Skip to main content
A workflow is a visual conversation flow that defines exactly how your AI Agent interacts with users — step by step. Instead of relying on open-ended AI responses alone, you design a graph of nodes (steps) connected by paths, giving you full control over what the agent says, what information it collects, which external systems it calls, and when the conversation ends. Workflows are the foundation of the Workflow Agent type — the advanced agent that follows a designed flow. If you want a simpler Knowledge Agent that answers freely from its knowledge base and instructions, see Agent Types for a comparison of the two agent types.

​
What a workflow looks like

A workflow is a directed graph:
  • Nodes are the individual steps — sending a message, asking a question, branching on a condition, calling a webhook, and more.
  • Edges are the paths that connect nodes, forming the conversation flow from start to finish.
  • Every workflow has exactly one Start node (where every conversation begins) and one or more End nodes (where a conversation path concludes).

​
Where workflows live

Workflows are attached to a Workflow Agent. You can find them on the Chat Workflows tab when you open a Workflow Agent in the dashboard. One agent can have many workflows, but only one workflow can be active for live conversations at a time. Activating a workflow automatically deactivates the previously active one; every other workflow stays inactive (whether draft or published) until you activate it. If a Workflow Agent has no active workflow — for example, immediately after creation or after deactivating all workflows — it falls back to standard AI answering using its instructions and knowledge base, behaving like a Knowledge Agent.
Workflows list

​
Next steps