Tool calling in AI Agents

Tool calling allows an AI Agent to perform real actions in external systems — not just generate responses. Interact with Gmail, Freshdesk, CRM systems, and more to automate end-to-end workflows.

Real system actions Instruction-driven Multi-app connectivity

What is tool calling?

Tool calling is the ability of an AI Agent to invoke predefined actions (tools) while executing its instructions. These tools represent specific operations such as creating records, updating data, sending emails, or applying labels.

Instead of only analyzing data, the agent can take action based on that analysis.

Tool actions include

Creating or updating records

Sending emails or messages

Applying labels or categories

Updating ticket fields

Calling external APIs

How tool calling works

Tool calling is driven by three main components that work together to determine when and how an action is executed.

Component 01

Instructions

Define when and how a tool should be used. The agent follows the logic written here to decide which tool to call.

Component 02

Tools

Actions configured from connected apps — creating labels in Gmail, updating tickets in helpdesk, sending messages.

Component 03

Trigger data

Data from the trigger — email content, ticket details — that the agent uses to decide what action to take and what values to pass.

Using tool calling in instructions

In the Instructions section, reference tools directly by typing /. This shows configured tools or discovers new ones across all connected apps — without leaving the screen.

You type

What appears

/

All tools currently configured for this agent

/create

All create-related actions — creating tickets, labels, tasks, and more across connected apps

/ticket

All ticket-related tools from multiple connected apps in one list

This lets you both use existing tools already added to the agent and discover new ones — without leaving the instructions screen.

When does the agent call a tool?

The agent calls a tool when all three of the following conditions are met. Behavior is fully controlled by writing clear and structured instructions.

Condition 01

Clear instruction

The instructions explicitly specify an action to take

Condition 02

Data available

The required input data is present from the trigger

Condition 03

Tool configured

The tool is properly set up with all required fields

Example workflow

A practical example showing how tool calling executes end to end — email categorization using Gmail.

Gmail · Labels

Email categorization

Receives email → Analyses content → Creates label → Applies label

Trigger fires

The agent receives a new email via the Gmail trigger

Analyses content

The AI model reads subject and body, classifies the email into a category

Creates label (if it does not exist)

Calls the Create Label tool in Gmail using the classified category name

Applies label to the email

Calls the Apply Label tool to tag the email, completing the workflow

Tools used

Gmail: Create Label Gmail: Apply Label

Best practices

Follow these to ensure your tool calling setup works reliably in production.

Write clear and step-by-step instructions

Explicitly mention when a tool should be used in the instructions

Ensure all tools are properly configured before activating the agent

Use restricted or allowed values for structured fields like category, status, and priority

Test the agent with sample data before activating in production

Why tool calling is important

Tool calling enables the AI Agent to move beyond generating outputs and actually perform tasks — automating real workflows, reducing manual effort, and connecting systems through intelligent decision-making.

Automates real workflows

Agents complete entire end-to-end processes without human intervention

Reduces manual effort

Repetitive tasks like tagging, updating records, and sending responses are handled automatically

Ensures consistency

Every execution follows the same logic, producing reliable and predictable results

Connects multiple systems

Intelligent decision-making bridges CRM, helpdesk, email, and other apps in one workflow

Get started

Ready to put your agent to work?

Configure tools, write instructions, and let your AI Agent act — automatically.

Get started free

​ Tool calling in AI Agents

Tool calling in AI Agents

What is tool calling?

How tool calling works

Using tool calling in instructions

When does the agent call a tool?

Example workflow

Best practices

Why tool calling is important

Tool calling in AI Agents