Complete guide to n8n

Getting started

Install n8n on your own server (free, open source) or use the paid cloud version at n8n.io. When you open the editor, you'll see a blank canvas where you build workflows by connecting nodes: a trigger node starts the workflow, and the following nodes run actions or call external services. For your first workflow, add a Trigger (for example, a webhook or a timer) and connect a simple action node (send an email, save to a spreadsheet). Once you understand the basic logic of connected nodes, you can add an AI Agent node so the AI decides what to do based on context, instead of following fixed rules.

Use cases

Advanced setup

1. Basic structure of an AI workflow
   A trigger node (webhook, timer) starts the workflow. The AI Agent node acts as the orchestration layer, using reasoning to decide which tools to invoke based on the request it receives.
2. Connecting tools to the agent
   Give the agent access to specific tools: one that queries your customer database, another that checks an order's status, another that creates a support ticket. The agent decides which one to use based on the request.
3. Memory management
   For agents handling multi-turn conversations, choose between window memory (last N messages), summary memory (a rolling summary), or external memory backed by Redis or Postgres.
4. Choosing the model provider
   n8n supports OpenAI, Anthropic, Mistral, Google Vertex AI, local models via Ollama, or any OpenAI-compatible endpoint through the generic node.

Tips

• Start with a simple workflow (a trigger plus one action) before adding an AI Agent: understanding the basic node logic makes it much easier to debug more complex workflows later.
• Give the agent only the tools it actually needs for the task: the more options it has, the more likely it is to get confused about which one to use.
• Use summary memory instead of full window memory when conversations get very long, so you don't saturate the model's context window.
• If data privacy is a concern, consider self-hosting n8n instead of using the cloud version: you keep full control over execution and data.
• Test each node on its own before connecting the full workflow: it makes it much easier to find the exact point where something fails.

FAQ