Skip to main content
This guide walks you through setting up your first Duckie agent and deploying it to handle customer conversations. By the end, you’ll have an AI agent responding to messages in your support channel.

​
Prerequisites

Before you begin, make sure you have:
  • A Duckie account (sign up here or book a demo)
  • Admin access to at least one support channel (Zendesk, Intercom, Slack, etc.)
  • Some documentation or help content for your agent to reference

​
Step 1: Connect Your Support Channel

First, connect the platform where you want Duckie to respond to customers.
1

Go to Connections

Navigate to Settings → Connections in your Duckie dashboard.
2

Find Your Platform

Browse by category (Ticketing, Messaging) or search for your platform.
3

Click Connect

Click Connect and complete the OAuth authorization flow.
Popular starting points:
  • Zendesk — Respond to support tickets
  • Intercom — Handle chat conversations
  • Slack — Answer questions in shared channels
For detailed setup instructions, see our integration guides.

​
Step 2: Add Knowledge

Give your agent information to work with by connecting your documentation.
1

Go to Knowledge

Navigate to Train → Knowledge in your dashboard.
2

Connect a Source

Click Add Source and connect your documentation platform (Notion, Confluence, Google Drive) or add a website URL.
3

Wait for Sync

Duckie will index your content. This usually takes 2-5 minutes depending on volume.
Quick start option: If you don’t have documentation ready, you can create a few custom knowledge articles directly in Duckie to get started.

​
Step 3: Create Your Agent

Now create the AI agent that will handle conversations.
1

Go to Agents

Navigate to Build → Agents and click Create Agent.
2

Name Your Agent

Give it a descriptive name like “Support Agent” or “Help Desk Bot”.
3

Select a Runbook

Choose the default support runbook or create your own later.
4

Assign Knowledge

Select the knowledge sources you connected in Step 2.
5

Save

Click Create to save your agent.

​
Step 4: Test in the Playground

Before deploying to real customers, test your agent in the playground.
1

Open Playground

Navigate to Test → Playground.
2

Select Your Agent

Choose the agent you just created from the dropdown.
3

Send Test Messages

Ask questions your customers would ask. Try:
  • “How do I reset my password?”
  • “What’s your refund policy?”
  • “I’m having trouble logging in”
4

Review Responses

Check that responses are accurate and helpful. Click on any response to see the execution details.
Iterate here: If responses aren’t quite right, adjust your agent’s knowledge or guidelines before deploying. The playground is your safe space to experiment.

​
Step 5: Deploy to Your Channel

Once you’re happy with playground results, deploy your agent to handle real conversations.
1

Go to Deploy

Navigate to Deploy and click Create Deployment.
2

Select Your Agent

Choose the agent you created and tested.
3

Configure Trigger

Select your connected channel and the events that should trigger the agent (e.g., “New ticket created”, “Message received”).
4

Start in Shadow Mode

Select Shadow mode to review responses before they go to customers.
5

Activate

Click Create and set the deployment to Active.
We recommend starting in Shadow mode. In shadow mode, Duckie drafts responses as internal notes instead of sending them directly to customers. This lets you review and build confidence before going live.

​
Step 6: Go Live

After reviewing shadow responses and confirming quality:
  1. Go to Deploy
  2. Click on your deployment
  3. Change mode from Shadow to Live
  4. Save
Your agent is now responding to customers!

​
What’s Next?

Add Guidelines

Shape your agent’s tone and communication style

Set Up Guardrails

Define escalation rules and restrictions

Monitor Performance

Track metrics and view run history

Explore Integrations

Connect more tools and channels

​
Need Help?

  • Documentation — You’re in the right place! Use the sidebar to explore.
  • Community — Join our Discord to ask questions and share tips.
  • Support — Email [email protected] for direct assistance.