How is an agent implemented in n8n?

I'll help you understand how agents are implemented in n8n.

In n8n, agents are implemented through the AI Agent node, which provides several types of agents based on LangChain. The main agent types available include:

1. Tools Agent (default) - This is the recommended agent for most use cases. It can understand the capabilities of different tools and determine which tool to use depending on the task.

2. Conversational Agent - This agent is designed for human-like conversations and can maintain context, understand user intent, and provide relevant answers.

3. OpenAI Functions Agent - This agent works specifically with OpenAI functions models and requires the OpenAI Chat Model.

4. Plan and Execute Agent - This agent creates a high-level plan to solve tasks and then executes the plan step by step.

5. ReAct Agent - This agent reasons about tasks, determines necessary actions, and executes them in a cycle until completion. It can break down complex tasks into smaller sub-tasks.

6. SQL Agent - This agent uses SQL databases as data sources and can convert natural language questions into SQL queries.

All agents in n8n require at least one tool sub-node to be connected to function properly. The agents run multiple times during workflow execution - for initial setup, calling tools, evaluating responses, and responding to users.

Unlike chains that follow a fixed sequence of operations, agents in n8n can make decisions about which actions to take based on the language model's understanding of the task at hand.