Settings on the General tab apply across all messaging channels unless overridden in a channel’s configuration. Anything set on a channel tab overrides the matching General setting for that channel only.
Channels
The Channels card is where messaging channels are switched on and off. Each channel has its own row with a status tag and toggle. A channel that needs setup before it can be activated shows Not configured with a prompt to complete setup; an activated channel shows Active. Deactivating a channel stops it sending and receiving messages, but keeps its configuration so it can be reactivated at any time. You are asked to confirm before a channel is deactivated.Language Model
Choose the primary LLM for messaging. All messaging channels use the same model. We recommend PolyAI’s Raven, designed for conversational AI and tuned for both voice and chat, with support for custom styling, reasoning and 24+ languages.Channel Behaviour
Tailor tone, format and etiquette for your messaging agent with the Messaging channel behaviour prompt. This optional multiline prompt applies across messaging channels, and sits between the agent’s core behaviour and any channel-specific style prompt.Prompts are applied in this order: channel style prompt > messaging channel behaviour > agent behaviour. Use this prompt for style shared across messaging channels, a channel tab’s style prompt for channel-specific style, and the Agent page to change what the agent knows or does.
Greeting
Set the Default greeting, the opening message used across messaging channels. Keep it short and welcoming, so it invites the first message rather than obscuring the conversation. Each channel tab can override this with its own greeting. Example: “Hello, how can I help?”Safety Filters
Configure filters to block harmful user inputs and generated outputs across four categories: violence, hate, sexual content and self-harm. Each category has its own toggle and a severity slider from Lenient to Strict; stricter settings increase protection against harmful content. Channel tabs can override these defaults per channel. See Safety filters for the full reference on categories, severity levels, and how filters interact with Guardrails. The jailbreak attack filter is always enabled and cannot be turned off. It blocks user inputs that attempt to bypass safety measures.Good to Know
- Changes apply on save. Messaging configuration takes effect immediately when saved, unlike settings that go through environment promotion.
- Markdown is supported. Chat messages render bold, italics, lists, links and code blocks across testing and review panels.
Branching behavior
Chat configuration changes take effect immediately when saved. Unlike other project settings, webchat does not follow the environment branching system, there is no Sandbox, Pre-release, or Live promotion step. Use the Test Agent Chat panel to verify changes before saving. Markdown support
Chat messages support formatted markdown rendering:- Bold and italics
- Bulleted and numbered lists
- Links
- Code blocks
Next steps
After configuring chat settings:- Configure the widget - Style and customize the chat widget appearance
- Deploy to your website - Generate and embed the script tag
Configure Each Channel
Channel-specific settings live on their own tabs and pages:Web chat
Widget configuration, disclaimers and privacy policies, and website deployment.
SMS
SMS behaviour, greetings and templates.