Quickstart

​

Prerequisites

• A BoostGPT account (sign up free)
• An AI agent created and trained with your content
• Access to your website’s code

​

Step 1: Create an Event

1. Go to your agent’s Events page in the dashboard
2. Click Create Event
3. Fill in the details:
   • Name: checkout_abandoned (use snake_case, no spaces)
   • Label: “Checkout Abandoned” (display name)
   • Description: “User started checkout but didn’t complete”
   • Prompt Template:
     A user abandoned their checkout. Cart value: {{cart_value}}.
     Help them complete their purchase.
     
4. Click Save

​

Step 2: Add the Chat Widget

<script src="https://embed.boostgpt.co/widget.js" data-bot-id="YOUR_AGENT_UUID"></script>

​

Step 3: Trigger on Friction Points

// Example: Trigger when user is about to leave checkout
window.addEventListener('beforeunload', function(e) {
  if (isOnCheckoutPage() && cartHasItems()) {
    boostgpt.trigger('checkout_abandoned', {
      cart_value: '$149.00',
      items: 3
    });
  }
});

// Example: Trigger after 30 seconds of inactivity on checkout
let timeout;
function resetTimer() {
  clearTimeout(timeout);
  timeout = setTimeout(function() {
    if (isOnCheckoutPage()) {
      boostgpt.trigger('checkout_abandoned', {
        cart_value: getCartValue()
      });
    }
  }, 30000);
}

document.addEventListener('mousemove', resetTimer);
document.addEventListener('keypress', resetTimer);

​

Step 4: Test Your Trigger

1. Open your website in a browser
2. Navigate to the page where you added the trigger
3. Trigger the condition (e.g., sit idle for 30 seconds)
4. You should see the trigger card appear above the widget bubble
5. Click it — the chat opens in the widget with context

​

Customizing the Trigger UI

boostgpt.trigger('checkout_abandoned', {
  cart_value: '$149.00'
}, {
  text: 'Need help checking out?',  // Message shown in the trigger card
  delay: 1000                       // Delay before showing (ms)
});

​

What Happens When User Clicks

1. The widget calls the BoostGPT API with the event and context
2. Your agent receives a message built from the prompt template + context
3. The agent generates a contextual response
4. The chat opens in the widget with the agent already responding

“Hi! I noticed you were about to leave without completing your order. Your cart has $149.00 worth of items. Is there anything I can help with? Maybe you have questions about shipping, returns, or payment options?”

​

Next Steps