Skip to content

3.1: Understanding Connections

Congratulations! You've built an AI agent that can answer questions, search knowledge, and use tools. But right now, there's one problem:

Your agent lives inside a table.

Customers can't talk to it directly. You have to manually add rows for each conversation. It's like having a brilliant employee who only works when you hand-deliver messages to their desk!

Let's fix that. Let's connect your agent to the real world.


3.1.1: What Are Connections?

💡 💡 Think of It Like This

Connections are communication channels - the ways customers can reach your agent.

Just like you might reach a business through:

  • Their website chat
  • WhatsApp
  • Email
  • Phone

Your agent can communicate through all these channels too!

Common channels you can connect:

Website Chat

Add a chat widget to your website for instant customer support

WhatsApp

Connect to WhatsApp Business and respond to messages automatically

Email

Monitor email inboxes and draft automatic responses

Slack

Create a bot for your team channels and answer questions

Telegram

Build Telegram bots for groups or direct messages

More Coming Soon

Discord, SMS, Instagram, and more channels

Each connection is like opening a new door for customers to reach your agent. Want to offer support via WhatsApp and website chat and email? Just add all three connections!


3.1.2: How Connections Work

When you add a connection to your table, something magical happens. Your agent goes from being a passive tool to an active system that responds in real-time.

Here's the flow:

How Connections Work

1

Connection Listens for Messages

Your connection monitors a channel (website, email, WhatsApp) and waits for new messages to arrive.

2

Creates a New Row

When a message arrives, the connection automatically creates a new row in your table with the message content.

3

Your Agent Processes It

Your columns automatically run: AI analyzes the message, searches knowledge, generates a response.

4

Response Goes Back

The response is automatically sent back through the same channel the message came from.

💡 Everything happens automatically - from the moment the message arrives to when the response is sent.

Let's break this down:

  1. Connection Listens - Your connection constantly monitors a channel (like your support@company.com inbox or your WhatsApp Business number)

  2. Creates a Row - The instant a message arrives, the connection automatically creates a new row in your table with the message content

  3. Agent Processes - Your columns spring into action: AI analyzes the message, searches knowledge, generates a response

  4. Response Sent - The response automatically goes back through the same channel

💡 🎉 This Changes Everything

Before connections: You manually add rows → Agent processes → You copy responses back

After connections: Customer sends message → Everything happens automatically → Customer gets response

You've gone from semi-automated to fully automated!


3.1.3: The Two-Part Setup

Every connection has two sides, like a conversation:

Two-Part Connection Setup

Incoming

How messages arrive

  • Configure which channel to listen to
  • Set up webhooks or monitoring
  • Choose which column receives the data
  • Automatically creates new rows
Example: "When email arrives at support@mycompany.com, create a new row and put the email content in the 'Message' column"

Outgoing

How responses are sent

  • Add a "Send" column to your table
  • Choose which column's content to send
  • Configure where to send it
  • Automatically sends when ready
Example: "Send the content from the 'AI Response' column back to the customer via email"

💡 Both parts work together: Incoming creates rows, your columns process them, Outgoing sends responses

Why Two Parts?

Think about any conversation:

  • Someone needs to receive the message (incoming)
  • Someone needs to send the response (outgoing)

Your agent is the same! You configure:

  1. How to receive messages (incoming connection)
  2. How to send responses (outgoing column)

Example: Email Support System

  • Incoming: "Monitor support@mycompany.com and create a row for each email"
  • Outgoing: "Send the AI response back to the customer's email address"

Example: Website Chat

  • Incoming: "When someone types in the chat widget, create a row"
  • Outgoing: "Send the AI response back to the chat widget"

3.1.4: Real-World Example

Let's see this in action with a customer support scenario:

Without Connections (Manual):

8:30 AM - Customer emails: "What's your return policy?"
9:15 AM - You check email
9:20 AM - You manually add a row: "Customer: john@example.com, Message: What's your return policy?"
9:21 AM - Agent generates response
9:25 AM - You copy the response and email it back
9:30 AM - Customer receives response (1 hour later)

With Connections (Automatic):

8:30:00 AM - Customer emails: "What's your return policy?"
8:30:02 AM - Connection creates row automatically
8:30:03 AM - Agent searches knowledge base
8:30:05 AM - Agent generates response
8:30:06 AM - Response sent automatically to customer
8:30:07 AM - Customer receives response (7 seconds later)

💡 The Impact

  • Response time: 1 hour → 7 seconds
  • 🎯 Accuracy: Always includes latest return policy
  • 😌 Your time: 15 minutes → 0 minutes
  • 🎉 Customer satisfaction: Much happier!

3.1.5: What You'll Learn Next

In the next sections, we'll set up connections for each channel:

  • Website Chat Widget - Add a chat bubble to your site
  • WhatsApp Integration - Connect to WhatsApp Business
  • Email Integration - Monitor email inboxes
  • Slack Integration - Create team bots
  • Telegram Integration - Build Telegram bots

Each one follows the same pattern:

  1. Set up incoming (how messages arrive)
  2. Set up outgoing (how responses are sent)
  3. Test it!

💡 Ready to Connect?

By the end of Part 3, your agent will be live and talking to customers across multiple channels automatically.

No more manual work. Just pure automation. ✨


Previous: 2.4 Customizing PersonalityNext: 3.2 Website Chatbot Widget

Built with VitePress