Build a Support Bot with Slack and Zendesk

What You’ll Build

Build an intelligent support bot that:

Monitors your Slack support channels for customer questions
Automatically checks ticket status in Zendesk when customers ask
Uses AI to respond to common questions without human intervention
Routes complex issues to your support team with proper context

This tutorial demonstrates real-world multi-service integration using Metorial’s MCP servers.

What you’ll learn:

Deploying multiple MCP servers (Slack and Zendesk)
Setting up OAuth for user authentication
Creating an AI agent that uses tools from multiple services
Implementing decision logic for automated responses

Before you begin:

Create a Metorial account
Create API keys
Slack workspace with admin access
Zendesk account with API access
OpenAI or Anthropic API key

Time to complete: 15-20 minutes

Prerequisites

Before building the support bot, ensure you have:

Metorial setup:
- Active Metorial account at app.metorial.com
- Project created in your organization
- Metorial API key (generate in Dashboard → Home → Connect to Metorial)
Slack workspace:
- Admin access to create OAuth apps
- Support channel where bot will listen (e.g., #customer-support)
Zendesk account:
- Admin access to create API tokens
- Active tickets for testing
AI provider:
- OpenAI API key (GPT-4 recommended) OR
- Anthropic API key (Claude Sonnet 4.5 recommended)
Development environment:
- Node.js 18+ (TypeScript) or Python 3.9+ installed
- Basic knowledge of async/await patterns

Architecture Overview

The support bot follows this workflow:

Listen: Monitor Slack channels for customer messages
Analyze: AI determines if message is about ticket status, common question, or complex issue
Act:
- Ticket status: Query Zendesk and post status to Slack
- Common question: AI generates and posts response to Slack
- Complex issue: Tag support team in dedicated channel with context

Services used:

Slack MCP Server: Read messages, post responses, send DMs
Zendesk MCP Server: Search tickets, read ticket details, add internal notes
AI Model: Decision-making and response generation

Step 1: Deploy Slack MCP Server

Deploy the Slack MCP server from Metorial’s catalog to enable your bot to interact with Slack.

Navigate to Server Catalog

In the Metorial Dashboard, go to Servers and search for “Slack”.

Deploy Slack Server

Click the Slack server, then click Deploy Server → Server Deployment.Give your deployment a descriptive name like “Support Bot Slack”.

Configure OAuth (Next Step)

Note your Server Deployment ID (shown on the deployment page). You’ll need this for OAuth setup.We’ll configure Slack OAuth in Step 3.

Keep your Slack deployment ID handy. You’ll use it when setting up OAuth and writing bot code.

Step 2: Deploy Zendesk MCP Server

Deploy the Zendesk MCP server to enable ticket querying and management.

Find Zendesk in Catalog

In the Metorial Dashboard, search for “Zendesk” in the Servers section.

Deploy Zendesk Server

Click Deploy Server → Server Deployment.Name it “Support Bot Zendesk”.

Configure API Authentication

Zendesk requires API token authentication:

In Zendesk Admin Center, go to Apps and Integrations → APIs → Zendesk API
Enable Token Access
Click Add API token, give it a description (e.g., “Metorial Support Bot”)
Copy the token and paste it in the Metorial deployment configuration
Enter your Zendesk subdomain (e.g., yourcompany.zendesk.com)

Save Configuration

Click Deploy. Note your Server Deployment ID for use in the code.

Step 3: Set Up OAuth Authentication

Your support bot needs permission to access your Slack workspace on behalf of your team.

