2025)

Major Updates

• ✅ Node.js 22+ requirement (ESM modules)
• ✅ TypeScript 5.7 with latest features
• ✅ Fastify 5.2 latest web server
• ✅ Chalk 5.6 & Inquirer 13 (ESM)
• ✅ Google Cloud AI Platform 5.13 latest SDK
• ✅ Updated to current Gemini models (2.5 Flash, 2.5 Pro, 3 Pro Preview)

Breaking Changes

• Requires Node.js 22+ (previously 18+)
• All imports now use ESM format
• Some dependency API changes (handled internally)