Skip to main content

Overview

The onboarding wizard guides you through creating a fully configured AI agent in just a few minutes. The process automatically discovers your website content, extracts your brand colors, and trains your agent on your knowledge base.
The onboarding wizard is designed to get you up and running quickly. You can always customize and add more content to your agent after the initial setup.

The 4-Step Process

1

Enter Your Domain

Provide your website URL to begin content discovery
2

Fetch & Select URLs

Review discovered pages and choose which content to import
3

Configure Branding

Customize your agent’s appearance with avatar, theme, and colors
4

Content Ingestion

Watch as your agent is created and trained on your content

Step 1: Enter Your Domain

The first step begins with identifying your website and naming your agent.

Agent Name

Give your agent a descriptive name that reflects its purpose. This name appears in the dashboard and helps you identify the agent among multiple agents in your workspace.
Choose a name that’s meaningful to your team. For example: “Sales Support Agent”, “Product Help Bot”, or “Customer FAQ Assistant”.

Website Domain

Enter your main website URL (e.g., https://yourdomain.com). The platform uses this to:
  • Discover pages: Automatically find all crawlable pages on your domain
  • Extract branding: Detect your brand colors for widget customization
  • Set context: Help your agent understand where it will be deployed
Input FormatExampleResult
Full URLhttps://example.comUses as-is
Domain onlyexample.comAdds https:// automatically
With subdomaindocs.example.comRespects subdomain

Step 2: Fetch & Select URLs

After entering your domain, the platform automatically discovers all available pages using intelligent web crawling.

URL Discovery

The crawler finds all public pages on your domain and displays them in a list. You’ll see:
  • Total pages found: The number of pages discovered
  • Page URLs: Full URLs of each discovered page
  • Selection controls: Options to select, deselect, or filter pages
The URL discovery typically finds up to 100 pages during onboarding. You can add more sources later from the Sources section.

Filtering Options

Use the built-in filters to refine your selection:
FilterPurposeExample
SearchFind specific URLs by keywordSearch for “product” to find product pages
ExcludeRemove URLs containing specific termsExclude “blog” to skip blog posts

Selecting Content

Choose pages that contain information your agent should know:
  • Product or service pages
  • FAQ and help documentation
  • About and contact pages
  • Pricing information
  • Feature descriptions
  • Terms of service and policies
  • Blog posts (unless relevant to support)
  • News articles or press releases
  • Login or account pages
  • Empty or placeholder pages
  • Duplicate content pages
Non-scrapable file types (images, PDFs, videos, etc.) are automatically filtered out. Only web pages with text content are included.

Managing Your Selection

  • Individual selection: Click checkboxes to select or deselect specific URLs
  • Bulk selection: Use shift+click to select a range of URLs
  • Remove URLs: Click the trash icon to remove unwanted URLs from the list
  • Clear all: Reset your selection and start fresh

Step 3: Configure Branding

This step customizes your agent’s visual appearance to match your brand identity.

Avatar Selection

Choose a 3D avatar that will represent your agent. The avatar:
  • Appears in the chat widget
  • Animates and speaks responses using text-to-speech
  • Creates a more engaging, human-like experience

Browse Gallery

Select from our collection of pre-made avatars with different personalities and styles.

Change Anytime

You can always change your avatar later from the agent settings.

Appearance Theme

Select between light and dark mode for your chat widget:
ThemeBest For
LightWebsites with light backgrounds, professional/corporate sites
DarkWebsites with dark backgrounds, tech/gaming sites, modern aesthetics

Brand Colors

The platform automatically extracts colors from your website:
  • Accent color: Your primary brand color
  • Dark color: A darker shade from your site
  • Light color: A lighter shade from your site
If automatic extraction doesn’t find your exact brand colors, you can manually enter a hex code or use the color picker.
You can customize your primary color by:
  1. Clicking extracted colors: Select from automatically detected brand colors
  2. Using the color picker: Choose any color visually
  3. Entering a hex code: Type an exact color value (e.g., #3B82F6)

Step 4: Content Ingestion

Once you click “Create Agent”, the platform begins building your AI assistant. This process runs automatically and shows real-time progress.

Progress Stages

The creation process includes several stages:
1

Creating Agent

Setting up your agent’s configuration and system prompt in the database.
2

Creating Widget

Generating the embeddable chat widget with your branding settings.
3

Preparing Sources

Initializing the knowledge base and preparing your selected URLs for processing.
4

Processing Pages

Crawling each selected URL, extracting text content, and creating embeddings for semantic search.
5

Building Knowledge

Indexing all processed content into the vector database for fast, accurate retrieval.

Progress Tracking

During ingestion, you’ll see:
MetricDescription
Overall progressPercentage of total completion
CompletedNumber of pages successfully processed
In ProgressPages currently being processed
FailedPages that couldn’t be processed (if any)
TotalTotal pages being trained
Processing time depends on the number of pages selected. Most agents complete in 1-3 minutes for 50-100 pages.

Handling Failures

If some pages fail to process:
  • The agent will still be created with successfully processed pages
  • You’ll see a summary of failed pages
  • You can retry failed sources later from the Sources section
  • Common failure reasons: pages blocked by robots.txt, authentication required, or temporary network issues

Completion

When processing finishes successfully:
  1. A success notification appears
  2. You’re automatically redirected to the Playground
  3. You can immediately start testing your agent with real questions

After Onboarding

Once your agent is created, you can continue customizing and improving it:

Troubleshooting

Possible causes:
  • The domain might be blocking crawlers via robots.txt
  • The website might require authentication
  • There could be a typo in the domain URL
Solutions:
  • Double-check the URL is correct
  • Try with or without www.
  • Contact your webmaster about crawler access
  • Manually add URLs after agent creation
Possible causes:
  • The website doesn’t have clear brand colors in its CSS
  • Colors are defined in images rather than code
  • The website is blocking external access
Solutions:
  • Use the manual color picker or enter your hex code directly
  • Check your brand guidelines for exact color values
Possible causes:
  • Individual pages may be blocking access
  • Content might be dynamically loaded (JavaScript-heavy)
  • Pages may have timed out during processing
Solutions:
  • The agent will work with successfully processed pages
  • Retry failed sources from the Sources section later
  • Consider using Q&A sources for critical content that won’t scrape
Possible causes:
  • Large number of pages selected
  • Pages with lots of content
  • Heavy traffic on the processing system
Solutions:
  • Be patient - large knowledge bases take longer to build
  • Consider starting with fewer pages and adding more later
  • The progress bar shows real-time status

Best Practices

Begin with 20-50 high-quality pages rather than importing everything. It’s easier to expand a working agent than to fix one with too much irrelevant content.
Take time to exclude pages that aren’t relevant to customer questions. Quality of sources matters more than quantity.
Use the Playground to ask real questions right away. This helps you identify gaps in your knowledge base early.
The default system prompt works well, but customizing it based on test results will improve your agent’s responses significantly.