n8n AI Agent Tutorial: Build Your First Agent

Steps

1. 01

   Create a workflow and add a Chat Trigger

   In n8n, create a new workflow. Add a Chat Trigger node so you have a chat interface to talk to your agent. This trigger gives you the 'Chat' button on the canvas that opens a local chat window for testing. (You can swap the trigger later — e.g. a webhook or schedule — once the agent works.)

   # On the n8n canvas:
   # 1. New workflow
   # 2. Add node -> 'Chat Trigger'

2. 02

   Add the AI Agent node

   Add the 'AI Agent' node — it is the core of adding AI to your workflow. In n8n the AI agent is represented as a node with some extra connection points underneath it (called sub-nodes) for a chat model, memory, and tools. Connect the Chat Trigger's output into the AI Agent node.

   # Add node -> 'AI Agent'
   # Connect: Chat Trigger -> AI Agent

3. 03

   Attach a Chat Model (required)

   AI agents require a chat model to be attached to process the incoming prompts. Under the AI Agent node, add a chat model sub-node for your provider (for example an OpenAI or Anthropic chat model), and set its credentials. Without a model attached, the agent cannot run.

   # Under the AI Agent node, add the chat model sub-node:
   #   AI Agent -> (Chat Model) -> e.g. 'OpenAI Chat Model'
   # Set the provider credential (your API key)

4. 04

   Add memory so the agent remembers the conversation

   Out of the box the model answers each message in isolation — tell it your name, and a few messages later it has forgotten. Attach a memory sub-node (for example 'Simple Memory') to the AI Agent so it retains conversation context across turns. After adding memory, repeat the conversation and the agent will remember earlier details like your name.

   # Under the AI Agent node, add the memory sub-node:
   #   AI Agent -> (Memory) -> 'Simple Memory'
   # (Simple Memory keeps recent turns; other memory types exist for longer context)

5. 05

   Give the agent tools

   Tools are what turn a chatbot into an agent — they let it take actions, not just talk. Attach one or more tool sub-nodes to the AI Agent (for example a search tool, an HTTP request tool, or another n8n node exposed as a tool). The agent decides when to call each tool based on the user's request. Start with one tool, confirm it works, then add more.

   # Under the AI Agent node, add a tool sub-node:
   #   AI Agent -> (Tool) -> e.g. an HTTP Request tool or a search tool
   # Add only the tools the agent needs for the task

6. 06

   Test your agent in chat

   Click the 'Chat' button near the bottom of the canvas. This opens a local chat window on the left and the AI agent's execution logs on the right, so you can watch which model calls and tool calls happen for each message. Send a message, confirm the agent responds, remembers context, and calls tools when appropriate. Iterate on the model, memory, and tools until it behaves the way you want.

   # Click 'Chat' on the canvas -> send a test message
   # Left: chat window   Right: agent logs (model + tool calls)

Troubleshooting

The AI Agent node errors that no chat model is connected

An agent cannot run without a chat model. Under the AI Agent node, add a chat model sub-node for your provider and set valid credentials. Confirm the credential is selected and the model name is one your account can access.

The agent does not remember earlier messages

Without a memory sub-node, each message is handled in isolation. Attach a memory sub-node (e.g. Simple Memory) to the AI Agent node. If it still forgets, check that memory is connected to the same agent and that you are using a trigger/session that preserves a conversation id.

The agent never calls a tool you attached

Make sure the tool sub-node is connected to the AI Agent's tool connection (not the main flow), and that the tool has a clear name/description so the model knows when to use it. Too many tools can also confuse the agent — start with one, confirm it is called, then add more.

Node names or UI do not match this guide exactly

n8n's AI features evolve between versions, so exact node names, sub-node labels, and the canvas UI can differ from this walkthrough. Use n8n's official Advanced AI documentation for your version as the source of truth; the structure (AI Agent + chat model + memory + tools) is stable even when labels change.