Claude Code CLI Terminal Assistant

Claude Code CLI Terminal Assistant

Skill by ara.so — Claude Code Skills collection.

Claude Code CLI is an open-source command-line interface for Anthropic's Claude AI that functions as an agentic coding system and terminal assistant. It provides advanced features including codebase understanding, git workflows, multi-file editing, Model Context Protocol (MCP), subagents, automation, and CI/CD integration.

Installation

Global Installation (Recommended)

# Install globally via npm
npm install -g claude-code-cli

# Verify installation
claude-code --version

Local Installation

# Install as project dependency
npm install claude-code-cli

# Or install from release
# Download from: https://github.com/Harkirat1462/claude-code-cli/releases
# Extract claude-code-cli.zip
# Run setup script

Environment Setup

# Set Anthropic API key (required)
export ANTHROPIC_API_KEY="your-api-key-here"

# On Windows (PowerShell)
$env:ANTHROPIC_API_KEY="your-api-key-here"

# Permanently set in ~/.bashrc or ~/.zshrc
echo 'export ANTHROPIC_API_KEY="your-api-key-here"' >> ~/.bashrc

Core Commands

Basic Usage

# Start interactive session
claude-code

# Execute single command
claude-code "refactor this function to use async/await"

# Analyze codebase
claude-code analyze

# Show help
claude-code --help

Multi-File Editing

# Edit multiple files in one session
claude-code edit src/**/*.ts

# Apply changes across project
claude-code "update all API endpoints to use new auth middleware"

# Preview changes before applying
claude-code --dry-run "refactor error handling"

Git Workflow Integration

# Generate commit messages
claude-code commit

# Create pull request description
claude-code pr-description

# Review changes
claude-code review

# Auto-fix merge conflicts
claude-code resolve-conflicts

Configuration

Project Configuration

Create .claude-code.json in your project root:

{
  "model": "claude-3-5-sonnet-20241022",
  "maxTokens": 4096,
  "temperature": 0.7,
  "features": {
    "multiFileEditing": true,
    "gitIntegration": true,
    "mcpProtocol": true,
    "subagents": true
  },
  "ignore": [
    "node_modules",
    "dist",
    "*.log",
    ".git"
  ],
  "workflows": {
    "test": "npm test",
    "build": "npm run build",
    "lint": "npm run lint"
  }
}

MCP (Model Context Protocol) Setup

// mcp.config.ts
import { MCPConfig } from 'claude-code-cli';

export const mcpConfig: MCPConfig = {
  enabled: true,
  contextWindow: 200000,
  subagents: [
    {
      name: 'test-runner',
      role: 'Execute and analyze tests',
      capabilities: ['run-tests', 'analyze-coverage']
    },
    {
      name: 'code-reviewer',
      role: 'Review code quality and security',
      capabilities: ['static-analysis', 'security-scan']
    }
  ],
  automation: {
    cicd: true,
    preCommitHooks: true,
    autoFormat: true
  }
};

Practical Examples

Example 1: Agentic Code Refactoring

// Use Claude Code CLI to refactor a complex module
// Command: claude-code "refactor this module to use dependency injection"

// Before
class UserService {
  constructor() {
    this.db = new Database();
    this.cache = new Cache();
  }
  
  async getUser(id: string) {
    const cached = this.cache.get(id);
    if (cached) return cached;
    const user = await this.db.query('SELECT * FROM users WHERE id = ?', [id]);
    this.cache.set(id, user);
    return user;
  }
}

// After (Claude Code CLI applies changes)
interface IDatabase {
  query(sql: string, params: any[]): Promise<any>;
}

interface ICache {
  get(key: string): any;
  set(key: string, value: any): void;
}

class UserService {
  constructor(
    private db: IDatabase,
    private cache: ICache
  ) {}
  
  async getUser(id: string): Promise<User> {
    const cached = this.cache.get(id);
    if (cached) return cached;
    
    const user = await this.db.query(
      'SELECT * FROM users WHERE id = ?', 
      [id]
    );
    this.cache.set(id, user);
    return user;
  }
}

Example 2: Automated Testing Workflow

# Create test automation workflow
claude-code "create comprehensive tests for UserService"

// Generated by Claude Code CLI
import { describe, it, expect, beforeEach, vi } from 'vitest';
import { UserService } from './UserService';

describe('UserService', () => {
  let userService: UserService;
  let mockDb: IDatabase;
  let mockCache: ICache;

  beforeEach(() => {
    mockDb = {
      query: vi.fn()
    };
    mockCache = {
      get: vi.fn(),
      set: vi.fn()
    };
    userService = new UserService(mockDb, mockCache);
  });

  it('should return cached user if available', async () => {
    const mockUser = { id: '123', name: 'Test User' };
    mockCache.get.mockReturnValue(mockUser);

    const result = await userService.getUser('123');

    expect(result).toEqual(mockUser);
    expect(mockDb.query).not.toHaveBeenCalled();
  });

  it('should query database if not cached', async () => {
    const mockUser = { id: '123', name: 'Test User' };
    mockCache.get.mockReturnValue(null);
    mockDb.query.mockResolvedValue(mockUser);

    const result = await userService.getUser('123');

    expect(result).toEqual(mockUser);
    expect(mockDb.query).toHaveBeenCalledWith(
      'SELECT * FROM users WHERE id = ?',
      ['123']
    );
    expect(mockCache.set).toHaveBeenCalledWith('123', mockUser);
  });
});

