n8n Workflows Are About to Get a LOT Smarter with Human Feedback

n8nRecent📅 2025-02-12

Watch on YouTube ↗

Multi-Agent Architecture

┌─────────────────────────────────────────────────────────────────────┐
│                     MULTI-AGENT SYSTEM                              │
├─────────────────────────────────────────────────────────────────────┤
│                                                                     │
│                     ┌───────────────────┐                           │
│                     │  ORCHESTRATOR     │                           │
│                     │  (Manager Agent)  │                           │
│                     └─────────┬─────────┘                           │
│                               │                                     │
│              ┌────────────────┼────────────────┐                    │
│              │                │                │                    │
│              ▼                ▼                ▼                    │
│   ┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐      │
│   │ Research Agent  │ │  Writer Agent   │ │ Reviewer Agent  │      │
│   │                 │ │                 │ │                 │      │
│   │  • Web Search   │ │  • Compose      │ │  • Validate     │      │
│   │  • Data Gather  │ │  • Format       │ │  • Improve      │      │
│   └────────┬────────┘ └────────┬────────┘ └────────┬────────┘      │
│            │                   │                   │                │
│            ▼                   ▼                   ▼                │
│   ┌─────────────────────────────────────────────────────────┐      │
│   │                    SHARED TOOLS                          │      │
│   │  [Web API]  [File System]  [Database]  [Code Executor]  │      │
│   └─────────────────────────────────────────────────────────┘      │
│                                                                     │
└─────────────────────────────────────────────────────────────────────┘

Agent Workflow

• Core concepts explained
• Step-by-step implementation
• Practical examples

Transcript

[00:00] I just built a multi-agent workflow in n8n that includes human feedback the problem with normal workflows is that each note will execute in sequence without giving the user a chance to provide any feedback or to request changes so that means that for a workflow like this that generates stories you have to wait for the entire workflow to complete before you can see what the premise is what the characters are the story outline Etc and if the agent got it wrong at any point during this process you have to restart the

[00:31] entire thing again this going to be a waste of time and of course you're paying for the tokens but with the introduction of human in the loop in n8n we can now create workflows where the agent will produce some output and it will then wait for the user to provide feedback and if the user requests any changes those changes will be passed back to the agent and again the agent will ask you for further input and once the user approves the flow we will move on to the next phase in our workflow let's St demonstrate this so I'm going to ask a story agent to write a harder

[01:03] story about a group of teenagers and a cabin in the woods in n8n we can see this human in the loop note is currently waiting for our feedback and back in telegram we can see the premise of our story and of course we can respond to it so if we want to we could change the premise or we can simply click on continue and back in n8n the Character Creator is now running and waiting for our feedback looking at telegram we can see that we got 1 two and three characters so let's say add a fourth character called slime Jones and ATN is

[01:35] currently generating this new character sheet and now it's waiting for human feedback we now get all those characters including the latest character so let's simply continue and in at end the story outliner now executed and is awaiting our feedback and we can also see the outline in telegram let's change the ending to be a cliffhanger and the ending was now changed to a cliffhanger ending finally let's approve this and now the final step in the process is being executed which is the story writer and this agent will simply generate the

[02:05] story and send it to me via email and now I have my final story including all of the revisions check out my channel if you want to learn how to build this workflow yourself