Creating your first AI agent
In HaloDesk, an AI agent is the combination of:
A chat widget (the AI chatbot users talk to on your website)
A dedicated Inbox (where escalated conversations become tickets)
Each agent has its own:
Knowledge base
Chat widget configuration
Inbox and ticket flow
Assignment rules and settings
This means:
One agent = one chatbot + one inbox
Multiple agents = separate chatbots and separate inboxes
This allows you to fully separate support flows by product, website, or use case.
Step 1: Create a new AI agent
From the HaloDesk dashboard:
Click Create new AI agent
Enter your website URL
Choose the agent use case (for example: Support or Sales)
The website URL is used to automatically train the agent on your existing content.
Step 2: How website training works
When you enter your website URL, HaloDesk will:
Scan the entire domain
Detect all pages under that domain
Fetch your logo and brand colors
Prepare all found pages as potential knowledge sources
In the next step, you can review and unselect any pages you don’t want the AI to learn from.
You can also skip pages like:
Legal pages
Marketing-only pages
Irrelevant blog posts
Step 3: Agent name vs Assistant name
This distinction is important.
Agent name
This is usually the name of your product or service (for example: HaloDesk or Pipedrive).Assistant name
This is the name of the AI itself (for example: HaloDesk AI Support).
When customers chat with your widget, messages will appear as coming from the assistant name.
Step 4: Customize the chat widget basics
Before finishing setup, you can:
Choose the agent name
Set the assistant name
Upload a profile picture for the chatbot
Customize the chat bubble icon
Select primary and button colors
Decide whether the header uses your primary color
These settings control how the widget looks on your website.
Step 5: Finish and train the agent
Once you click Finish and train agent:
You’ll be taken to the Sources page
All selected links, files, and text will be listed
You can start training immediately or make changes first
Training makes your agent usable.
Creating multiple agents
You can repeat this process to create:
Separate agents per product
Different agents for different websites
Dedicated agents for support and sales
Each agent has its own:
Knowledge base
Inbox
Settings
Widget configuration