Building Multi-Agent AI Systems with Agent Studio

What is Agent Studio?

Agent Studio is a visual, no-code builder within your CRM that enables you to create sophisticated multi-agent AI systems. This canvas-style interface allows you to orchestrate multiple AI agents and rule-based steps within a single workflow, combining autonomous AI reasoning with deterministic logic without requiring any programming knowledge.

Key Benefits

Using Agent Studio provides several advantages for building intelligent automation:

• Visual drag-and-drop interface requiring no coding experience
• Hybrid automation that blends AI reasoning with predictable logic
• Conditional branching based on variables or AI intent analysis
• Version control with one-click publishing and rollback capabilities
• Built-in testing environment to validate workflows before deployment
• Free tier for unlimited agents with pay-per-use LLM token pricing

Core Components

Visual Builder Interface

The Agent Studio canvas provides an intuitive workspace where you can design complete workflows by dragging nodes onto the grid and connecting them. Right-click context menus allow you to rename or reorder elements, while clean layouts and connection lines make even complex systems easy to visualize.

Node Types

Agent Studio offers two primary node types that can be combined in any configuration:

• AI Agent Node: Connects to language models and tools like web search, knowledge base access, content generation, and external APIs for dynamic reasoning and data retrieval
• Sequential Node: Handles rule-based actions such as API calls, form validation, or webhook triggers where predictable, repeatable behavior is required

Conditional Routing

Instead of using separate router blocks, the connections between nodes themselves contain the logic. You can write simple expressions or use AI intent matching to determine which path the workflow should follow, enabling sophisticated branching within a compact interface.

Knowledge Base Integration

Attach multiple knowledge sources including PDFs, CSV files, FAQs, or entire websites to any AI Agent Node. This retrieval-augmented generation (RAG) approach ensures responses are grounded in your specific content while still leveraging the creative capabilities of language models.

Global Variables and Runtime Inputs

Store API keys, environment constants, or branding guidelines as global variables that can be referenced throughout your workflow. You can also pass runtime variables from calling applications or chat sessions to personalize each execution.

Version Control and Testing

Every save creates a draft version of your agent system. When ready, you can publish to production with one click or instantly roll back to any previous version if issues arise. The testing panel allows you to run complete workflows including language model calls and tool invocations before deploying to end users.

Getting Started

Accessing Agent Studio

To begin building your first multi-agent system:

1. Navigate to the AI Agents section in your CRM's left sidebar
2. Select the Agent Studio tab
3. You'll land on the dashboard where all your agents are listed

Creating Your First Agent

Click the Create Agent button in the top-right corner to open the canvas workspace. The interface is divided into three main areas:

• Left panel: Available nodes (building blocks)
• Center: Canvas (workflow construction area)
• Right panel: Settings (step configuration)

Building Your Workflow

Start by dragging an AI Agent node from the left panel onto the canvas. Click the node to configure its settings, including selecting the language model and defining the agent's behavior through prompts. Connect nodes by dragging from the connector dots to establish your workflow sequence.

Add sequential nodes for rule-based actions as needed. Use the Variables section at the top of the screen to create global variables for API keys, user data, or other reusable values.

Testing Your Agent

Before deploying your agent system, use the Test button to run simulations with sample inputs. Review response quality, tool usage, and overall flow behavior to ensure everything functions as intended.