Frequently Asked Questions

KitePin is an AI-powered customer support platform that lets you deploy voice and chat agents on your website. These agents can handle customer inquiries 24/7, answer questions, take orders, schedule appointments, and more.

Here's how it works:

• Create an agent and define its persona, knowledge, and behavior
• Connect your OpenAI API key (BYOK model)
• Customize the widget's appearance to match your brand
• Add a simple script to your website
• Your AI agent is live and ready to help customers!

You can have a basic agent up and running in 5-10 minutes. The setup wizard guides you through naming your agent, selecting your industry, and configuring basic settings.

For more advanced customization (detailed prompts, knowledge base, webhooks, function calling), plan for 30-60 minutes depending on complexity.

No coding skills required for basic setup! KitePin is designed to be user-friendly with a visual interface for configuring agents, customizing appearance, and managing settings.

For installation, you'll just need to paste a single line of script code into your website. If you use website builders like WordPress, Shopify, or Wix, this is usually as simple as adding it to a "custom code" section.

Advanced features like webhooks and function calling may require some technical knowledge, but our documentation walks you through each step.

To get started with KitePin, you'll need:

• A KitePin account – Sign up at go.kitepin.com
• An OpenAI API key – Get one from platform.openai.com
• A website – Where you'll embed the agent widget
• Your business context – Information about what you want the agent to help with

Yes! Every agent gets a unique preview URL (e.g., go.kitepin.com/x/your-code) where you can test conversations before embedding on your live site. You can also toggle agents between active and inactive states at any time.

KitePin offers a limited feature free trail. When you buy any plan, we offer14-day money-back guarantee. Purchase any plan, test it thoroughly, and if it doesn't meet your needs, request a full refund within 14 days — no questions asked.

BYOK means you connect your own OpenAI API key directly to KitePin. You pay OpenAI at their standard rates with zero markup from us. This typically saves you 60-80% compared to platforms that mark up API costs.

You maintain full control over your API usage, billing, and can set your own spending limits directly in your OpenAI account.

With BYOK, you pay OpenAI directly at their standard rates:

• Text conversations: ~$0.02-0.06 per conversation (varies by length and model)
• Voice conversations: ~$0.05-0.15 per minute

These costs depend on the model you choose (gpt-realtime-mini vs gpt-4o-realtime) and conversation length. There's no markup from KitePin.

Yes! Pay once and access KitePin forever. This includes all current features and future updates. The only ongoing cost is your API usage, which you pay directly to OpenAI.

We're able to offer this because of the BYOK model — you cover the AI compute costs, we cover the platform.

Plans differ primarily by the number of agents and access to advanced features:

• Solo ($99): 1 agent, all core features
• Starter ($169): 2 agents, all core features
• Pro ($399): 5 agents, team members
• Business ($599): 10 agents, function calling, custom domain
• Agency ($999): 20 agents, all features

All plans include voice + text support, full customization, knowledge base, and analytics.

Yes! You can upgrade to a higher plan at any time. You'll only pay the difference between your current plan and the new one. Contact support to initiate an upgrade.

Absolutely. We offer a 14-day money-back guarantee. If KitePin isn't right for you within the first 14 days, contact us for a full refund — no questions asked.

Yes! White labeling is available for enterprise customers at $4,999. This includes complete brand removal, custom domain support, and dedicated onboarding. Contact us to discuss your requirements.

Creating an agent is simple:

• Go to your Agents dashboard and click "Add Agent"
• Name your agent and describe its goal
• Select your industry (helps with pre-configured settings)
• Configure the 6 tabs: Agent, Appearance, Behaviour, Knowledge, Automation, and Install
• Add your OpenAI API key
• Save and your agent is ready!

Each agent has 6 configuration tabs:

• Agent: Domain, language, first message, context/prompt, OpenAI key and model
• Appearance: Widget placement, avatar, colors, logo, and label customization
• Behaviour: Text-only mode, timeouts, daily limits, voice selection, conversation memory
• Knowledge: User variables, global knowledge, contextual and page-wise knowledge
• Automation: Summaries, email notifications, webhooks
• Install: Embed script code, custom domain setup

A good agent prompt (called "Context" in KitePin) should include:

• Role definition: Who the agent is (e.g., "You are a friendly customer service agent for...")
• Tone guidelines: How it should communicate (professional, casual, friendly)
• Boundaries: What it should NOT do (e.g., "Do not make up information")
• Specific instructions: How to handle common scenarios

Pro tip: Use the "Improve with AI" button to enhance your prompt automatically!

Yes! In your Agents list, click the actions menu (three dots) on any agent and select "Duplicate." This creates an exact copy that you can then customize for different purposes or domains.

KitePin supports OpenAI's realtime models:

• gpt-realtime-mini: Faster, more cost-effective — great for high-volume use
• gpt-4o-realtime: More capable, better for complex conversations

You can select your preferred model in the Agent configuration tab.

In the Agent tab, you'll find a "First message" field. This is the greeting your agent uses when a conversation starts. For example: "Hi! I'm here to help. What can I assist you with today?"