Install Dependencies

Install the Metorial SDK and your chosen AI provider:

npm install metorial @metorial/anthropic @anthropic-ai/sdk

Create OAuth Session

Run this code to generate the Slack OAuth URL:

import { Metorial } from 'metorial';

const metorial = new Metorial({
  apiKey: process.env.METORIAL_API_KEY
});

async function setupSlackOAuth() {
  const slackOAuth = await metorial.oauth.sessions.create({
    serverDeploymentId: 'your-slack-deployment-id',
    callbackUrl: 'https://yourapp.com/oauth/callback' // Optional
  });

  console.log('Authorize Slack here:', slackOAuth.url);
  console.log('OAuth Session ID:', slackOAuth.id);

  // Wait for authorization
  await metorial.oauth.waitForCompletion([slackOAuth.id]);
  console.log('✓ Slack authorized!');

  // Save slackOAuth.id for future use
  return slackOAuth.id;
}

setupSlackOAuth();

Authorize in Browser

Open the printed OAuth URL in your browser
Select your Slack workspace
Review and approve the permissions
You’ll be redirected to your callback URL (or see a confirmation page)

Store OAuth Session ID

Save the OAuth session ID securely. You’ll reuse it for all future bot operations without re-authorizing.For production apps, store OAuth session IDs in your database per user/workspace.

OAuth vs API Keys:

Slack uses OAuth (user authorization)
Zendesk uses API tokens (configured in deployment)

You only need OAuth setup for Slack.

Step 4: Build the Support Bot Core Logic

Create the main bot that connects to both Slack and Zendesk with AI-powered decision making.

import { Metorial } from 'metorial';
import { metorialAnthropic } from '@metorial/anthropic';
import Anthropic from '@anthropic-ai/sdk';

const metorial = new Metorial({
  apiKey: process.env.METORIAL_API_KEY!
});

const anthropic = new Anthropic({
  apiKey: process.env.ANTHROPIC_API_KEY!
});

// Store your deployment IDs and OAuth session ID
const SLACK_DEPLOYMENT_ID = 'your-slack-deployment-id';
const ZENDESK_DEPLOYMENT_ID = 'your-zendesk-deployment-id';
const SLACK_OAUTH_SESSION_ID = 'your-slack-oauth-session-id';

async function runSupportBot(userMessage: string, slackChannel: string) {
  await metorial.withProviderSession(
    metorialAnthropic,
    {
      serverDeployments: [
        {
          serverDeploymentId: SLACK_DEPLOYMENT_ID,
          oauthSessionId: SLACK_OAUTH_SESSION_ID
        },
        {
          serverDeploymentId: ZENDESK_DEPLOYMENT_ID
        }
      ],
      streaming: false
    },
    async ({ tools, callTools, closeSession }) => {
      // AI analyzes the message and decides what action to take
      const messages: Anthropic.MessageParam[] = [
        {
          role: 'user',
          content: `You are a support bot. A customer wrote: "${userMessage}"

Analyze this message and take appropriate action:

1. If asking about ticket status (e.g., "what's the status of ticket #1234"):
   - Use Zendesk tools to search for and retrieve ticket details
   - Post a summary to Slack channel ${slackChannel}

2. If asking a common question (e.g., "how do I reset my password"):
   - Generate a helpful response
   - Post it to Slack channel ${slackChannel}

3. If it's complex or requires human attention:
   - Post to #support-escalations channel
   - Tag @support-team
   - Include customer's original message

Take action now using the available tools.`
        }
      ];

      let response = await anthropic.messages.create({
        model: 'claude-sonnet-4-5',
        max_tokens: 2048,
        messages,
        tools
      });

      // Handle tool calls in agentic loop
      while (response.stop_reason === 'tool_use') {
        const toolUseBlocks = response.content.filter(
          (block): block is Anthropic.ToolUseBlock => block.type === 'tool_use'
        );

        // Execute tools via Metorial
        const toolResults = await callTools(toolUseBlocks);

        // Add assistant response and tool results to conversation
        messages.push(
          { role: 'assistant', content: response.content },
          {
            role: 'user',
            content: toolResults.map((result, i) => ({
              type: 'tool_result' as const,
              tool_use_id: toolUseBlocks[i].id,
              content: JSON.stringify(result)
            }))
          }
        );

