antigravity-gateway

v2.1.0

Published

2 months ago

Universal AI gateway - access Claude & Gemini models via OpenAI or Anthropic-compatible API through Antigravity Cloud Code

0High
0Medium
0Low

johninnl

claude gemini anthropic openai antigravity proxy gateway ai llm cursor continue aider cline litellm

Antigravity Gateway

Universal AI Gateway - Access Claude and Gemini models through any OpenAI or Anthropic-compatible client, powered by Antigravity's Cloud Code.

banner

┌─────────────────────────────────────────────────────────────────────────┐
│                         YOUR AI TOOLS                                   │
│  Cursor • Cline • Continue • Aider • Claude Code • Gemini CLI • etc.   │
└─────────────────────────────────────────────────────────────────────────┘
                                    │
                                    ▼
                    ┌───────────────────────────────┐
                    │     Antigravity Gateway       │
                    │  OpenAI + Anthropic API       │
                    └───────────────────────────────┘
                                    │
                                    ▼
                    ┌───────────────────────────────┐
                    │   Antigravity Cloud Code      │
                    │   Claude • Gemini models      │
                    └───────────────────────────────┘

Features

Triple API Support: OpenAI Chat Completions (/v1/chat/completions), OpenAI Responses (/v1/responses), and Anthropic (/v1/messages) endpoints
Multiple Models: Access Claude Sonnet 4.5, Opus 4.5, and Gemini 3 Flash/Pro models
Extended Thinking: Full support for reasoning/thinking models
Multi-Account Load Balancing: Add multiple Google accounts for higher throughput
Universal Compatibility: Works with any OpenAI or Anthropic-compatible client

Quick Start

Docker (Recommended)

docker pull ghcr.io/johnneerdael/antigravity-gateway:latest

docker run -it -p 8080:8080 \
  -v $(pwd)/data:/root/.config/antigravity-gateway \
  ghcr.io/johnneerdael/antigravity-gateway:latest

Docker Compose

services:
  antigravity-gateway:
    image: ghcr.io/johnneerdael/antigravity-gateway:latest
    container_name: antigravity-gateway
    ports:
      - "8080:8080"
    volumes:
      - ./data:/root/.config/antigravity-gateway
    environment:
      - FALLBACK=true
    restart: unless-stopped

NPM

npx antigravity-gateway start

# Or install globally
npm install -g antigravity-gateway
agw start

Adding Google Accounts

The gateway requires Google accounts with Antigravity access. Add accounts via OAuth:

# Interactive (opens browser)
agw accounts add

# Headless servers (manual code entry)
agw accounts add --no-browser

# List configured accounts
agw accounts list

Client Configuration

The gateway exposes three API endpoints:

OpenAI Chat Completions: http://localhost:8080/v1/chat/completions
OpenAI Responses API: http://localhost:8080/v1/responses
Anthropic-compatible: http://localhost:8080/v1/messages

Available Models

| Model ID | Type | Context | Output | Thinking | |----------|------|---------|--------|----------| | claude-sonnet-4-5-thinking | Claude | 200K | 16K | ✓ | | claude-opus-4-5-thinking | Claude | 200K | 16K | ✓ | | claude-sonnet-4-5 | Claude | 200K | 8K | ✗ | | gemini-3-flash | Gemini | 1M | 16K | ✓ | | gemini-3-pro-high | Gemini | 1M | 16K | ✓ | | gemini-3-pro-low | Gemini | 1M | 16K | ✓ | | gemini-3-pro-image | Gemini | 1M | 16K | ✓ | | gemini-2.5-pro | Gemini | 1M | 16K | ✓ | | gemini-2.5-flash | Gemini | 1M | 16K | ✓ | | gemini-2.5-flash-thinking | Gemini | 1M | 16K | ✓ | | gemini-2.5-flash-lite | Gemini | 1M | 8K | ✗ |

Note: Model availability depends on your Antigravity account. Run curl http://localhost:8080/v1/models to see your available models.

AI Coding Tools Configuration

Cursor

Settings → Models → Add Model:

Provider: OpenAI Compatible
Base URL: http://localhost:8080/v1
API Key: any-value
Model: gemini-3-flash

Continue.dev

Edit ~/.continue/config.json:

{
  "models": [{
    "title": "Antigravity Gateway",
    "provider": "openai",
    "model": "gemini-3-flash",
    "apiBase": "http://localhost:8080/v1",
    "apiKey": "any-value"
  }]
}

Cline (VS Code)

Settings → Cline → API Provider:

Provider: OpenAI Compatible
Base URL: http://localhost:8080/v1
API Key: any-value
Model ID: claude-sonnet-4-5-thinking

Roo Code

Settings → Custom API:

API Endpoint: http://localhost:8080/v1/chat/completions
API Key: any-value
Model: gemini-3-flash

Kilo Code

Settings → Provider → OpenAI Compatible:

Base URL: http://localhost:8080/v1
API Key: any-value
Model: claude-sonnet-4-5-thinking

TRAE

Configure in settings:

Provider: OpenAI
Base URL: http://localhost:8080/v1
API Key: any-value

OpenCode

Edit ~/.opencode/config.json:

{
  "provider": {
    "type": "openai",
    "baseUrl": "http://localhost:8080/v1",
    "apiKey": "any-value",
    "model": "gemini-3-flash"
  }
}

Goose

Edit ~/.config/goose/config.yaml:

provider: openai
model: gemini-3-flash
api_base: http://localhost:8080/v1
api_key: any-value

Aider

aider --openai-api-base http://localhost:8080/v1 \
      --openai-api-key any-value \
      --model openai/gemini-3-flash

Or set environment variables:

export OPENAI_API_BASE=http://localhost:8080/v1
export OPENAI_API_KEY=any-value
aider --model openai/gemini-3-flash