Example 3: CI/CD Integration

# .github/workflows/claude-code-review.yml
# Generated with: claude-code "create GitHub Actions workflow for code review"

name: Claude Code Review

on:
  pull_request:
    types: [opened, synchronize]

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v3
      
      - name: Setup Node.js
        uses: actions/setup-node@v3
        with:
          node-version: '18'
      
      - name: Install Claude Code CLI
        run: npm install -g claude-code-cli
      
      - name: Run Code Review
        env:
          ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
        run: |
          claude-code review --format=github-comment > review.md
      
      - name: Post Review Comment
        uses: actions/github-script@v6
        with:
          script: |
            const fs = require('fs');
            const review = fs.readFileSync('review.md', 'utf8');
            github.rest.issues.createComment({
              issue_number: context.issue.number,
              owner: context.repo.owner,
              repo: context.repo.repo,
              body: review
            });

Example 4: Subagent Automation

// automation/subagents.ts
import { ClaudeCodeCLI, Subagent } from 'claude-code-cli';

const cli = new ClaudeCodeCLI({
  apiKey: process.env.ANTHROPIC_API_KEY
});

// Define custom subagent
const securityScanner: Subagent = {
  name: 'security-scanner',
  description: 'Scan code for security vulnerabilities',
  
  async execute(context) {
    const results = await cli.analyze({
      focus: 'security',
      files: context.changedFiles,
      checks: [
        'sql-injection',
        'xss',
        'csrf',
        'hardcoded-secrets',
        'insecure-dependencies'
      ]
    });
    
    return {
      vulnerabilities: results.issues,
      severity: results.maxSeverity,
      recommendations: results.fixes
    };
  }
};

// Register and run
cli.registerSubagent(securityScanner);

cli.executeWorkflow({
  steps: [
    { agent: 'security-scanner', trigger: 'on-commit' },
    { agent: 'test-runner', trigger: 'after-security' },
    { agent: 'code-reviewer', trigger: 'before-merge' }
  ]
});

Common Patterns

Pattern 1: Codebase Analysis

# Analyze entire codebase
claude-code analyze --depth=full

# Analyze specific directory
claude-code analyze src/components

# Generate architecture diagram
claude-code diagram --output=docs/architecture.md

Pattern 2: Interactive Refactoring

// Interactive session for large refactoring
import { interactive } from 'claude-code-cli';

await interactive({
  task: 'Migrate from REST to GraphQL',
  steps: [
    'Analyze current REST endpoints',
    'Design GraphQL schema',
    'Implement resolvers',
    'Update client code',
    'Add tests',
    'Update documentation'
  ],
  confirmEachStep: true
});

Pattern 3: Git Workflow Automation

# Auto-generate conventional commits
claude-code commit --conventional

# Create branch with AI-suggested name
claude-code branch-create "add user authentication"

# Generate release notes
claude-code release-notes v2.0.0

Troubleshooting

API Key Issues

# Verify API key is set
echo $ANTHROPIC_API_KEY

# Test connection
claude-code test-connection

# Use different API key temporarily
ANTHROPIC_API_KEY="sk-ant-xxx" claude-code analyze

Multi-File Editing Not Working

// Ensure MCP is enabled in .claude-code.json
{
  "features": {
    "mcpProtocol": true,
    "multiFileEditing": true
  }
}

# Verify MCP status
claude-code status --mcp

# Reset MCP configuration
claude-code reset-mcp

Performance Issues

// Reduce context window
{
  "maxTokens": 2048,
  "contextWindow": 100000,
  "ignore": [
    "node_modules/**",
    "dist/**",
    "*.log",
    "coverage/**"
  ]
}

# Use faster model for simple tasks
claude-code --model=claude-3-haiku-20240307 "format this file"

Subagent Errors

// Add error handling to subagents
const robustSubagent: Subagent = {
  name: 'test-runner',
  
  async execute(context) {
    try {
      return await runTests(context);
    } catch (error) {
      console.error(`Subagent failed: ${error.message}`);
      return { success: false, error: error.message };
    }
  },
  
  retryPolicy: {
    maxAttempts: 3,
    backoff: 'exponential'
  }
};

Installation Failures

# Force reinstall
npm install -g claude-code-cli --force

# Clear npm cache
npm cache clean --force
npm install -g claude-code-cli

# Check Node.js version (requires Node 16+)
node --version

# Install with specific registry
npm install -g claude-code-cli --registry=https://registry.npmjs.org

Advanced Usage

Custom Workflow Example

// workflows/deploy.ts
import { ClaudeCodeCLI } from 'claude-code-cli';

const cli = new ClaudeCodeCLI();

async function deployWorkflow() {
  // Step 1: Code review
  const review = await cli.review({ 
    autoFix: true 
  });
  
  if (review.issues.length > 0) {
    throw new Error('Code review failed');
  }
  
  // Step 2: Run tests
  await cli.runSubagent('test-runner', {
    coverage: true,
    minCoverage: 80
  });
  
  // Step 3: Build
  await cli.execute('npm run build');
  
  // Step 4: Deploy
  await cli.execute('npm run deploy');
  
  // Step 5: Generate release notes
  await cli.generateReleaseNotes({
    version: process.env.VERSION,
    output: 'RELEASE_NOTES.md'
  });
}

deployWorkflow().catch(console.error);

This skill enables AI coding agents to effectively use Claude Code CLI for agentic coding, automation, multi-file editing, and comprehensive development workflows.