n8n just dropped native support for MCP servers. For a deeper discussion, let's chat here. I imagine there are about to be tons of YouTube videos on this topic, so today I break down what these nodes really allow us to do - without the hype. We’ll discuss how the two new MCP nodes work, and how to connect them to your Claude desktop setup. I’ll also cover the current limitations and where I think this is heading for the future of AI Agents in n8n.

I am building an n8n workflow that acts as a personal assistant using WhatsApp as the trigger. The main workflow (Personal Agent) processes incoming WhatsApp messages, filters for voice or text input, and includes an initial check to ensure the message is coming from my number rather than someone else’s. It then delegates tasks to various sub-agents (Calendar, Email, Contacts, etc.) via tool workflows. The Problem: After manually triggering the workflow via WhatsApp, the Personal Agent is being executed additional times automatically. This causes the Window Buffer Memory node to receive three or more inputs instead of the expected single item (this occurred before I even implemented the number-checking filter). Now, with the number-checking filter added, that node produces an error instead of the buffer memory. These extra inputs lead to various errors and unexpected behavior. Question: Has anyone experienced similar issues with multiple triggers from WhatsApp events in n8n? I would appreciate any insights or suggestions on how to resolve these issues. Thank you in advance! PS: I am using the cloud hosted version of n8n.

Hi all, I am probably overlooking the obvious but I am stuck trying to aggregate multiple runs from one node into one record in n8n I have a process that will search my email for relevant communication related to a specific search. Assuming there is more than one result, it will download all items and then summarize them one by one. The output could be anywhere between 2 and 50 records. In the next step, I want to load them all into an AI agent in one run, so the agent does not analyse them individually. The individual fields option in the aggregate function seemed to be the most obvious option, but it just passes them through one by one. I am using the N8N cloud hosted latest version. Appreciate any advise you can give.

Hello everyone, I wanted to know if someone could give me an approximation of the cost of the ChatGPT API, for exemple : the GPT-4 Mini model, to run a complete AI agent assistant ?

Let's get to know each other! Comment below sharing where you are in the world, a career goal you have, and something you like to do for fun. 😊