Overview

An AI agent is an intelligent system that can perceive its environment, make decisions, and take actions to achieve specific goals. This tutorial demonstrates building a research agent using n8n that can scrape websites, summarize content, and save information to Notion.

What is an AI Agent?

AI agents are intelligent systems that combine perception, decision-making, action, and memory to accomplish tasks autonomously. Unlike simple chatbots, agents can use tools, make plans, and adapt to achieve their goals.

Key Components of AI Agents

• Perception: Gathering information from various sources
• Decision-making: Using LLMs to understand and plan tasks
• Action: Executing tasks through tools and integrations
• Memory: Maintaining context and learning from interactions

Prerequisites

• n8n instance (cloud or self-hosted)
• Browserless account for web scraping
• Google AI API key
• Discord webhook (optional for notifications)
• Notion account (for saving research)

Step 1: Set Up Chat Trigger

Start by adding a Chat Trigger node to your n8n workflow. This will be the entry point for interacting with your AI agent. Configure it to accept user input and initiate the agent's workflow.

Step 2: Configure AI Agent Node

Add an AI Agent node and connect it to your chat trigger. Choose between different agent types based on your needs - Tools Agent for structured tasks or Conversational Agent for more natural interactions.

Step 3: Define Agent's Goal

Write a clear system message that defines what your agent should do. For example: 'You are a research assistant that helps users find and summarize information from websites. Use the available tools to scrape content, create summaries, and save important information.'

System Message:
You are a research assistant that can:
1. Scrape content from any URL provided
2. Summarize the key information
3. Save important findings to Notion
4. Send notifications when tasks are complete

Always ask for clarification if the user's request is ambiguous.

Step 4: Add Web Scraping Tool

Configure a Browserless tool to enable your agent to scrape websites. This gives your agent the ability to gather information from any URL.

{
  "url": "{url}",
  "gotoOptions": {
    "waitUntil": "networkidle0"
  }
}

Step 5: Create Notion Save Tool

Add a Notion integration tool that allows your agent to save summarized information to your Notion workspace. Configure the database and properties you want to populate.

Step 6: Set Up Notifications

Add a Discord or Slack notification tool so your agent can inform you when tasks are completed. This is especially useful for long-running research tasks.

Step 7: Test and Refine

Test your agent with various scenarios. Start with simple requests and gradually increase complexity. Monitor how the agent uses its tools and refine the system message and tool descriptions as needed.

Best Practices

• Use clear, explicit system messages to guide agent behavior
• Define tools with precise names and descriptions
• Start simple and add complexity gradually
• Test thoroughly with edge cases
• Monitor token usage and costs
• Implement error handling for failed tool calls