Claude Code

Claude Code is Anthropic's AI coding assistant, providing conversational code generation, explanation, and editing capabilities directly in your terminal.

Overview

This guide covers using Claude Code with a direct Anthropic API subscription. For using Claude Code with Databricks-hosted models, see Claude Code Router (CCR) →.

Authentication

Claude Code offers multiple authentication options. Choose the method that best fits your needs.

Method 1: Claude Pro or Max Plan (Recommended)

# Run claude and login with your Claude credentials
claude

When prompted during setup, log in with your Claude Pro (20 USD/month) or Claude Max (100 USD/month) account. This provides: - Unified subscription for both Claude Code and web interface - No additional API billing - Automatic usage tracking

Method 2: Claude Console (Pay-as-you-go)

# Run claude and complete OAuth process
claude

Connect through the Claude Console at console.anthropic.com with active billing. A "Claude Code" workspace is automatically created for usage tracking and cost management.

Method 3: API Key

# Set your Anthropic API key
export ANTHROPIC_API_KEY="sk-ant-..."

# Add to ~/.bashrc for persistence
echo 'export ANTHROPIC_API_KEY="sk-ant-..."' >> ~/.bashrc

# Run claude (will use API key automatically)
claude

Get your API key from the Anthropic Console.

Note: If ANTHROPIC_API_KEY is set, Claude Code will use it instead of your Claude subscription.

Switching Authentication

To change authentication methods:

# Logout
claude logout

# Login with preferred method
claude login

Usage

# Run Claude Code
claude

Note: The cc command is reserved for Claude Code when using CCR (Databricks-hosted). When using direct Anthropic API, use claude.

Features with Anthropic API

When using direct Anthropic API, you get full access to:

• ✅ Extended thinking modes - Use ultrathink, think, and other thinking keywords for deeper reasoning
• ✅ All Claude models - Access to Sonnet, Opus, Haiku variants
• ✅ Latest features - Immediate access to new Anthropic features
• ✅ Full API capabilities - No limitations from proxy transformations
• ✅ 200K+ context window - Large context for complex codebases

Model Selection

# Use specific model
export CLAUDE_MODEL="claude-sonnet-4-20250514"

# Add to ~/.bashrc
echo 'export CLAUDE_MODEL="claude-sonnet-4-20250514"' >> ~/.bashrc

Available Models

• claude-sonnet-4-20250514 - Latest Sonnet 4 (balanced performance)
• claude-opus-4-20250514 - Most capable (if available)
• claude-haiku-4-20250305 - Fastest and most efficient

Common Usage Examples

Interactive Mode

$ claude
Claude Code v1.0
Type 'exit' to quit

> explain this code
[paste code]

> write a function to calculate fibonacci
[Claude generates code]

> edit the function to use memoization
[Claude modifies code]

Command Line Mode

# One-off command
claude "write a function to reverse a string"

# With file context
claude "explain index.py" < index.py

# Save output to file
claude "write a REST API client" > api_client.py

Multi-turn Conversations

Claude Code maintains context across multiple turns:

> write a class to manage users

[Claude generates User class]

> add a method to validate email addresses

[Claude adds method to existing class]

> write unit tests for the User class

[Claude generates test suite]

Advanced Usage

With File Context

# Provide file as input
claude "add docstrings" < my_script.py

# Or reference files in prompt
claude "explain the main function in app.py"

Custom Instructions

# Set preferences
export CLAUDE_CODE_INSTRUCTIONS="Always include type hints and docstrings"

claude "write a function to sort a list"

Using Extended Thinking

With Anthropic API, you can use extended thinking modes:

# Use ultra thinking for complex problems
claude "ultrathink: design a distributed caching system with consistency guarantees"

# Use standard thinking for moderate complexity
claude "think: optimize this database query for performance"

Note: Thinking modes are not available when using Databricks-hosted Claude via CCR.

Best Practices

1. Be Specific

# Good
> write a Python function that reads a CSV file, filters rows where column 'status' is 'active', and returns a list of dictionaries

# Less optimal
> read csv file

2. Iterate Incrementally

> write a basic User class

[review output]

> add email validation to the User class

[review output]

> add password hashing

3. Provide Context

> I'm working on a Flask app. Write a route handler for user registration.

