Skip to main content
Conversation nodes let you build AI-powered, multi-turn conversations with people across any communication channel. They handle everything from opening a channel and routing messages, to running AI logic and sending replies.

Why conversation nodes exist

Most workflow automations are one-shot: a trigger fires, actions run, and the workflow ends. But many real-world tasks require back-and-forth dialogue — qualifying a lead, resolving a support ticket, collecting form data, or getting approval. Conversation nodes give your workflows the ability to hold stateful, multi-turn conversations with people while the rest of the workflow continues running.

How they work together

Conversation nodes follow a layered architecture. Each layer has a clear responsibility:
LayerNodesPurpose
ChannelChannel Dispatcher, Ask a PersonOpen a communication channel and manage message polling
ActorCommunication Actor, LLM ActorProcess conversations with AI
MessagingSend Message, Clear ConversationSend outbound messages and manage conversation data

Two architecture patterns

Separate the channel from the AI logic. This gives you more control — you can route different conversations to different actors, chain actors together, or add logic between them.

All-in-one pattern

The Ask a Person combines channel + AI in a single node. Simpler to set up for straightforward use cases.

Supported channels

All conversation nodes support the same set of communication channels:
ChannelInitiator modeResponder mode
Web WidgetYes
GmailYesYes
OutlookYesYes
SlackYesYes
Slack (OAuth)YesYes
Slackbot (built-in)Yes
SMS (Twilio)YesYes
WhatsApp BusinessYes
Microsoft TeamsYes
Initiator mode means the workflow starts the conversation (e.g., sends the first email). Responder mode means the workflow waits for someone to reach out.

Choosing the right nodes

Use Channel Dispatcher + LLM Actor. The LLM Actor gives you direct control over the model, system prompt, and tools without needing to configure a skillset.
Use Channel Dispatcher + Communication Actor. The Communication Actor uses the Skillset Engine, which is designed for collecting structured data through natural conversation.
Use Ask a Person. It combines channel management and AI logic in a single node — good for simple workflows where you don’t need to route between different actors.
Use Channel Dispatcher + multiple Communication Actors or LLM Actors with a Condition or Multi-Condition node to route based on message content, user data, or AI classification.
Use Send Message. It sends a message through an existing conversation opened by a Channel Dispatcher or Ask a Person.

All conversation nodes

Channel Dispatcher

Open a communication channel and emit events when messages arrive.

Ask a Person

All-in-one channel + AI conversation manager.

Communication Actor

Handle conversations using the Skillset Engine for structured data collection.

LLM Actor

Handle conversations using an LLM with tool calling.

Send Message

Send a message through an active conversation.

Clear Conversation

Reset conversation messages, metadata, or state.