Skip to main content
The Communication Actor processes conversations using the Skillset Engine. It connects downstream of a Channel Dispatcher and handles the AI side of the conversation — collecting structured data, answering questions from a knowledge base, and managing conversation flow. Think of it as the “brain” assigned to a conversation channel. The Channel Dispatcher is the ears and mouth; the Communication Actor decides what to say.

​
When to use

  • Structured data collection — gather specific fields (name, company, budget) through natural conversation
  • Knowledge-powered Q&A — answer questions using your knowledge base documents
  • Form-like workflows — replace web forms with conversational data collection
  • Custom actors — use pre-built or custom-configured skillset actors
  • Conversation continuation — keep talking after data collection for follow-up questions
If you need direct LLM control with a system prompt and tools instead of the Skillset Engine, use LLM Actor instead.

​
How it works

When a conversation_updated event fires from the Channel Dispatcher, the Communication Actor:
  1. Finds its upstream Channel Dispatcher by traversing the workflow graph
  2. Gets the conversation ID from the event data
  3. Starts the skillset if this is the first message (or reconnects if already started)
  4. Proxies the user’s message to the Skillset Engine for processing
  5. Sets itself as the active actor so subsequent messages route directly to it

​
Actor lifecycle

The Communication Actor manages its own lifecycle:
  • Start: When first triggered, it starts the skillset for this conversation
  • Re-entry: On subsequent messages, it proxies the new message to the already-running skillset
  • Handoff: If another actor was previously active, it replays the last user message so the new actor has context
  • End: When the skillset emits conversationEnded or transfer_to_human, the actor clears itself and fires the corresponding workflow event

​
Custom tools

You can connect Custom Tool nodes to the Communication Actor’s tools handle. These become LLM tools that the skillset can call during conversation: Custom tools let the AI take actions during conversation — look up data, run calculations, or trigger external systems — without leaving the conversation flow.

​
Example: lead qualification with knowledge base

Qualify leads while answering product questions from your docs:
1

Set up the channel

Add a Channel Dispatcher with your preferred channel (e.g., Web Widget in responder mode).
2

Add a Communication Actor

Connect a Communication Actor to the Channel Dispatcher’s conversation_updated handle. Select your Lead Qualification skillset.
3

Attach a knowledge base

In the actor settings, select a knowledge base. Enable Continue conversation so the AI can answer follow-up questions after collecting all required fields.
4

Handle results

Connect the conversationEnded handle to downstream nodes that process the collected data (e.g., create a CRM contact, send a notification).

​
Example: multi-actor routing

Route conversations to different actors based on intent:

​
Example: actor handoff chain

Chain actors together — when one finishes, the next takes over the same conversation: When the second actor starts, it replays the user’s last message so it can pick up the conversation naturally.

​
Suppressing completion messages

When a Communication Actor has a conversationEnded or transfer_to_human edge connected to another node, it automatically suppresses the skillset’s default completion message. This prevents the user from seeing a “goodbye” message before the workflow continues to the next step.

​
Settings

​
skillsetId
string
required
The skillset to use for conversation handling. Can be a built-in skillset or a custom actor created in your workspace.
​
skillsetConfiguration
object
Skillset-specific settings. Common options include:
  • formFields — fields to collect during conversation
  • knowledgeBaseId — knowledge base for Q&A capabilities
  • continueConversation — keep the conversation active after data collection (requires a knowledge base)
  • inputVariables — pre-populated variable values
All values support variable interpolation.
​
enableVisualizer
boolean
default:"false"
Enable Magic Window for visual content during conversation.

​
Outputs

​
conversationId
string
The ID of the conversation this actor processed.
​
result
object
Result data from the skillset, including collected fields.
​
success
boolean
Whether the skillset processed the conversation successfully.

​
Events emitted

EventWhen it fires
conversationEndedThe skillset completed its task and ended the conversation
transfer_to_humanThe AI decided the conversation needs a human agent

Channel Dispatcher

The channel node that feeds conversations to this actor.

LLM Actor

Alternative actor that uses raw LLM + tools instead of the Skillset Engine.

Skillset Engine

Learn how skillsets work under the hood.

Clear Conversation

Reset conversation state between actor handoffs.