I’m trying to follow the tutorial “n8n Developer Agent: Build Workflows with LLMs” (published October 15th on the n8n blog) and I’m stuck on one specific piece.
In the tutorial, there’s a node called “n8n Developer” that sits between the Chat Trigger and the tools (Brain/Memory/Developer Tool). I cannot find this node anywhere in my node search panel.
Here’s what I’m building:
∙ An AI agent that reads a PDF strategy document from Google Drive
∙ Uses an LLM to generate n8n workflow JSON
∙ Automatically creates those workflows via the n8n API
What I have set up:
∙ Chat Trigger
∙ Basic LLM Chain
∙ OpenRouter Chat Model (GPT-4.1 mini)
∙ Execute Workflow (sub-workflow)
∙ Google Drive (Download file)
∙ Create a Workflow node
My question:
Is the “n8n Developer” node the same as the AI Agent node, just renamed? Or is it a separate node I need to install? Could it be unavailable on the Starter plan?
I’m on Windows (not Mac) and on the n8n Starter plan on n8n Cloud.
Any help appreciated
Hey @Toshi
This “n8n Developer” node in the tutorial is actually the “AI Agent” node and is only renamed for easy understanding in the tutorial, not a node type you need to install.
Secondly, your starter plan does NOT restrict which nodes you use. For cloud Starter plans, community nodes are available.
Hope this helps!
Hi @Toshi
nice idea, your flow seems pretty much okay, but even the smartest models lack when creating complete workflows, i would recommend adding an MCP like:
And use AI agent to connect everything up, for now i would recommend testing how the output is actually coming before directly using n8n API for automated generation, consider using AI agent with MCP + Claude Models, GPTs are nice but claude models when it comes to creating n8n workflow JSONs are really smart. Let me know if you need any help with that.
Hi Anshul! Thank you so much for responding and for the detailed explanation. Yes, I would love your help with this!
To confirm what I have so far:
∙ Chat Trigger
∙ Basic LLM Chain (which I now know needs to be swapped for the AI Agent node)
∙ OpenRouter Chat Model (GPT-4.1 mini)
∙ Google Drive (downloading my TikTok strategy PDF)
∙ Execute Workflow (sub-workflow)
∙ Create a Workflow node
I actually do have an Anthropic API key already so I’m open to switching to Claude as the model.
Can you help me with:
- Setting up the n8n MCP
- Connecting it to the AI Agent node
- Switching from GPT to Claude
I’m on Windows and using n8n Cloud on the Starter plan. I really appreciate your offer to help — looking forward to getting this working!
@Toshi Cool! No need to switch to claude models i recommend GPTs, for the n8n MCP i provided the link to, it is really easy to setup, just get there create your API key and they have some decent free credits also, and once that is done just add a MCP node as tool to the AI agent and add that as a instruction into your System prompt,
For model changing i do not think so there is much you need to do, just remove the openrouter node and add the anthropic node.
Depending on your goal you can do a bunch of stuff with those, if all you want is to summaries that PDF just use vector vector database:
Let me know if this helped.
for your 3 questions @Toshi — switching to Claude is easy, just swap OpenRouter Chat Model for the Anthropic Chat Model node and paste your API key in credentials. claude-sonnet-4 works well for workflow generation, better reasoning on structured JSON than gpt-4.1 mini in my experience.
for n8n MCP on cloud — go to Settings > n8n API, enable it, and you’ll get an MCP server URL. then in your workflow add an “MCP Client Tool” subnode to the AI Agent and point it at that URL with your API key. thats how the agent can create/read workflows via MCP.
one thing worth knowing though: for workflow creation specifically the n8n API tool (HTTP Request node calling POST /api/v1/workflows) gives you more predictable results than MCP — you have full control over the JSON payload and its easier to debug when the LLM generates something invalid.