How to Create a Sequential Chat Flow with Multiple AI Agents?

[The_Bakery]

Hi there,

Looking to know if anyone has being cappable to use the Chat Trigger Node connected to more than one.

Hello everyone!
I’m trying to build a chat flow in n8n that involves multiple AI Agents (for example, 3 different AI Agent nodes), each of which should only be triggered after the user has completed a task or conversation with the previous agent.

What I’m Trying to Achieve

• I have a Chat Trigger node that receives user input.
• I want AI Agent 1 to handle the first step, then wait for the user to confirm or complete something.
• After that confirmation, AI Agent 2 should start, and only then AI Agent 3 should run.

The Problem

• Currently, when I connect the AI Agent nodes in sequence, they all fire automatically one after another. It seems like the next agent is being triggered immediately instead of waiting for a user confirmation.
• This leads to the agents “talking over each other” or receiving responses as if they were the user.

What I’ve Tried

1. Connecting the nodes in a simple linear sequence (AI Agent 1 → AI Agent 2 → AI Agent 3).
2. Attempting to insert an IF node to check if the user has typed something like “Ok, next step” to move on, but I might be missing something in my setup.
3. Looking into storing a “state” variable in Simple Memory, but I’m unsure how to make the flow pause and wait for the user’s next message.

What I Need Help With

• How do I structure my workflow so that AI Agent 2 only runs after the user explicitly finishes the conversation with AI Agent 1?
• Is there a recommended pattern for using Wait or IF nodes with Chat Trigger flows so that the next agent only fires upon a specific user response?
• Any best practices for managing “state” or “step completion” across multiple AI Agents in n8n?

[solomon]

I can think of two possible solutions:

1. Master agent with sub-workflows

One agent that decides who to call, depending on the context of the conversation. Other agents will be connected to it as tools, by using sub-workflows.

2. Database to store conversation state

Give the AI Agents the tool to insert/update data in a database and add a field for the status/step/stage of the conversation.

Once the conversations reaches a different stage, the AI can change the field in the database.

In your workflow you would have a simple SWITCH node that checks the stage of each session_id and routes the conversation for the correct agent.

You can also use Google Sheets to test that, if a database is not something you are used to work with.
.

If my reply answers your question, please remember to mark it as a solution.