Configure your chatbot’s behavior, AI model, notifications, and advanced options.
The Settings section lets you configure how your chatbot behaves, which AI model it uses, rate limiting, domain restrictions, and more.
In your chatbot dashboard, click Configuration in the sidebar.
Click on the Settings tab.
The Basic tab contains core configuration options for your chatbot.
Your unique chatbot identifier. You’ll need this ID to connect to your agent using Zapier and other integrations.
The fallback response when your agent cannot find a relevant answer in its training data. You can customize this message or reset it to the default.
Select the language your agent uses when replying. Set to Auto to automatically detect and respond in the visitor’s language.
Choose how often you receive email summaries with your chatbot’s top answered and unanswered questions. Options include:
When enabled, your agent can display images and videos directly in messages. When disabled, the agent will provide a link to the media instead.
Allow visitors to provide feedback on responses by selecting thumbs up or thumbs down after each message. This helps you identify which responses need improvement.
Allow the agent to understand the current page content when responding to questions. The agent will analyze the page where the chat is embedded to provide more relevant answers.
Allow visitors to upload files to the agent with their messages. Your agent can process these files and use them to answer questions.
Identify the country of your visitors when they interact with your chatbot. When enabled, the visitor’s country is determined using network-level headers and displayed across your dashboard, inbox, and contacts pages.
Privacy-first by design:
When enabled, country data appears in:
Visitor Geolocation is available on all plans.
Configure the AI model powering your chatbot and customize its behavior.
Select which AI model your agent uses to generate answers and perform actions. Available models include various OpenAI GPT versions and other providers.
Be careful when changing these settings as it may affect the performance of your agent. Test your agent after making changes.
Customize the behavior, tone, and personality of your agent. The default instructions define:
When enabled, country data appears in:
Control the randomness of responses on a scale from 0 to 1:
For support chatbots, lower values (0-0.3) are typically recommended for accuracy.
Display a notice message that users must accept before starting a chat. This is useful for terms and conditions, privacy notices, or other important information.
Toggle this on to display a notice message before users can start chatting.
Enter your notice message (up to 250 characters). This message supports basic markdown formatting for links and emphasis.
The Settings section includes several other tabs with dedicated documentation:
Configure lead capture forms and fields.
Set up human handoff and support notifications.
Multi-language support and message localization.
Connect to third-party services and platforms.
The Advanced tab contains voice features, restriction levels, rate limiting, and domain restrictions.
Enable voice input and audio responses for your chatbot. For detailed configuration, see Voice & TTS.
Control how strictly your agent adheres to its training data:
Max (Default)
Agent will only answer based on the training data
Low
Agent will attempt to use training data, but can also use some general ChatGPT knowledge
None
No restrictions, agent will answer any query
Prompt Guard pre-screens every visitor message using AI classification to detect prompt injection and jailbreak attempts before they reach your chatbot. When a malicious message is detected, the chatbot responds with the default answer instead of processing the harmful prompt.
Enable Prompt Guard by toggling it on in the Advanced settings tab.
Prompt Guard adds a small amount of latency to each message as it runs the classification check. This is typically imperceptible to visitors.
Prompt Guard is available on Startup plans and above.
PII Redaction automatically masks sensitive identifiers in visitor messages before they are stored or sent to the AI model. Detected values are replaced with placeholders so the conversation stays readable while the sensitive data never reaches the AI provider or your conversation history.
The following identifiers are detected and masked:
National ID numbers (PESEL)
[PESEL]
Payment card numbers
[CARD]
Bank account numbers (IBAN)
[IBAN]
BLIK payment codes
[BLIK]
Passwords
[REDACTED]
Enable PII Redaction by toggling it on in the Advanced settings tab. Every redaction is also recorded in your audit log as a “PII Redaction” entry showing how many values were masked and their types, never the values themselves.
Detection uses pattern matching with checksum validation, so it adds no latency to responses. Values that are unusually formatted (for example, digits separated by words) may not be detected, so treat PII Redaction as a strong safeguard rather than an absolute guarantee.
PII Redaction is available on Growth plans and above.
Prevent abuse by limiting the number of messages a visitor can send within a 5-minute window.
Control where your chatbot can be loaded:
Control which pages your chatbot appears on by filtering based on URL paths.
Page URL filtering is applied client-side. It is bypassed in preview mode so you can always test your chatbot.
Block specific IP addresses or ranges from interacting with your chatbot.
IP address blocking is applied server-side. It is bypassed in preview mode so you can always test your chatbot.
Actions in this section are irreversible and cannot be undone.
Permanently delete your chatbot and all its data. This removes:
Go to Configuration → Settings → Advanced.
Find the Danger Zone section at the bottom.
Click the red Delete Agent button and confirm.
