🤖 Tutorial: Customer Service Bot with Claude

Home › Docs › Claude Tutorials › Tutorial: Customer Service Bot with Claude
Claude Tutorials

Tutorial: Customer Service Bot with Claude

You are my X/Twitter customer service automation expert. I want to use XActions' customer service bot to automate responses to mentions, handle inquiries, and provide excellent support. Help me configure and run the full CS automation system.

Context

I'm using XActions (https://github.com/nirholas/XActions), an open-source X/Twitter toolkit. The src/automation/customerService.js is a browser-based bot that:

  • Monitors mentions, DMs, and replies in real-time
  • Auto-categorizes incoming messages (issue, thanks, pricing, question, etc.)
  • Responds with category-appropriate templates
  • Enforces business hours
  • Tracks all interactions with analytics
  • Has configurable rate limits and approval modes

What I Need You To Do

Phase 1: Understand the Customer Service Bot

The bot runs in the browser console on x.com. Here's how it works:

  1. Setup flow:

    • Navigate to x.com (logged into your business account)
    • Open DevTools → Console
    • Paste core.js (the XActions foundation)
    • Paste customerService.js
    • The bot starts monitoring immediately

  2. Bot architecture:

    Notification Check Loop
↓
Fetch @mentions from x.com/notifications/mentions
↓
For each mention:
  → Extract customer username + message text
  → Auto-categorize (keyword matching)
  → Generate templated response
  → Check rate limits + business hours
  → Send reply (or prepare for approval)
↓
Wait CHECK_INTERVAL_SECONDS
↓
Repeat

Phase 2: Configure the Bot

Here's every configuration option. Help me customize each section.

Account Configuration:

const ACCOUNTS = `
my_business_account
`.trim().split('\n').map(line => {
  const [username, password] = line.trim().split(':');
  return { username: username?.toLowerCase(), password };
}).filter(a => a.username);
  • Format: One account per line
  • Just the username if already logged in
  • username:password format if switching accounts

Monitoring Configuration:

MONITOR: {
  mentions: true,      // Watch @mentions
  dms: true,           // Watch DMs
  replies: true,       // Watch replies to your tweets
  keywords: true,      // Watch brand keywords
},

BRAND_KEYWORDS: [
  // 'your_brand_name',
  // 'your_product_name',
],

Response Settings:

RESPONSE: {
  autoReply: true,          // Auto-reply to mentions
  autoReplyDM: false,       // Auto-reply to DMs (careful!)
  requireApproval: true,    // Show response and wait for confirmation
  markAsRead: true,         // Mark notifications as read
},

Business Hours:

BUSINESS_HOURS: {
  enabled: true,
  timezone: 'America/New_York',
  start: 9,                 // 9 AM
  end: 17,                  // 5 PM
  days: [1, 2, 3, 4, 5],    // Mon-Fri (0=Sunday)
  outsideHoursMessage: "Thanks for reaching out! We're outside business hours (Mon-Fri 9AM-5PM ET). We'll respond next business day! 🙏",
},

Outside hours: bot sends the outsideHoursMessage instead of a category-specific response.

Rate Limits:

CHECK_INTERVAL_SECONDS: 60,      // How often to check notifications
RESPONSE_DELAY_SECONDS: 5,       // Wait before replying (looks natural)
MAX_RESPONSES_PER_HOUR: 20,
MAX_RESPONSES_PER_DAY: 100,

Phase 3: Response Templates

The bot uses categorized templates with placeholders:

const TEMPLATES = {
  greeting: [
    "Hi {customer}! Thanks for reaching out. How can I help? 😊",
    "Hey {customer}! We're here to help. What can we assist with?",
  ],
  issue: [
    "Hi {customer}, sorry about the issues. Please DM us details so we can help!",
    "Hey {customer}, we want to fix this! DM us your order details.",
  ],
  thanks: [
    "Thank you {customer}! We appreciate your kind words! 🙏❤️",
    "Wow, thanks {customer}! Comments like yours make our day! 🌟",
  ],
  question: [
    "Great question {customer}! Let me help. [CUSTOMIZE]",
  ],
  pricing: [
    "Hi {customer}! For pricing, check our site or DM for a quote! 💰",
  ],
  support: [
    "Hi {customer}! For fastest support, DM us or email support@company.com 📧",
  ],
  availability: [
    "Hi {customer}! We'll check availability and DM you!",
  ],
  escalate: [
    "Hi {customer}! Let's continue in DMs for better assistance. Send us a message! 🙌",
  ],
};

Placeholder variables:

  • {customer} → replaced with @username
  • {original} → replaced with their message text

Customize for my business:

  1. Replace generic templates with brand-voice responses
  2. Add product-specific categories
  3. Include links to FAQ, support pages, etc.
  4. Add seasonal/promotional templates