# Better than:
> write a function for user registration

4. Use Thinking Modes for Complex Tasks

# For architectural decisions
> ultrathink: should I use microservices or monolith for this e-commerce platform?

# For optimization problems
> think: how can I reduce the time complexity of this algorithm?

Custom Commands

Claude Code supports custom commands that you can define in your project. Commands are stored in ./.claude/commands/ directory and allow you to create reusable prompts and workflows.

Setting Up Commands

Create a commands directory in your project:

mkdir -p .claude/commands

Creating a Command

Commands are simple text files with prompts. Create a file in .claude/commands/:

Example: .claude/commands/review.md

Review this code for:
- Bugs and potential errors
- Performance issues
- Security vulnerabilities
- Code style and best practices
- Maintainability improvements

Provide specific, actionable feedback.

Example: .claude/commands/test.md

Generate comprehensive pytest unit tests for this code.

Include:
- Happy path tests
- Edge cases
- Error handling
- Mock external dependencies
- Docstrings for each test

Example: .claude/commands/explain.md

Explain this code in detail:
- What it does
- How it works
- Key algorithms or patterns used
- Any potential gotchas or limitations

Using Commands

Once created, use commands in the Claude Code interactive interface with slash commands:

$ claude
Claude Code v1.0
Type 'exit' to quit

> /review
[paste your code here]

Claude will now review the code based on the review.md command template.

> /test
[paste your code here]

Claude will generate pytest tests based on the test.md command template.

> /explain
[paste your code here]

Claude will explain the code based on the explain.md command template.

Command Best Practices

1. Be specific: Include clear instructions in your commands
2. Use markdown: Format commands with markdown for better readability
3. Add context: Include relevant details like coding standards or project conventions
4. Organize by task: Create separate commands for different workflows (review, test, refactor, etc.)

Example Command Library

Create a library of commands for common tasks:

.claude/commands/
├── review.md          # Code review
├── test.md            # Generate tests
├── refactor.md        # Refactoring suggestions
├── optimize.md        # Performance optimization
├── document.md        # Add documentation
└── security.md        # Security audit

Configuration

API Key Management

Store your API key securely:

# In ~/.bashrc or ~/.zshrc
export ANTHROPIC_API_KEY="sk-ant-..."

# Or use a secrets manager
export ANTHROPIC_API_KEY=$(cat ~/.secrets/anthropic_key)

Custom Settings

# Set default model
export CLAUDE_MODEL="claude-sonnet-4-20250514"

# Set custom instructions
export CLAUDE_CODE_INSTRUCTIONS="Use type hints, write docstrings, follow PEP 8"

# Set timeout (in seconds)
export CLAUDE_TIMEOUT=120

Troubleshooting

API Key Issues

# Check if API key is set
echo $ANTHROPIC_API_KEY

# Test API key
curl https://api.anthropic.com/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "anthropic-version: 2023-06-01" \
  -H "content-type: application/json" \
  -d '{"model":"claude-sonnet-4-20250514","max_tokens":1024,"messages":[{"role":"user","content":"Hello"}]}'

Connection Issues

# Check network connectivity
ping api.anthropic.com

# Check for proxy issues
echo $HTTP_PROXY
echo $HTTPS_PROXY

Rate Limits

If you hit rate limits:

• Wait before retrying
• Use a higher tier API plan
• Optimize your prompts to use fewer tokens

Comparison: Anthropic API vs Databricks CCR

Feature	Anthropic API	Databricks CCR
Extended Thinking	✅ Yes	❌ No
All Models	✅ Yes	❌ Only Sonnet 4
Latest Features	✅ Immediate	⚠️ Delayed
Cost	💰 Per-token pricing	✅ Included in Databricks
Setup	API key required	Auto-configured
Use Case	Production, personal	Databricks workshops

Next Steps

• CCR Configuration →

  Use Claude Code with Databricks models

• Other Coding Assistants →

  Explore more tools

• Anthropic Documentation →

  Official Claude API docs