Learn what a multi-agent system is and discover the powerful benefits of having specialized AI agents work together! In this video, I'll explain the concept of multi-agent systems and why they're often more effective than single AI solutions. Then I'll walk you through a step-by-step example of creating a simple sub-tool workflow in n8n that acts as a specialized agent, which your main parent agent can call whenever it needs that specific functionality. Perfect for beginners - no coding required, just n8n's visual drag-and-drop interface. Whether you're new to AI automation or looking to build more sophisticated systems, this tutorial will show you how to create AI agents that can intelligently delegate tasks to specialized tools.

Hi, @Nate Herk could you share the links to flows like https://youtu.be/SvhYQSa-9m8?si=bIlqkLIlFfJIadBu and the other mcp YouTube video you made? Don't seem to find them. I would like to get better understanding of these flows in the way you demonstrate them in your videos Thanks! Jan

Hi, Please advise howto set multiple inputs for the AI Agent in n8n (webhook AND chat AND Telegram AND email) in 1 flow. Have not been able to figure it out yet. Thanks, J

Subject: Enhanced Prompting Flow for Personal Assistant with @AI Integration Hi, I watched your video on Personal Assistant, and I’m excited about the new one on prompting—definitely going to check it out! I had a few thoughts regarding the integration of @AI and best practices for prompting within the Personal Assistant flow. Would you be able to share the updated flow with @AI included and offer insights on refining prompting strategies? Here are a few specific ideas I thought might enhance functionality: 1. **Email Responses**: Tailor responses to mirror the sender's tone, automatically setting the appropriate salutation and greeting. Integrating a vector database for contextual analysis could pull in relevant information from prior emails to enrich responses. 2. **Calendar Events**: Using a vector database could also help provide background when setting up agenda items. Once a user ID is retrieved, contacts (from Gmail or another source) could be queried to find additional context about attendees or previous interactions, allowing for a more contextually rich prompt. 3. **Personal Data Trust**: A major added value would be allowing users to trust the assistant with specific, personal information tied to their ID. For example, users could add licenses, passwords, or other private information, enabling the assistant to remember and provide context-based responses, much like the personalized responses we see with ChatGPT’s latest updates. This could vastly improve prompt relevance and response accuracy. 4. **Grounding the AI (Similar to Microsoft Copilot)**: Before responding to any prompt, the assistant could consult a vector database for user-specific details (e.g., Jan is a male of [age], tech-savvy, etc.). By grounding each interaction in the user’s data, responses can be much more precise and personalized. Would love to hear your thoughts on these ideas and any recommendations you have for refining the prompting flow! Kind regards, Jan

Hi, This video shows to use structured output parser and also some other tips. Master Multi-AI Agent Workflows in N8N | Unlock the Secrets to Advanced Automation Would appreciate your view/take on that. I do have concerns when publishing output to WordPress which is different than LinkedIn or X. Is the output parser then the way to go? I struggle with making Agent 2.0 working correctly (adding my own nodes like WordPress, LinkedIn post, Agenda etc). Formatting is always off, sometimes * appears, sometimes \n in output etc. @Nate Herk would love to see a version 3.0 of your AI Agent with all lessons learned (like @FromAI, type of AI agent -function -apparently the best?/conversational/tool-, prompting practises "you have these tools", correctly handling subflows -with again subflows since first agent should just handle like email to one subflow but that subflow could need again multiple tools) Thanks for your input! J