If you leave this empty, the agent will wait for the user to speak first.

Yes! Each agent has an Active/Inactive toggle. When inactive, the widget won't appear on your website. This is useful for:

• Testing agents before going live
• Temporarily disabling an agent during updates
• Managing multiple agents for different campaigns

Function calling lets your agent connect to external APIs or databases during conversations. For example, a restaurant agent could call a function to fetch menu items, check availability, or place orders.

You define functions with:

• Name and description
• Parameters (with types: String, Integer, Boolean)
• Required vs optional parameters

This feature is available on Business and Agency plans.

Yes! KitePin supports both AI chat and AI voice agents. Your customers can type or speak naturally, and your AI agent responds in real-time with human-like voice or text responses.

Voice conversations use OpenAI's realtime voice models for natural, low-latency interactions.

Absolutely! Toggle "Text Only Mode" in the Behaviour tab to disable voice and run a pure text chat widget. This is ideal for:

• Users who prefer typing
• Environments where audio isn't appropriate
• Reducing API costs (text is cheaper than voice)

KitePin offers multiple voice options from OpenAI, including:

• Alloy: Neutral, crisp, general-purpose
• And other voices with different tones and characteristics

Select your preferred voice in the Behaviour tab under "Voice Selection."

AI Turn Detection determines when the user has finished speaking and it's the agent's turn to respond. When enabled, the AI automatically detects natural pauses in conversation rather than using fixed timing.

This creates more natural, human-like conversations — especially important for voice interactions.

Yes! In the Behaviour tab, you can set:

• Call timeout (silence): End call if user doesn't respond for X seconds
• Max conversation duration: Maximum seconds a conversation can last
• Daily call limit: Maximum number of agent calls per day

These help control costs and prevent abuse.

KitePin supports multiple languages including English, Spanish, French, Hindi, and more. You can:

• Set a default language for your agent
• Customize all widget labels in any language
• Let the AI respond in the language the customer uses

The underlying OpenAI models support 50+ languages for natural conversations.

The Global Knowledge Base is information that your agent can access across all conversations. This is where you add:

• Company information and FAQs
• Product details and pricing
• Policies and procedures
• Any static information the agent should know

Think of it as your agent's permanent memory about your business.

When Contextual Knowledge is enabled, your agent automatically reads and understands the content of the page where it's embedded. This means:

• On a product page, it knows that product's details
• On a pricing page, it can answer pricing questions
• On a blog post, it understands the article content

Toggle this in the Knowledge tab with "Use Contextual Knowledge."

Page-wise Knowledge lets you define specific information for specific URLs. For example:

• /pricing → Detailed pricing FAQs and comparison info
• /product/widget → Technical specs for that product
• /contact → Support hours and contact details

This gives you granular control over what your agent knows on each page.

User Variables let you pass dynamic information from your website to the agent. Common examples:

• username – Personalize greetings ("Hi John!")
• plan – Know if user is on free or paid plan
• geoloc – Location-aware responses
• cart_total – Help with checkout questions

Define variables in the Knowledge tab and pass them via the embed script.

Yes, if you enable Conversation Memory in the Behaviour tab. When enabled:

• The agent remembers past interactions with returning users
• Context carries over between sessions
• Provides more personalized experiences

Note: This uses more tokens per conversation. You can control data retention period separately.

You can position the widget in multiple locations:

• Bottom Right – Most common, standard chat position
• Bottom Left – Alternative corner placement
• Top Right – Above-the-fold visibility
• Top Left – Alternative top placement

Configure this in the Appearance tab under "Placement."

Yes! You have several avatar options:

• Animation: Choose from themes like Royal Aqua, Sunset Glow, Forest Mist, Ocean Wave
• Custom Image: Upload your own avatar image
• None: No avatar displayed

For animations, you can also customize the two gradient colors to match your brand.

Absolutely! In the Appearance tab, you can customize:

• Animation Colors: Color 1 and Color 2 for gradient animations
• Widget Color: Main widget background/accent
• Buttons Color: Call-to-action button colors

Use the color pickers to match your brand exactly.

Yes! Every text label in the widget is customizable:

• Main Label (e.g., "Need Help?")
• Call / New Call buttons
• Accept / Dismiss Terms
• Status messages (Listening, Speaking, Connecting)
• Input placeholder text
• Conversation ended messages
• Error messages

Find these under the "Language" section in the Appearance tab.

Yes! You can upload your company logo in the Appearance tab. The logo appears in the widget header, reinforcing your brand identity during conversations.

Yes! Enable "Show Terms" in the Behaviour tab. Users must accept your terms before starting a conversation. You can customize:

• Terms header text
• Accept button label
• Dismiss button label

This is important for compliance in many industries.

When enabled, KitePin automatically generates an AI summary after each conversation ends. Summaries include:

• Main purpose of the conversation
• Key points discussed
• Any action items or outcomes

You can customize the summary prompt to focus on what matters most to your business.

