Skip to main content

​
What is Claude Code?

Claude Code is a terminal-based AI development assistant created by Anthropic. It enables developers to accelerate coding workflows by using natural language to generate code, troubleshoot issues, refactor existing implementations, and perform a wide range of engineering tasks directly from the command line.
AGCloud provides a fully Anthropic-compatible API endpoint. Claude Code will display model names like claude-sonnet-4-6, but the actual model running behind the scenes is your chosen model (e.g., GLM-5, Kimi-K2.5, etc.).
With AGCloud, you can use 160+ AI models from top providers (Claude, OpenAI, Gemini, GLM, Kimi, MiniMax, DeepSeek, and more) directly in Claude Code — all with a single API key and no code changes required.

​
Two Ways to Use Claude Code

​
Step 1: Install Claude Code CLI

Requires Node.js 20+
Install Claude Code via NPM package manager: 
npm install -g @anthropic-ai/claude-code
Check if the installation was successful. 
claude --version # display version, e.g.: 2.x.x (Claude Code)
After installation, you can start with Claude Code AI. 
cd your-project
claude
During your first run, Claude Code will prompt you to log in to your Anthropic account. You can skip this step.

​
Step 2: Configure AGCloud API

​
Option A: Quick Setup (Environment Variables)

The fastest way to get started. Set these environment variables before launching Claude Code:
To make this permanent, add the following lines to your shell profile (~/.bashrc, ~/.zshrc, or ~/.profile).
export ANTHROPIC_BASE_URL=https://aiandgpu.com
export CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS='1'
export ANTHROPIC_AUTH_TOKEN=<YOUR_API_KEY>
export ANTHROPIC_MODEL='claude-sonnet-4-6'
export ANTHROPIC_DEFAULT_HAIKU_MODEL='claude-haiku-4-5-20251001'
export ANTHROPIC_DEFAULT_SONNET_MODEL='claude-sonnet-4-6'
export ANTHROPIC_DEFAULT_OPUS_MODEL='claude-opus-4-6'
It is recommended to set CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS='1' for AWS Bedrock protocol compatibility, otherwise your requests may be slower. For more environment variables, refer to: Claude Code Env Vars
✅ Make configuration effective: 
# bash users
source ~/.bashrc

# zsh users
source ~/.zshrc
Edit the Claude Code settings file at ~/.claude/settings.json: 
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "CLAUDE_CODE_DISABLE_EXPERIMENTAL_BETAS": "1"
  }
}
If you manually edit settings.json, you must restart Claude Code for the changes to take effect. If running inside an IDE plugin, restart the IDE as well.

​
Step 3: Choose Your Model

By default, Claude Code maps its internal model tiers to specific models. With AGCloud, you can customize which model powers each tier.

​
Default Model Mapping

Claude Code TierEnvironment VariableDefault ModelDescription
Opus (Heavy tasks)ANTHROPIC_DEFAULT_OPUS_MODELclaude-opus-4-6Complex reasoning, architecture
Sonnet (Main)ANTHROPIC_DEFAULT_SONNET_MODELclaude-sonnet-4-6Primary coding model
Haiku (Light tasks)ANTHROPIC_DEFAULT_HAIKU_MODELclaude-haiku-4-5-20251001Quick completions, small edits

​
Model Configuration Examples

Add the following to ~/.claude/settings.json:

​
Use Claude Models

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "claude-opus-4-6",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "claude-sonnet-4-6",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "claude-haiku-4-5-20251001"
  }
}

​
Use GLM Models

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "glm-5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "glm-4.5-air"
  }
}

​
Use Kimi Models

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "kimi-k2.5"
  }
}

​
Use MiniMax Models

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "MiniMax-M2.5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "MiniMax-M2.5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "MiniMax-M2.5"
  }
}

​
Use DeepSeek Models

{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "deepseek-v3.2",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "deepseek-v3.2",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v3.2"
  }
}
The best approach is to assign different models to different tiers based on their strengths: 
{
  "env": {
    "ANTHROPIC_BASE_URL": "https://aiandgpu.com",
    "ANTHROPIC_AUTH_TOKEN": "<YOUR_API_KEY>",
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "glm-5",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "kimi-k2.5",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "deepseek-v3.2"
  }
}
Mix and match for best results! Use a powerful model like GLM-5 for complex tasks (Opus), a balanced model like Kimi-K2.5 for main coding (Sonnet), and a fast model like DeepSeek-V3.2 for quick completions (Haiku).

​
Step 4: Start with Claude Code

After configuration, start Claude Code in your project directory: 
cd your-project
claude
You can now use natural language to interact with your codebase: 
> Help me refactor the authentication module to use JWT tokens

> Find and fix the bug in the payment processing logic

> Write unit tests for the user service

> Explain the architecture of this project
Claude Code will show model names like claude-sonnet-4-20250514 in the UI, but the actual model running is your configured AGCloud model (e.g., GLM-5, Kimi-K2.5, etc.). This is expected behavior.

​
Available Models

Below are the recommended coding models provided by AGCloud:
Model NameUse ForSpeed
claude-opus-4-6Complex reasoning, full-stack development🌟🌟🌟🌟🌟
claude-sonnet-4-6Balanced coding, debugging🌟🌟🌟🌟🌟
claude-haiku-4-5-20251001Quick completions, lightweight tasks🌟🌟🌟🌟
AGCloud supports 160+ models. You can use any model from the model list as long as it supports the chat completions API.

​
FAQ

​
1. How to switch models?

You can switch models at any time by updating the environment variables in ~/.claude/settings.json and restarting Claude Code. Alternatively, set environment variables directly in your terminal before launching: 
ANTHROPIC_DEFAULT_SONNET_MODEL=glm-5 claude

​
2. Why does the UI still show “Claude” model names?

This behavior is intentional. Claude Code relies on Anthropic’s native model identifiers internally. AGCloud functions as a passthrough layer—it accepts requests using Claude model names and forwards them to the model you’ve configured. The underlying capabilities remain the same; only the displayed model name changes.

​
3. Manual config changes not taking effect?

If you edited ~/.claude/settings.json manually:
  1. Close Claude Code completely (not just the current session)
  2. If using the IDE plugin, restart the IDE
  3. Verify the JSON syntax is valid
  4. Re-launch Claude Code
# Verify your config
cat ~/.claude/settings.json | python3 -m json.tool
We recommend using the latest stable version of Claude Code for the best compatibility: 
npm update -g @anthropic-ai/claude-code

​
5. Connection issues?

If you encounter connection errors:
  1. Verify your API key is valid: visit AGCloud Dashboard
  2. Check that ANTHROPIC_BASE_URL is set to https://aiandgpu.com (no trailing slash)
  3. Test connectivity:
curl https://aiandgpu.com/v1/messages \
  -H "Authorization: Bearer <YOUR_API_KEY>" \
  -H "Content-Type: application/json" \
  -d '{"model":"glm-5","max_tokens":20,"messages":[{"role":"user","content":"Hi"}]}'