Direct access to Claude Code, CodeX, Gemini CLI from China, no VPN needed.
One API Key for all top-tier coding models.
Anthropic official CLI AI coding assistant with AIToken direct access from China.
Visit Node.js official site to download the LTS version (.msi), or use winget:
winget install OpenJS.NodeJS.LTS
Verify installation:
node --version npm --version
as administratorrun command line：
npm install -g @anthropic-ai/claude-code
In C:\Users\YourUsername\.claude\ directory, create settings.json：
{
"env": {
"ANTHROPIC_AUTH_TOKEN": "Your AIToken API Key",
"ANTHROPIC_BASE_URL": "https://aitoken.sbs/v1"
}
}
Open command line，cd to your project directory：
cd your-project-folder claude
Using Homebrew:
brew install node
npm install -g @anthropic-ai/claude-code
Create ~/.claude/settings.json：
{
"env": {
"ANTHROPIC_AUTH_TOKEN": "Your AIToken API Key",
"ANTHROPIC_BASE_URL": "https://aitoken.sbs/v1"
}
}
cd your-project-folder && claude
For Ubuntu/Debian, using NodeSource:
curl -fsSL https://deb.nodesource.com/setup_lts.x | sudo -E bash - sudo apt-get install -y nodejs
sudo npm install -g @anthropic-ai/claude-code
Create ~/.claude/settings.json，Same content as above。
cd your-project-folder && claude
OpenAI GPT-based code assistant, configured to use AIToken service.
as administratorRun：
npm install -g @openai/codex@latest
In C:\Users\YourUsername\.codex\ create two files：
config.toml
model_provider = "aitoken" model = "gpt-4o" model_reasoning_effort = "high" network_access = "enabled" [model_providers.aitoken] name = "AIToken" base_url = "https://aitoken.sbs/v1" wire_api = "responses" requires_openai_auth = true
auth.json
{
"OPENAI_API_KEY": "Your AIToken API Key"
}
cd your-project-folder codex
npm install -g @openai/codex@latest
In ~/.codex/ create config.toml and auth.json（Same as Windows）。
cd your-project-folder && codex
sudo npm install -g @openai/codex@latest
In ~/.codex/ create config.toml and auth.json（Same as Windows）。
cd your-project-folder && codex
Google official CLI tool with AIToken direct access from China.
npm install -g @google/gemini-cli
In C:\Users\YourUsername\.gemini\ create two files：
.env
GOOGLE_GEMINI_BASE_URL=https://aitoken.sbs/v1 GEMINI_API_KEY=Your AIToken API Key GEMINI_MODEL=gemini-2.5-pro
settings.json
{
"ide": { "enabled": true },
"security": {
"auth": { "selectedType": "gemini-api-key" }
}
}
cd your-project-folder gemini
npm install -g @google/gemini-cli
In ~/.gemini/ create .env and settings.json（Same as Windows）。
cd your-project-folder && gemini
sudo npm install -g @google/gemini-cli
In ~/.gemini/ create .env and settings.json（Same as Windows）。
cd your-project-folder && gemini
Universal setup for VSCode, Cursor, Trae and other editors.
Open VSCode → Extensions (Ctrl+Shift+X) → Search and install:
Works with Continue.dev, Cline, Tongyi Lingma, and other plugins supporting OpenAI API format:
// API Base URL https://aitoken.sbs/v1 // API Key Your AIToken API Key // Recommended Models deepseek-ai/DeepSeek-V3 ← Best value, top Chinese performance deepseek-ai/DeepSeek-R1 ← Deep reasoning, math/code deepseek-ai/DeepSeek-V4-Flash ← Ultra-fast response
DeepSeek V3 → deepseek-ai/DeepSeek-V3 ¥7/M Token DeepSeek R1 → deepseek-ai/DeepSeek-R1 ¥14/M Token DeepSeek V4 Flash → deepseek-ai/DeepSeek-V4-Flash ¥7/M Token DeepSeek V3.1 → deepseek-ai/DeepSeek-V3.1-Terminus ¥14/M Token Qwen 3.5 27B → Qwen/Qwen3.5-27B ¥7/M Token MiniMax M2.5 → MiniMaxAI/MiniMax-M2.5 ¥7/M Token
Sign up and get 1,000 free Tokens. Connect to top-tier AI coding tools in 3 minutes.
One Key for all models. Pay-as-you-go, zero management overhead.