Yes! Enable "Send Alert" in the Automation tab to receive email notifications when conversations end. You'll get:

• Conversation summary
• Key details and outcomes
• Link to full transcript

Enter your email address and you'll stay informed without checking the dashboard constantly.

Webhooks let you send conversation data to external systems automatically. Configure:

• Webhook URL: Your endpoint that receives the data
• JSON Structure: Define what data to extract and send
• Prompt: Instructions for AI to extract specific information

Use webhooks to update your CRM, trigger Zapier workflows, create support tickets, and more.

Yes! Use webhooks to integrate with any CRM that accepts webhook data:

• Salesforce, HubSpot, Pipedrive
• Zapier (connects to 5000+ apps)
• Custom CRM systems via API

Define your JSON payload structure and KitePin will send lead data automatically after each conversation.

You control what data gets extracted. Common examples:

• Customer name and contact info
• Product interests or order details
• Questions asked and issues raised
• Lead qualification scores
• Appointment preferences

Write a prompt that tells the AI exactly what to look for and structure in the JSON output.

The KitePin dashboard provides comprehensive analytics:

• Total Users: Unique visitors who interacted with your agent
• Active Users: Currently engaged users
• Total Conversations: Number of chat sessions
• Average Duration: How long conversations typically last
• Total Duration: Cumulative conversation time

Filter by date range and specific agents to drill down into the data.

Yes! The Conversations section shows all chat sessions with:

• Full transcript of every message
• Session name and ID
• Timestamp and duration
• Session variables captured

Search by date range, agent, or keywords to find specific conversations.

The Usage Statistics table shows token consumption per session:

• Text tokens: Input, cached, output, and total
• Audio tokens: Input, cached, output, and total
• Combined total: Overall token usage

This helps you estimate costs since you pay OpenAI directly based on token usage.

Yes! You can download your analytics data as CSV files. This is useful for:

• Creating custom reports
• Importing into spreadsheets or BI tools
• Archiving conversation data
• Sharing insights with your team

Look for the "Download CSV" button in the dashboard.

Yes! Plans with team member support (Pro and above) let you invite colleagues. Go to Team Members and click "Add Team Member" to invite someone via email.

Each team member gets their own login credentials.

You can control access per agent with three permission levels:

• Dashboard: View analytics and metrics for the agent
• Modify: Edit agent settings and configuration
• Conversations: View and manage conversation logs

Assign different permissions for each agent to each team member.

Yes! Each team member has an "Is Active" toggle. Deactivating a member:

• Revokes their access immediately
• Preserves their account for potential reactivation
• Keeps their activity history intact

Use the "Show Active" filter to view only active team members.

Go to User Profile to update your account:

• Profile tab: Photo (max 2MB), name, email, phone
• Reset Password tab: Change your password securely

Changing your email requires confirming your current password.

Yes! Your API key is encrypted and stored securely. It's only used to make calls to OpenAI on your behalf and is never exposed to end users or third parties.

For additional security, you can set usage limits directly in your OpenAI account.

You control data retention in the Behaviour tab. Options include:

• 30 Days
• 60 Days
• 90 Days
• Custom periods

After the retention period, conversation logs are automatically deleted.

Yes! Toggle "Keep Logs" off in the Behaviour tab if you don't want to store conversation transcripts. This is useful for:

• Privacy-sensitive industries
• Compliance requirements
• Minimizing data storage

Note: Without logs, you won't have access to conversation history or analytics.

Yes! Business and Agency plans include custom domain support. Instead of go.kitepin.com, your agent can run on your own domain like chat.yourbrand.com.

Setup requires adding a CNAME record in your DNS settings. Instructions are provided in the Install tab.

KitePin provides tools to help with GDPR compliance:

• Configurable data retention periods
• Terms & Conditions display before chat
• Ability to disable logging entirely
• Data export capabilities

However, you're responsible for ensuring your specific implementation meets your legal requirements.

Check these common issues:

• Agent is inactive: Make sure the Active toggle is ON
• Wrong domain: Verify the domain in Agent settings matches your website
• Script not added: Ensure the embed script is in your HTML
• Caching: Clear your browser cache and try again
• Script placement: The script should be before the closing </body> tag

Voice issues are usually caused by:

• Text Only Mode enabled: Check Behaviour tab and disable it
• Browser permissions: User must allow microphone access
• HTTPS required: Voice only works on secure (https://) connections
• Browser compatibility: Use Chrome, Edge, or Safari for best results

Improve response accuracy by:

• Refine your prompt: Add clearer instructions and boundaries
• Add knowledge: Include more information in Global Knowledge Base
• Use constraints: Tell the agent what NOT to do (e.g., "Don't make up information")
• Try a better model: gpt-4o-realtime is more capable than gpt-realtime-mini
• Use "Improve with AI": Let AI enhance your prompt

If you're seeing API errors:

• Check your API key: Make sure it's valid and has sufficient credits
• OpenAI status: Check status.openai.com for outages
• Rate limits: You may have hit your OpenAI usage limits
• Daily call limit: Check if you've exceeded your KitePin daily limit setting

If issues persist, contact support with your agent ID and error messages.