I’m new to n8n and experimenting with the AI Agent Node. I was wondering if it’s possible to create agents dynamically (for example, during a workflow run) and then have these agents communicate or collaborate with each other.
As an example:
Imagine a “department” with different “employees” or “roles,” each acting as an agent. They all receive the same question and then work together by exchanging messages until they agree on a final answer.
A few questions I have:
Can new agents be created automatically without setting them up one by one in the UI?
Is it possible for these agents to exchange information directly, like having a conversation or working together on a task?
How would you handle memory or context so they can “remember” what was said or done?
I’m not sure what the best approach is, so any explanations, examples, or resources would be really helpful.
I’m not sure if n8n is friendly with real-time applications. But I would suggest to:
Create different workflows where each workflow has an AI Agent with a different model or access of tools,
Then utilize these workflows using ‘Call N8N Workflow Tool’ in your main workflow.
This would allow your main AI Agent to receive opinions from other AI Agents. But it’s a one-timer. Like a group of people expressing themselves to the boss for once and then letting the boss decide a final answer.
Thanks for the idea! That approach makes sense for getting multiple agents’ input, but it doesn’t yet solve my main challenge. I still need a mechanism to create, configure, and clean up agents at runtime, rather than having them pre-built in separate workflows.
Just wanted to follow up on this — I’m still exploring ways to dynamically create and manage AI agents within n8n, especially to enable them to collaborate and maintain some memory/context during a workflow.
Has anyone tried something similar or found creative workarounds? I’d really appreciate any tips, examples, or even just a direction to explore further.