        // Continue conversation
        response = await anthropic.messages.create({
          model: 'claude-sonnet-4-5',
          max_tokens: 2048,
          messages,
          tools
        });
      }

      // Get final text response
      const finalText = response.content
        .filter((block): block is Anthropic.TextBlock => block.type === 'text')
        .map(block => block.text)
        .join('\n');

      console.log('Bot action completed:', finalText);

      await closeSession();
    }
  );
}

// Example usage
runSupportBot(
  "What's the status of ticket #1234?",
  "#customer-support"
);

What this code does:

Creates a provider session with both Slack and Zendesk tools available
Sends customer message to AI with instructions on how to handle different scenarios
AI decides and executes tools:
- Searches Zendesk if ticket status requested
- Posts to Slack with responses
- Routes to support team if needed
Handles multi-step tool calls in an agentic loop until complete

This example uses Claude’s agentic capabilities—the AI autonomously decides which tools to call and when. You don’t need to write routing logic yourself.

Step 5: Check Ticket Status

Let’s test the bot with a ticket status inquiry. Scenario: Customer asks “What’s the status of my ticket #1234?” The bot will:

Parse ticket number from message
Call Zendesk’s search_tickets or get_ticket tool
Format ticket details (status, assignee, last update)
Post to Slack

// Test ticket status check
await runSupportBot(
  "Hi! Can you check the status of ticket #1234?",
  "#customer-support"
);

// Expected flow:
// 1. AI calls Zendesk tool: zendesk_get_ticket(ticket_id: 1234)
// 2. AI formats response: "Ticket #1234 is currently 'In Progress' and assigned to Sarah. Last updated 2 hours ago."
// 3. AI calls Slack tool: slack_post_message(channel: "#customer-support", text: "...")

The AI autonomously chains tool calls:

First calls zendesk_get_ticket
Receives ticket data
Then calls slack_post_message with formatted response

This is the power of agentic AI with multi-tool access.

Step 6: AI-Powered Auto-Response

Test the bot with a common question that doesn’t require tools. Scenario: Customer asks “How do I reset my password?” The bot will:

Recognize this as a common question
Generate helpful response using its knowledge
Post directly to Slack

// Test auto-response for common question
await runSupportBot(
  "How do I reset my password?",
  "#customer-support"
);

// Expected flow:
// 1. AI recognizes common question (no Zendesk needed)
// 2. AI generates response with password reset steps
// 3. AI calls Slack tool: slack_post_message(channel: "#customer-support", text: "To reset your password: 1. Go to...")

Customizing Auto-Responses:For production bots, enhance the system prompt with:

Your company’s knowledge base articles
FAQs and standard responses
Links to help documentation
Escalation criteria (e.g., “always escalate billing questions”)

Step 7: Route Complex Issues to Humans

Test the bot with a complex issue that needs human attention. Scenario: Customer reports “Your API is returning 500 errors and our production is down!” The bot will:

Recognize this as urgent and complex
Post to escalation channel
Tag support team
Include original message and urgency

// Test escalation for complex issue
await runSupportBot(
  "Your API is returning 500 errors and our production is down!",
  "#customer-support"
);

// Expected flow:
// 1. AI recognizes urgency and complexity
// 2. AI calls Slack tool: slack_post_message(
//      channel: "#support-escalations",
//      text: "🚨 URGENT: Production issue reported by customer..."
//      mentions: ["@support-team"]
//    )
// 3. AI calls Slack tool: slack_post_message(
//      channel: "#customer-support",
//      text: "I've escalated this to our engineering team. Someone will assist you shortly."
//    )

Configuring Escalation Rules:Update the system prompt to define when to escalate:

Keywords: “down”, “urgent”, “billing”, “refund”, “legal”
Customer sentiment: angry or frustrated tone
Business rules: questions about enterprise plans, integrations

Step 8: Test End-to-End

Run a complete test to verify all functionality.

