Claude Code

​

Quick start

# 1. Install
npm install -g @anthropic-ai/claude-code

# 2. Configure (set environment variables)
export ANTHROPIC_BASE_URL=https://dashscope-intl.aliyuncs.com/apps/anthropic
export ANTHROPIC_API_KEY=sk-xxx
export ANTHROPIC_MODEL=qwen3.5-plus

# 3. Test (start Claude Code and type a message)
claude
# Then type: Write a hello world in Python

​

Configuration

​

Basic setup

• API endpoint: https://dashscope-intl.aliyuncs.com/apps/anthropic
• Authentication: API key required (get yours here)
• Model selection: Any Qwen model via the Anthropic-compatible API

​

Model configuration

1. During conversation: Use /model <model_name> for temporary changes

/model qwen3.5-plus

2. At startup: Use --model flag for single session

claude --model qwen3.5-plus

3. Environment variables: Set different models for task complexity

export ANTHROPIC_DEFAULT_OPUS_MODEL="qwen3.5-plus"     # Complex tasks
export ANTHROPIC_DEFAULT_SONNET_MODEL="qwen3.5-plus"   # Daily tasks
export ANTHROPIC_DEFAULT_HAIKU_MODEL="qwen3-coder-next" # Simple tasks

4. Settings file: Create settings.json in project root

{
  "env": {
    "ANTHROPIC_DEFAULT_OPUS_MODEL": "qwen3.5-plus",
    "ANTHROPIC_DEFAULT_SONNET_MODEL": "qwen3.5-plus",
    "ANTHROPIC_DEFAULT_HAIKU_MODEL": "qwen3-coder-next"
  }
}

​

Platform-specific setup

# Add to ~/.zshrc (zsh) or ~/.bash_profile (bash)
export ANTHROPIC_BASE_URL="https://dashscope-intl.aliyuncs.com/apps/anthropic"
export ANTHROPIC_API_KEY="YOUR_DASHSCOPE_API_KEY"
export ANTHROPIC_MODEL="qwen3.5-plus"

# Apply changes
source ~/.zshrc  # or source ~/.bash_profile

# PowerShell (persistent)
[Environment]::SetEnvironmentVariable("ANTHROPIC_API_KEY", "YOUR_KEY", [EnvironmentVariableTarget]::User)
[Environment]::SetEnvironmentVariable("ANTHROPIC_BASE_URL", "https://dashscope-intl.aliyuncs.com/apps/anthropic", [EnvironmentVariableTarget]::User)
[Environment]::SetEnvironmentVariable("ANTHROPIC_MODEL", "qwen3.5-plus", [EnvironmentVariableTarget]::User)

​

First launch setup

1. Open ~/.claude.json (macOS/Linux) or C:\Users\%USERNAME%\.claude.json (Windows)
2. Set hasCompletedOnboarding to true:

{
  "hasCompletedOnboarding": true
}

3. Restart Claude Code

​

Limitations

• Does not support: Direct file uploads (use file paths in conversation)
• Requires: Node.js v18.0 or later

​

Examples

​

Basic code generation

claude> Write a Python function to validate email addresses

claude> Refactor this code for better performance: [paste code]

​

Project-level assistance

# Start in your project directory
cd /path/to/project
claude

# Claude Code automatically scans and understands your codebase
claude> Add error handling to all API endpoints

claude> Create unit tests for the auth module

​

Model switching

# Use faster model for simple tasks
/model qwen3-coder-flash
claude> Format this JSON file

# Switch to powerful model for complex tasks
/model qwen3.5-plus
claude> Design a scalable microservices architecture for this app

​

Troubleshooting

Solution: Verify environment variables are set correctly. Run echo $ANTHROPIC_BASE_URL to check.

Solution:

• Start Claude Code from specific project directories
• Use /compact to summarize long conversations
• Use /clear before starting new tasks
• Be specific in your requests to avoid unnecessary file scans

Solution: Edit ~/.claude.json and set "hasCompletedOnboarding": true

Solution: Check model name spelling. Use models from the supported list.

​

Token optimization tips

• Project directory scanning
• Code file analysis
• Conversation history maintenance

1. Focused directories: Start from your specific project folder
2. Manual summarization: Type /compact at 95% context limit
3. Clear instructions: Specific requests minimize file scanning
4. Context resets: Type /clear between unrelated tasks
5. Smaller tasks: Break complex work into steps

​

Related resources

• Coding Plan: Setup with subscription →
• Models: Available models →
• API docs: Anthropic-compatible reference →
• Official docs: Claude Code documentation →