Getting Started with Agent Studio

Introduction to Agent Studio

Agent Studio is the visual environment within your CRM for designing intelligent, automated assistants known as agents. It provides a flexible, node-based canvas where you can define how an agent is activated, how it processes information, the decisions it makes, and the actions it performs. This article introduces the core concepts and interface to help you begin building your own agents.

Note: Agent Studio is distinct from other AI tools in the platform, such as AI Studio. Agent Studio is focused on creating automated agents that respond to events and perform tasks, while AI Studio is designed for conversational, prompt-driven creation of web assets.

Key Benefits

Using Agent Studio offers several advantages for automating complex interactions:

• Visual Builder: Design agent logic by dragging and dropping nodes onto a canvas.
• Event-Driven: Trigger agents based on specific CRM events like new chat messages, form submissions, or tags applied to contacts.
• Intelligent Routing: Direct conversations along different paths using AI-driven intent detection or custom rules.
• Integrated Tools: Empower agents to search your Knowledge Base, query the web, or connect to external systems via APIs.
• Multi-Format Output: Create agents that can generate text, images, audio, or video.
• Data Capture: Systematically collect information like email addresses or phone numbers during an interaction.

Understanding the Interface

Agent Studio Dashboard

You can access Agent Studio by navigating to AI Agents → Agent Studio in your CRM. The dashboard that appears is your central management hub for all agents. Here you can:

• View a list of all existing agents in your account.
• See their status (Draft or Published).
• Check creation and last-updated dates.
• Organize agents into folders.
• Create a new agent or browse pre-built templates.

Creating a New Agent

From the dashboard, click Create Agent. You will be presented with two main options:

1. Start from Scratch: Opens a blank canvas for you to build an agent entirely from the ground up.
2. Use a Template: Lets you select a pre-configured agent blueprint from the Template Library, which you can then customize.

Agent Template Library

The Template Library contains ready-made agent structures designed for common business scenarios. Using a template is often the quickest way to get started, as it provides a functional framework that includes triggers, AI logic, routing, and actions. You can browse categories, preview the flow structure, and install any template to use as a foundation for your own agent.

The Builder Canvas

When you open or create an agent, you enter the visual builder canvas. This is your primary workspace where you construct the agent's logic by:

• Dragging nodes from the library onto the canvas.
• Connecting these nodes to define the sequence of operations.
• Configuring the settings for each individual step.

An agent's flow is built from left to right, beginning with a trigger and concluding with a final action or endpoint.

Core Building Blocks

Start Trigger

Every agent must begin with a Start Trigger. This node defines the specific event that will activate the agent, such as a new chat message, a form submission, or a tag being added to a contact. Without a configured trigger, the agent will not run.

Node Categories

Nodes are the fundamental components used to build an agent on the canvas. Each node performs a specific function. Key node categories include:

• AI Agent: This is the core intelligence node. It interprets incoming context or user input and generates responses based on the instructions you provide.
• Sequential: Groups multiple steps that must be executed in a strict, predetermined order.
• End Node: Signifies the conclusion of the agent's execution for a given interaction.
• Router: Creates decision points, directing the flow down different paths based on set conditions or AI-detected intent.
• Search Knowledge Base: Enables the agent to retrieve answers from your connected Knowledge Base documentation.
• Search Web: Allows the agent to fetch current information from public online sources to inform its responses.
• MCP Server: Connects the agent to supported external tool servers for advanced integrations and functionality.