Claude Code Setup & Best Practices

Claude Code is Anthropic's official agentic coding tool that runs directly in your terminal. It can understand your codebase, edit files, run commands, and help you build complex projects—all through natural conversation. By connecting Claude Code to UniterRouter, you gain access to multiple AI providers, unified billing, and significant cost savings.

Why Use UniterRouter with Claude Code?

Prerequisites

• Terminal access (macOS, Linux, or WSL on Windows)
• Node.js 18+ installed
• An UniterRouter API key (get one here)

Quick Setup

npm install -g @anthropic-ai/claude-code

brew install claude-code

export ANTHROPIC_BASE_URL="https://api.uniterouter.ai"
export ANTHROPIC_API_KEY="your-uniterouter-api-key"

$env:ANTHROPIC_BASE_URL = "https://api.uniterouter.ai"
$env:ANTHROPIC_API_KEY = "your-uniterouter-api-key"

# Add to your shell config file
echo 'export ANTHROPIC_BASE_URL="https://api.uniterouter.ai"' >> ~/.zshrc
echo 'export ANTHROPIC_API_KEY="your-uniterouter-api-key"' >> ~/.zshrc
source ~/.zshrc

claude-code --version
claude-code "Hello, can you see my current directory?"

Best Practices for Claude Code

1. Optimize Token Usage

Claude Code can consume significant tokens for large codebases. Here's how to optimize:

# Use .claudeignore to exclude unnecessary files
echo "node_modules/
dist/
*.log
.git/
coverage/
__pycache__/" > .claudeignore

Recommended .claudeignore patterns:

• node_modules/, vendor/, venv/ - Dependencies
• dist/, build/, out/ - Build outputs
• *.min.js, *.min.css - Minified files
• .git/, .svn/ - Version control
• *.lock, package-lock.json - Lock files (optional)

2. Use Focused Context

Instead of letting Claude Code scan your entire project, provide specific context:

# Good: Specific file context
claude-code "Review the authentication logic in src/auth/login.ts"

# Avoid: Vague requests that scan everything
claude-code "Review my code"

3. Enable Response Caching

UniterRouter automatically caches identical requests. Structure your prompts consistently to maximize cache hits:

# These will cache efficiently
claude-code "Explain this function: $(cat src/utils/helper.ts)"
claude-code "Explain this function: $(cat src/utils/helper.ts)"  # Cache hit!

4. Leverage Model Routing

With UniterRouter, you can switch models based on task complexity:

# For simple tasks, use faster/cheaper models
export ANTHROPIC_MODEL="claude-3-5-haiku-latest"
claude-code "Format this JSON file"

# For complex reasoning, use powerful models
export ANTHROPIC_MODEL="claude-sonnet-4-20250514"
claude-code "Refactor this authentication system for better security"

Advanced Configuration

Custom API Settings

Create a configuration file for persistent settings:

# ~/.claude-code/config.json
{
  "apiBaseUrl": "https://api.uniterouter.ai",
  "defaultModel": "claude-sonnet-4-20250514",
  "maxTokens": 8192,
  "temperature": 0.7
}

Project-Specific Configuration

Add a .claude-code.json to your project root:

{
  "model": "claude-sonnet-4-20250514",
  "systemPrompt": "You are a senior developer working on a Next.js e-commerce platform. Follow TypeScript best practices and ensure all code is type-safe.",
  "ignorePatterns": [
    "node_modules",
    "*.test.ts",
    ".env*"
  ]
}

Using Different Models for Different Tasks

UniterRouter allows you to access any model through the same endpoint:

Cost Optimization Tips

1. Monitor Usage

Check your spending in real-time:

# View current session costs (if supported)
claude-code --stats

# Or check the UniterRouter dashboard
open https://www.uniterouter.ai/dashboard

2. Set Spending Limits

Configure API key limits in UniterRouter to prevent unexpected costs:

1. Go to Dashboard → API Keys
2. Edit your API key
3. Set daily/monthly spending limits
4. Enable alerts for threshold notifications

3. Use Caching Strategically

UniterRouter's caching can reduce costs by up to 90% for repeated queries:

• Same code explanations → Cached
• Repeated formatting requests → Cached
• Common refactoring patterns → Cached

Troubleshooting

Connection Issues

# Test API connectivity
curl -X POST https://api.uniterouter.ai/v1/messages \
  -H "Content-Type: application/json" \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -d '{"model": "claude-sonnet-4-20250514", "max_tokens": 100, "messages": [{"role": "user", "content": "Hi"}]}'

Model Not Found

If you get a model not found error, check available models:

curl https://api.uniterouter.ai/v1/models \
  -H "Authorization: Bearer $ANTHROPIC_API_KEY"

Rate Limiting

UniterRouter handles rate limits automatically with intelligent retry logic. If you hit limits frequently, consider:

1. Enabling caching (automatic)
2. Using batch operations
3. Upgrading your plan for higher limits

Security Best Practices

Protect Your API Keys

# Never commit API keys
echo "ANTHROPIC_API_KEY=*" >> .gitignore

# Use environment variables or secret managers
# macOS Keychain
security add-generic-password -a "$USER" -s "uniterouter-api-key" -w "your-key"

# Retrieve it
export ANTHROPIC_API_KEY=$(security find-generic-password -a "$USER" -s "uniterouter-api-key" -w)

Use IAM Rules

Configure fine-grained access control in UniterRouter:

• Restrict models per API key
• Set rate limits
• Enable IP allowlists
• Configure spending caps

See API Keys & IAM Rules for detailed configuration.

Summary

Using Claude Code with UniterRouter gives you:

• Multi-model flexibility - Access Claude, GPT-4, Gemini, and more
• Cost savings - Up to 50% reduction with caching
• Unified billing - One dashboard for all AI spending
• Better reliability - Automatic failover between providers
• Enhanced security - IAM rules and spending controls

Start coding smarter today by getting your UniterRouter API key.