# Agents

An agent can:

* Handle or delegate tasks based on reasoning from LLM.
* Be triggered through chat, workflows, or directly from other agents.
* Continuously evaluate and improve its own process (the "agentic loop").

With Agents, you can embed AI-driven decision-making directly into your apps, workflows, and automations.

### Accessing Agents

You can set up and access agents in several ways:

* **From Triggers**: Choose *Agent* as an action when configuring triggers.

{% @arcade/embed flowId="qM5YHy90gQHp6vJewhG5" url="<https://app.arcade.software/share/qM5YHy90gQHp6vJewhG5>" %}

* **Search**: Type “Agent Chat” in the component search box and drag it into your page.

{% @arcade/embed flowId="dzyDUvcYhYQSoDLzku7K" url="<https://app.arcade.software/share/dzyDUvcYhYQSoDLzku7K>" %}

* **Left Menu**: In the left-hand menu, click the Agents icon and then select Add Agent.

{% @arcade/embed flowId="2pdlEYxOU9QEVjbWVItM" url="<https://app.arcade.software/share/2pdlEYxOU9QEVjbWVItM>" %}

{% hint style="info" %}

### When to Use Agents

Agents are useful when you need more than simple automation rules. They are best for:

* **Complex workflows** where steps depend on reasoning or context.
* **Customer-facing chatbots** that can pull data from multiple sources.
* **Task delegation** when actions may involve multiple systems or tools.
* **Dynamic decisions** that go beyond pre-defined logic.

If your task is straightforward and always follows the same pattern, a regular workflow or automation may be enough. If it requires flexibility and decision-making, an agent is the better choice.
{% endhint %}