Skip to main content
A Chat Agent is an AI-powered conversational assistant that you create and deploy to interact with your customers across multiple channels. Each Chat Agent is backed by a large language model (LLM) and can be trained on your specific business knowledge to answer questions, qualify leads, schedule meetings, and more — all without human intervention.

How It Works

When a visitor sends a message — whether through your website widget, WhatsApp, email, or another channel — the Chat Agent:
  1. Receives the message from the connected channel.
  2. Retrieves relevant context from your knowledge base (web pages, documents, FAQs).
  3. Generates a response using the LLM, following the instructions and tone you configured.
  4. Sends the reply back to the visitor on the same channel.
If the conversation requires a human, the agent can escalate to a live team member in your Inbox.

Supported Channels

Each Chat Agent is configured for a specific channel. You can create multiple agents for different channels.
ChannelDescription
WebsiteEmbedded chat widget on your website. Includes full appearance customization, bubble messages, and visitor tracking.
EmailResponds to inbound emails via Mailgun or Gmail. Supports thread-based replies and BCC configuration.
SMSHandles text message conversations via Twilio.
WhatsAppConnects to WhatsApp Business API for automated messaging with template support.
ZaloIntegrates with Zalo chat.
Zalo OAConnects to Zalo Official Account for business messaging.
MessengerResponds to Facebook Messenger conversations via your Facebook Page.
Channel types available when creating a new Chat Agent

Agent Types (Engine Configuration)

When creating a Chat Agent, you choose one of two engine types:

Single Prompt (Simple Prompt)

Best for most use cases. You provide a set of instructions (objectives) that tell the agent how to behave in all conversations. The agent uses these instructions along with your knowledge base to generate responses. Use when:
  • Your conversations follow a general pattern.
  • You want a quick setup with natural, flexible responses.
  • Your agent handles support, FAQ, or general lead qualification.

Conversational Flow

A node-based visual designer where you build a structured conversation tree. Each node represents a stage in the conversation with its own instructions and transition rules. Use when:
  • You need a strict, step-by-step conversation path.
  • Different branches require completely different behavior.
  • You’re building a qualification funnel with specific routing logic.
Flow editor showing a multi-node conversation design

Key Capabilities

Every Chat Agent can be configured with:
  • Knowledge Base — Upload documents, crawl web pages, and add FAQs so the agent answers from your content.
  • Lead Capture — Collect contact information (name, email, phone) during conversations.
  • Meeting Scheduling — Let visitors book meetings directly from the chat.
  • Human Escalation — Automatically hand off to a live agent when the AI can’t handle the request.
  • Tools & Integrations — Connect to external APIs, CRMs, and trigger custom actions.
  • Message Approval — Optionally require human approval before the agent sends a message.
  • Analytics — Track conversation metrics, feedback, and agent performance.

What’s Next