AI Agents with n8n: RAG, MCP, and Intelligent Automation

Level: Intermediate–advanced
Requirements: Basic n8n knowledge. No in-depth AI knowledge required.

This would be the course with the most commercial potential.

Unit 1. Generative AI applied to automation

• Generative AI and LLMs.
• How a language model works.
• Tokens.
• Context.
• Prompts.
• Temperature and parameters.
• Structured Outputs.
• Models from OpenAI, Anthropic, Google, and other providers.
• Traditional automation vs. AI-powered automation.
• When to use AI and when not to.
• Costs and limitations.

Practice: Automatically classify and process information using an LLM.

Unit 2. Creating workflows with AI

• n8n AI nodes.
• Chat models.
• Prompt templates.
• Structured Output Parser.
• Classification.
• Information extraction.
• Summarization.
• Content generation.
• Document analysis.
• Email processing.
• Sentiment analysis.
• Integrating AI into traditional workflows.

Practice: Create an intelligent system to automatically process customer emails.

Unit 3. AI Agents in n8n

• What is an agent?
• Difference between chatbot and agent.
• AI Agent.
• Model.
• Memory.
• Tools.
• Custom tools.
• Using APIs as tools.
• Agents that perform actions.
• Human-in-the-loop.
• Limits and permissions.
• Designing secure agents.

Practice: Create an agent that queries information and executes actions in external applications.

Unit 4. RAG and business knowledge

• The problem of LLM knowledge.
• What is RAG?
• Embeddings.
• Chunking.
• Vector databases.
• Vector stores.
• Document ingestion.
• PDF, Word, web, and other sources.
• Semantic retrieval.
• Retriever.
• Generation based on corporate documents.
• Knowledge updates.
• Response evaluation.
• Reducing hallucinations.

Practice: Create an assistant that responds using internal documentation.

Unit 5. MCP and external tools

• What is Model Context Protocol?
• Client-server architecture.
• MCP Client.
• MCP Server.
• Tools.
• Resources.
• Exposing tools through MCP.
• Using external services from agents.
• n8n as a consumer and provider of tools.
• Security and permissions.
• Business cases for MCP.

Practice: Connect an agent with external tools using MCP.

Unit 6. Advanced agent architectures

• Deterministic workflows + agents.
• When to use each approach.
• Specialized agents.
• Agents with multiple tools.
• Orchestrators.
• Routing.
• Multi-agent architectures.
• Short and long-term memory.
• Conversation state.
• Subworkflows as tools.
• Agents that call other processes.
• Execution of complex business processes.

Practice: Create several specialized agents coordinated from n8n.

Unit 7. Production, security, and governance of AI agents

This part seems especially important to differentiate a professional course from a YouTube one.

• Prompt injection.
• Improper access to tools.
• Information protection.
• Personal data.
• Credentials.
• Limiting agent actions.
• Human approval.
• Auditing.
• Logging.
• Error management.
• Timeouts.
• Cost control.
• Token consumption.
• Result evaluation.
• Testing AI workflows.
• Observability.
• Monitoring.
• Best practices for bringing agents to production.

Final project

Build a complete business agent that:

1. receives a request;
2. identifies the intent;
3. consults documentation using RAG;
4. uses tools or APIs;
5. executes an action via n8n;
6. requests human authorization for sensitive actions;
7. logs the result.