Gemini Server

A lightweight FastAPI server that exposes an Anthropic-style /v1/messages API backed by Google's Gemini web interface. This allows you to use Gemini models with any tool that supports the Anthropic API format (like Claude Code, Cursor, Continue.dev, etc.).

• Anthropic-Compatible API: Use Gemini with tools expecting Anthropic's API format
• Asynchronous & Fast: Built with FastAPI and Uvicorn for high performance
• No Official API Key Required: Uses your browser cookies instead of paid API credits
• Streaming Support: Full SSE (Server-Sent Events) streaming responses
• Multiple Models: Supports gemini-3-flash and gemini-3-pro
• Simple Configuration: Easy cookie-based authentication
• Lightweight: Minimal dependencies

• Python 3.8+
• A Google account with access to Gemini
• Chrome/Firefox browser to extract cookies

git clone https://github.com/IMApurbo/gemini_server
cd gemini-server

pip install -r requirements.txt

Or install directly:

pip install fastapi uvicorn gemini-webapi python-multipart

The server requires two cookies to authenticate with Gemini:

1. Open gemini.google.com in Chrome/Firefox

2. Log in to your Google account

3. Open Developer Tools:

   • Chrome: F12 or right-click → Inspect
   • Firefox: F12 or right-click → Inspect Element

4. Find the cookies:

   Chrome:

   • Go to the Application tab
   • Click Cookies → https://gemini.google.com
   • Look for __Secure-1PSID and __Secure-1PSIDTS

   Firefox:

   • Go to the Storage tab
   • Click Cookies → https://gemini.google.com
   • Look for __Secure-1PSID and __Secure-1PSIDTS

5. Copy the full value of each cookie (they're long strings)

On first run, the server will prompt you for the cookies:

python server.py

It will ask:

Gemini cookie '__Secure-1PSID' is not set in config.json.
Enter __Secure-1PSID: [paste your cookie here]
Gemini cookie '__Secure-1PSIDTS' is not set in config.json.
Enter __Secure-1PSIDTS: [paste your cookie here]

You can also create a config.json file manually:

{
    "secure_1psid": "YOUR_1PSID_VALUE",
    "secure_1psidts": "YOUR_1PSIDTS_VALUE"
}

You can customize the server using environment variables:

export SERVER_HOST="0.0.0.0"  # Default: 0.0.0.0
export SERVER_PORT="8765"     # Default: 8765

python server.py

You should see output like:

Starting Anthropic-style server...
INFO:     Started server process [12345]
INFO:     Waiting for application startup.
INFO:     Application startup complete.
INFO:     Uvicorn running on http://0.0.0.0:8765

curl http://localhost:8765/

Response:

{
  "status": "ok",
  "message": "Gemini-server Anthropic-style server is running."
}

If Claude Code supports custom API endpoints, add this to your configuration:

{
  "api": {
    "base_url": "http://localhost:8765",
    "api_key": "dummy-key"
  }
}

export ANTHROPIC_API_URL="http://localhost:8765/v1/messages"
export ANTHROPIC_API_KEY="dummy-key"

When running Claude Code, specify the endpoint:

claude --api-url http://localhost:8765 --api-key dummy-key

Health check endpoint.

List available models.

Response:

{
  "data": [
    {"id": "gemini-3-flash", "object": "model"},
    {"id": "gemini-3-pro", "object": "model"}
  ]
}

Send a message to Gemini.

Request:

{
  "model": "gemini-3-flash",
  "messages": [
    {"role": "user", "content": "Hello, how are you?"}
  ],
  "system": "You are a helpful assistant.",
  "stream": false
}

Response (non-streaming):

{
  "id": "msg_abc123def456",
  "type": "message",
  "role": "assistant",
  "model": "gemini-3-flash",
  "content": [
    {
      "type": "text",
      "text": "Hello! I'm doing well, thank you for asking..."
    }
  ],
  "stop_reason": "end_turn",
  "stop_sequence": null,
  "usage": {
    "input_tokens": 12,
    "output_tokens": 25
  }
}

curl -X POST http://localhost:8765/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [
      {"role": "user", "content": "What is the capital of France?"}
    ]
  }'

curl -X POST http://localhost:8765/v1/messages \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [
      {"role": "user", "content": "Tell me a short story"}
    ],
    "stream": true
  }'

import requests

response = requests.post(
    "http://localhost:8765/v1/messages",
    json={
        "model": "gemini-3-flash",
        "messages": [
            {"role": "user", "content": "Explain quantum computing in simple terms"}
        ]
    }
)

print(response.json()["content"][0]["text"])

If you see errors like:

RuntimeError: Could not find SNlM0e token - your cookies are likely invalid or expired

Your cookies have expired. Simply get fresh cookies from Gemini and update config.json.

• Stay logged into Google in your browser
• Visit gemini.google.com occasionally
• The free tier has usage limits, so use judiciously

If the server isn't accessible:

• Check if it's running: ps aux | grep server.py
• Check the port: netstat -tulpn | grep 8765
• Try changing the port: export SERVER_PORT=8766

Gemini has usage limits. If you hit them, you'll receive errors. Wait a few minutes and try again.

• Local Network: The server binds to 0.0.0.0 by default, making it accessible on your network
• No Authentication: The server doesn't implement API key validation
• Cookie Storage: Your cookies are stored in plaintext in config.json
• Recommendations:
  • Use a firewall to restrict access
  • Change HOST to 127.0.0.1 for local-only access
  • Don't expose the server to the public internet

┌─────────────┐     HTTP Request     ┌──────────────────┐
│   Client    │  ──────────────────> │   FastAPI App    │
│ (Claude Code│   /v1/messages       │  (server.py)     │
│  Cursor,    │                      │                  │
│  etc.)      │                      └────────┬─────────┘
└─────────────┘                               │
                                               ▼
                                      ┌──────────────────┐
                                      │  GeminiClient    │
                                      │ (gemini_webapi)  │
                                      └────────┬─────────┘
                                               │
                                               ▼
                                      ┌──────────────────┐
                                      │  Gemini Web      │
                                      │  (gemini.google) │
                                      └──────────────────┘

• fastapi: Web framework
• uvicorn: ASGI server
• gemini-webapi: Gemini web client
• python-multipart: Form data parsing

Contributions are welcome! Please feel free to submit issues and pull requests.

MIT License - feel free to use this for any purpose.

• Claude Code: Use Gemini as a replacement for Claude
• Cursor IDE: Use Gemini with Cursor's AI features
• Continue.dev: Use Gemini with Continue's coding assistant
• Custom Tools: Any tool that supports Anthropic's API format
• Testing: Test Gemini without paying for API credits
• Development: Develop applications with Gemini's capabilities

If you encounter issues:

1. Check the troubleshooting section
2. Verify your cookies are correct
3. Check the logs for error messages
4. Open an issue on GitHub

To completely remove the server:

# Remove the directory
rm -rf gemini-server

# Or if using pip
pip uninstall fastapi uvicorn gemini-webapi

• Gemini Official Site
• gemini-webapi Documentation
• Anthropic API Reference

Happy Coding! 🚀