Phase 4: Auto-Categorization

The bot detects message categories using keyword matching:

const KEYWORD_CATEGORIES = {
  issue: ['problem', 'issue', 'broken', 'doesn\'t work', 'not working',
          'error', 'bug', 'complaint', 'disappointed', 'frustrated',
          'terrible', 'awful', 'worst', 'refund', 'cancel'],
  thanks: ['thank', 'thanks', 'awesome', 'amazing', 'love', 'great',
           'excellent', 'perfect', 'best', 'fantastic', 'wonderful'],
  pricing: ['price', 'cost', 'how much', 'pricing', 'discount',
            'coupon', 'deal', 'sale', 'expensive', 'cheap'],
  availability: ['available', 'in stock', 'restock', 'when',
                  'back in stock', 'sold out', 'inventory'],
  question: ['?', 'how do', 'how can', 'what is', 'where', 'when',
             'why', 'can you', 'could you', 'does it', 'is it'],
};

Customization ideas:

  • Add industry-specific keywords
  • Add product names to relevant categories
  • Create new categories (e.g., "shipping", "returns", "demo")
  • Adjust keyword priority order (first match wins)

Phase 5: Runtime Controls

Once the bot is running, use these console commands:

Command Description
stopCS() Stop the bot
csStats() Print interaction statistics
csTemplates() List all available templates
csRespond('category') Get a random response from a category (copied to clipboard)
csRespond('thanks', 'Custom message') Use a custom message

Stats output:

╔═══════════════════════════════════════════════════╗
║  📊 Customer Service Stats                        ║
╠═══════════════════════════════════════════════════╣
║  Total Interactions: 47                           ║
║  Today: 12   │  This Hour: 3                      ║
╠───────────────────────────────────────────────────╣
║  By Category:                                     ║
║    issue          : 15                            ║
║    thanks         : 12                            ║
║    question       : 10                            ║
║    pricing        : 5                             ║
║    greeting       : 5                             ║
╚═══════════════════════════════════════════════════╝

Phase 6: Approval Mode

When requireApproval: true:

  1. Bot detects a new mention
  2. Categorizes the message
  3. Generates a response
  4. Types the response into the reply box
  5. Logs: "📝 Response ready. Press Enter to send, Escape to cancel"
  6. You review and send or cancel

This is the recommended mode because:

  • Prevents embarrassing auto-replies
  • Lets you customize before sending
  • Catches categorization errors
  • Human-in-the-loop quality control

Phase 7: Full Setup Walkthrough

Here's the complete setup for MY business:

  1. Prepare core.js — Copy src/automation/core.js

  2. Customize customerService.js:

    • Set ACTIVE_ACCOUNT to your X handle
    • Configure BRAND_KEYWORDS for your brand
    • Set your BUSINESS_HOURS
    • Start with requireApproval: true
    • Customize TEMPLATES with your brand voice
    • Add product-specific categories to KEYWORD_CATEGORIES

  3. Go to x.com → DevTools → Console

  4. Paste core.js → Wait for "✅ Core loaded!"

  5. Paste customized customerService.js

  6. Watch the bot start:

    ╔═══════════════════════════════════════════════╗
║  🎧 XActions Customer Service Bot              ║
║  Active Account: @your_business                ║
║  Monitoring: Mentions ✅ │ DMs ✅ │ Replies ✅ ║
║  Auto-Reply: ✅ │ Approval: ✅                 ║
║  Business Hours: ✅ (9:00-17:00)               ║
╚═══════════════════════════════════════════════╝

Phase 8: Advanced Customization Ideas

  1. Multi-language support:

    • Detect language from incoming message
    • Maintain templates in multiple languages
    • Route to language-specific categories

  2. Priority detection:

    • Flag urgent issues (words: "emergency", "down", "broken")
    • Escalate VIP customers (verified accounts, high followers)
    • Fast-track returning customers

  3. CRM-like features using sessionStorage:

    // Track customer interaction history
const customerHistory = storage.get('cs_customer_history') || {};
// Check if returning customer
if (customerHistory[customer]) {
  // Personalize greeting
}

  4. Integration with MCP:
    Use MCP tools alongside the bot:

    "Check sentiment of our last 20 mentions"
"Get analytics on our response rate"
"Search for unresolved complaints"

My Business Details

(Replace before pasting)

  • Business X handle: @mybusiness
  • Brand keywords: brand1, product1, product2
  • Business hours: 9AM-5PM EST, Mon-Fri
  • Support email: support@mybusiness.com
  • Products/services offered: [describe]
  • Common customer issues: [list]
  • Brand voice: Professional / Casual / Friendly / Formal

Generate the fully customized customerService.js with my business details filled in.

⚡ Explore XActions

100% free and open-source. No API keys, no fees, no signup.

Browse All Documentation