Cody (Sourcegraph)

Settings → Cody → Enterprise:

Server Endpoint: http://localhost:8080
Access Token: any-value

Claude Code CLI

Create ~/.claude/settings.json:

{
  "env": {
    "ANTHROPIC_BASE_URL": "http://localhost:8080",
    "ANTHROPIC_API_KEY": "any-value"
  }
}

Then run: claude

Gemini CLI

export GEMINI_API_BASE=http://localhost:8080/v1
export GEMINI_API_KEY=any-value
gemini chat

Cherry Studio

Settings → API Configuration:

Provider: OpenAI Compatible
API URL: http://localhost:8080/v1
API Key: any-value
Model: gemini-3-flash

Factory Droid

Configure provider:

Type: OpenAI
Base URL: http://localhost:8080/v1
API Key: any-value

Crush

Settings → AI Provider:

Provider: OpenAI Compatible
Endpoint: http://localhost:8080/v1
Key: any-value

SDK Integration

OpenAI Python SDK

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8080/v1",
    api_key="any-value"
)

# Chat Completions API
response = client.chat.completions.create(
    model="gemini-3-flash",
    messages=[{"role": "user", "content": "Hello!"}]
)
print(response.choices[0].message.content)

# Responses API
response = client.responses.create(
    model="gemini-3-flash",
    input="Tell me a joke about programming."
)
print(response.output[0].content[0].text)

OpenAI JavaScript SDK

import OpenAI from 'openai';

const client = new OpenAI({
  baseURL: 'http://localhost:8080/v1',
  apiKey: 'any-value'
});

const response = await client.chat.completions.create({
  model: 'gemini-3-flash',
  messages: [{ role: 'user', content: 'Hello!' }]
});
console.log(response.choices[0].message.content);

Anthropic Python SDK

import anthropic

client = anthropic.Anthropic(
    base_url="http://localhost:8080",
    api_key="any-value"
)

message = client.messages.create(
    model="claude-sonnet-4-5-thinking",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Hello!"}]
)
print(message.content[0].text)

cURL

# OpenAI Chat Completions format
curl http://localhost:8080/v1/chat/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer any-value" \
  -d '{
    "model": "gemini-3-flash",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

# OpenAI Responses API format
curl http://localhost:8080/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer any-value" \
  -d '{
    "model": "gemini-3-flash",
    "input": "Tell me a three sentence bedtime story about a unicorn."
  }'

# OpenAI Responses API with streaming
curl http://localhost:8080/v1/responses \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer any-value" \
  -d '{
    "model": "gemini-3-flash",
    "input": "Write a haiku about coding.",
    "stream": true
  }'

# Anthropic format
curl http://localhost:8080/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: any-value" \
  -d '{
    "model": "claude-sonnet-4-5-thinking",
    "max_tokens": 1024,
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

LiteLLM Integration

Create litellm-config.yaml:

model_list:
  - model_name: claude-sonnet-4-5-thinking
    model_info:
      mode: chat
      supports_vision: true
      supports_function_calling: true
      max_input_tokens: 200000
      max_output_tokens: 16000
    litellm_params:
      model: openai/claude-sonnet-4-5-thinking
      api_base: http://antigravity-gateway:8080/v1
      api_key: "not-needed"
      stream: true

  - model_name: gemini-3-flash
    model_info:
      mode: chat
      supports_vision: true
      supports_function_calling: true
      max_input_tokens: 1048576
      max_output_tokens: 16384
    litellm_params:
      model: openai/gemini-3-flash
      api_base: http://antigravity-gateway:8080/v1
      api_key: "not-needed"
      stream: true

API Endpoints

| Endpoint | Method | Description | |----------|--------|-------------| | /v1/chat/completions | POST | OpenAI Chat Completions API | | /v1/responses | POST | OpenAI Responses API | | /v1/messages | POST | Anthropic Messages API | | /v1/models | GET | List available models | | /health | GET | Health check | | /account-limits | GET | Account quotas (add ?format=table) | | /refresh-token | POST | Force token refresh |

Multi-Account Load Balancing

Add multiple Google accounts for higher throughput and automatic failover:

agw accounts add  # Add first account
agw accounts add  # Add second account
agw accounts add  # Add third account

The gateway automatically:

Uses sticky account selection for prompt cache efficiency
Switches accounts when rate limited
Waits for short rate limits (≤2 min)
Falls back to alternate models when all accounts exhausted

Check account status:

curl "http://localhost:8080/account-limits?format=table"

Environment Variables

| Variable | Default | Description | |----------|---------|-------------| | PORT | 8080 | Server port | | DEBUG | false | Enable debug logging | | FALLBACK | false | Enable model fallback on quota exhaustion |

Migration from v1.x

If upgrading from antigravity-claude-proxy:

Config path changed: ~/.config/antigravity-proxy → ~/.config/antigravity-gateway
Package name changed: antigravity-claude-proxy → antigravity-gateway
Docker image changed: ghcr.io/johnneerdael/antigravity-claude-proxy → ghcr.io/johnneerdael/antigravity-gateway
New CLI alias: agw (short for antigravity-gateway)

Copy your existing accounts:

cp -r ~/.config/antigravity-proxy/* ~/.config/antigravity-gateway/

Troubleshooting

"No accounts available"

Add Google accounts: agw accounts add

Rate limited

Add more accounts for load balancing
Enable fallback mode: agw start --fallback
Wait for quota reset (shown in error message)

Connection refused

Ensure the gateway is running: agw start

Model not found

Use exact model IDs from /v1/models endpoint. Common models:

gemini-3-flash (recommended for speed)
claude-sonnet-4-5-thinking (recommended for quality)

License

MIT License - see LICENSE

Credits

Based on work from: