Turn an n8n Workflow Into a Real Agent

A regular n8n workflow is like an assembly line: data enters at one end, passes through a fixed sequence of steps, and exits at the other. It is reliable, fast, and great for predictable tasks. But what if the task requires decisions? What if the number of steps depends on what the AI finds?

That is the difference between a workflow and an agent. An agent can call tools, inspect the results, and decide what to do next, looping until it has finished the job. n8n's AI Agent node makes this straightforward to build.

Workflows vs. Agents: What Changes

In a standard n8n workflow, you define every step in advance. The flow is deterministic: step 1 always leads to step 2. This is ideal for recurring data pipelines and integrations where the logic does not change.

An agent workflow replaces that fixed sequence with a reasoning loop. The AI Agent node receives a task, decides which tool to call, calls it, observes the result, and repeats until it believes the task is complete. The number of steps is not fixed: it depends on what the agent encounters at runtime.

This makes agents powerful for research tasks, multi-step lookups, and anything where the path depends on intermediate results.

The AI Agent Node: How It Works

Add the AI Agent node to your canvas. You will configure four things: the model (Claude or GPT-4), the system prompt, the tools, and the memory setting.

Tools are the key addition. Each tool is a capability the agent can invoke, such as an HTTP call, a code block, a web search, or a database query. The agent decides autonomously which tool to call based on the task. You define the tools; the agent decides when to use them.

Set the system prompt to define the agent's persona and scope. Here is a solid starting point for a research agent:

Defining Tools for Your Agent

Tools in n8n are defined as sub-workflows or as node types attached to the Agent node. Common tools include HTTP Request (call any external API), Code (run custom JavaScript), and the built-in web search integration.

Here is an example of a custom HTTP tool definition that lets the agent look up a customer record in your CRM:

Tool Name: lookup_customer
Description: Look up a customer by email address. Returns name, plan, and account status.

HTTP Request:
  Method: GET
  URL: https://api.yourcrm.com/v1/customers
  Query Params:
    email: {{ $fromAI('email', 'The customer email address to look up') }}
  Headers:
    Authorization: Bearer YOUR_CRM_API_KEY
    Accept: application/json

The $fromAI() function is n8n's way of letting the agent fill in the tool parameter at runtime. When the agent decides to call this tool, it provides the email address from context.

A Working Example: Support Ticket Resolver

Here is a concrete agent workflow to build. Trigger: a new ticket arrives via webhook from your support system. The Agent node receives the ticket text and has access to three tools: a knowledge base search, a customer lookup, and a Slack notifier.

The agent reads the ticket, searches the knowledge base for a matching answer, looks up the customer's plan, and either drafts a reply (if it found an answer) or posts to a Slack escalation channel (if it did not). The number of tool calls varies per ticket. No fixed branching logic is required.

Build this once and it handles tier-1 support triage automatically, 24 hours a day.