nœuds / builtin / flow
Respond to Chat
Send a message to the chat that started the run
Ce contenu n’est pas encore disponible dans votre langue.
Send a message back to the chat that started the run. Use it to control exactly what the chat shows, or to ask a follow-up question and wait for the answer.
What it does
Section titled “What it does”In a workflow started by a Chat Trigger, the reply in the chat comes from a reply step: a Respond to Chat step, or else the last AI step on the trigger’s path. The reply streams into the chat as it is written.
Respond to Chat sends its Message. Then, depending on After sending:
- Carry on: the run continues straight away.
- Wait for the reply: the run pauses until the person writes the next chat message. That message becomes this step’s output, and the run continues from here.
With Add to chat memory on (the default), the message is also stored in the Chat Memory connected to the Chat Trigger, so an agent later in the chat remembers it.
When no chat is open for the run (it was started another way), the message is only stored in the chat memory and the run log says so.
When the message streams
Section titled “When the message streams”In an open chat, the message streams (words appear as they are written, with Draft reply is writing…) only when it is exactly the reply of the AI step linked straight to this step, for example {{ $('Draft reply').response }} or {{ $input.response }}. Otherwise it is sent whole once it is ready: when a step in between (Guardrails, Edit Fields…) can change the text, or when the message is more than one AI step’s reply. The step’s settings say which (Streams into the open chat as Draft reply writes., or Sent whole: … with the reason). Nothing else changes: the run and the stored message are the same either way.
When to use it
Section titled “When to use it”- A fixed reply: “Got it, your ticket is #1234.”
- Combining several results into one message, after steps that are not AI steps.
- A short back-and-forth: ask for a missing detail, wait, then use the answer.
Inputs and settings
Section titled “Inputs and settings”| Réglage | Type | Ce que cela fait |
|---|---|---|
| Message | Long text | |
| After sending | Choice | Options: Carry on, Wait for the reply. Par défaut : Carry on |
| Add to chat memory | Toggle | Also store the message in the Chat Memory connected to the Chat Trigger. Par défaut : On |
Write the Message with data from earlier steps, for example Done! I added {{ $input.count }} rows.
Outputs
Section titled “Outputs”-
Output
Each output item has these fields:
| Field | Type | What it holds |
|---|---|---|
| message | Text | The message sent to the chat |
| chatInput | Text | The user’s answer (Wait for the reply) |
{ "message": "…", "chatInput": "…"}{ "message": "Which project?", "chatInput": "Website redesign" }chatInput is set only with Wait for the reply: it holds the person’s answer.
Dependencies and credentials
Section titled “Dependencies and credentials”None. This node needs no connection and has no slots for other nodes.
Example workflow
Section titled “Example workflow”- Chat Trigger receives “Add a task”.
- Respond to Chat: “Which project?”, Wait for the reply.
- Todoist creates the task in
{{ $input.chatInput }}. - Respond to Chat:
Added to {{ $('Respond to Chat').chatInput }}.
Common issues
Section titled “Common issues”- The chat shows the agent’s text instead: the reply step is the Respond to Chat step only when the run reaches it. Check that it sits on the path the run takes.
- Stop in the chat: stopping keeps the part of the reply already written and marks it as stopped.
Related nodes
Section titled “Related nodes”- Chat Trigger: starts the run from a chat.
- Tools Agent: an AI reply with tools.
- Wait For Approval: pause for a decision instead of a message.