Monitor Slack Channel

Open your Slack workspace and watch the #customer-support channel.

Simulate Customer Messages

Run the bot with different test messages:

// Test all scenarios
await runSupportBot("What's the status of ticket #1234?", "#customer-support");
await runSupportBot("How do I reset my password?", "#customer-support");
await runSupportBot("API is down in production!", "#customer-support");

Verify Bot Behavior

Check that:

✓ Ticket status posted to Slack with correct details
✓ Password reset instructions posted clearly
✓ Urgent issue escalated to #support-escalations
✓ Acknowledgment posted to customer in original channel

Check Zendesk Integration

Verify the bot can read tickets:

Go to your Zendesk dashboard
Create a test ticket
Ask the bot about that ticket ID
Confirm details match

Debugging Tips:If tools aren’t being called:

Check OAuth session is active (try re-authorizing)
Verify deployment IDs are correct
Check API tokens for Zendesk haven’t expired
Review Metorial dashboard logs under Monitoring

What’s Next?

Congratulations! You’ve built a production-ready support bot that integrates Slack and Zendesk with AI-powered decision making.

Enhancements to Try

Add Message Monitoring

Use Slack’s Events API to trigger bot on new messages automatically instead of manual calls.

Sentiment Analysis

Enhance AI prompt to detect frustrated customers and escalate proactively.

Create Zendesk Tickets

Allow bot to create new tickets when customers report issues without ticket IDs.

Multi-Language Support

Add language detection and translation for global support teams.

Learn More

SDK Documentation

Explore advanced SDK features and patterns.

OAuth Guide

Learn more about managing user authorizations.

Monitoring Guide

Monitor tool calls and debug issues.

Production Considerations

Before deploying to production:

Error handling: Add try/catch blocks and retry logic
Rate limiting: Implement backoff for high-volume channels
Security: Store API keys in environment variables or secrets manager
Monitoring: Set up alerts for failed tool calls
Testing: Create automated tests for common scenarios

Need help? Email us at support@metorial.com.

Metorial 101

Sample Projects

API

Build a Support Bot with Slack and Zendesk

What You’ll Build

Prerequisites

Architecture Overview

Step 1: Deploy Slack MCP Server

Step 2: Deploy Zendesk MCP Server

Step 3: Set Up OAuth Authentication

Step 4: Build the Support Bot Core Logic

Step 5: Check Ticket Status

Step 6: AI-Powered Auto-Response

Step 7: Route Complex Issues to Humans

Step 8: Test End-to-End

What’s Next?

Enhancements to Try

Add Message Monitoring

Sentiment Analysis

Create Zendesk Tickets

Multi-Language Support

Learn More

SDK Documentation

OAuth Guide

Monitoring Guide

Production Considerations

Metorial 101

Sample Projects

API

​What You’ll Build

​Prerequisites

​Architecture Overview

​Step 1: Deploy Slack MCP Server

​Step 2: Deploy Zendesk MCP Server

​Step 3: Set Up OAuth Authentication

​Step 4: Build the Support Bot Core Logic

​Step 5: Check Ticket Status

​Step 6: AI-Powered Auto-Response

​Step 7: Route Complex Issues to Humans

​Step 8: Test End-to-End

​What’s Next?

​Enhancements to Try

Add Message Monitoring

Sentiment Analysis

Create Zendesk Tickets

Multi-Language Support

​Learn More

SDK Documentation

OAuth Guide

Monitoring Guide

​Production Considerations

What You’ll Build

Prerequisites

Architecture Overview

Step 1: Deploy Slack MCP Server

Step 2: Deploy Zendesk MCP Server

Step 3: Set Up OAuth Authentication

Step 4: Build the Support Bot Core Logic

Step 5: Check Ticket Status

Step 6: AI-Powered Auto-Response

Step 7: Route Complex Issues to Humans

Step 8: Test End-to-End

What’s Next?

Enhancements to Try

Learn More

Production Considerations