Create your first AI Agent

AI Agents

Create your first AI Agent

Set up a trigger, define instructions, connect services, add tools, and activate your agent — by the end you will have a fully functional AI Agent automating real workflows.

9 steps Step-by-step guide Beginner friendly

Before you begin

Before creating the agent, take a moment to plan. Defining what your agent should do upfront will make every step faster and more accurate.

Think through the following before you open the builder:

What the agent should do

What input it will receive

What actions it should perform

Which event will trigger it

Example

Automatically categorize support tickets and update fields in a helpdesk system.

Steps in this guide

Plan your agent

Add a trigger

Create connections

Select AI model

Define instructions

Add tools

Configure tool fields

Save & activate

Test your agent

Step-by-step guide

Plan your AI Agent Prep

Before creating the agent, clearly define what it should accomplish. A well-planned agent is faster to configure and more reliable in production.

Example

Automatically categorize support tickets and update fields in a helpdesk system.

Add a trigger Required

The trigger determines when the agent runs. Without a trigger, the agent will never execute.

To add a trigger

Click Add Trigger

Select a trigger source

Apps (Gmail, Freshdesk, CRM) Webhook Scheduler

Example

Freshdesk: New Ticket — fires every time a new support ticket is created.

Create required connections Required

Before using triggers, models, or tools, you must create connections. Connections allow the agent to interact with external systems securely.

To create a connection

Select the app, model provider, or tool

Click Create Connection

Provide required credentials

Authentication types

OAuth

API Key

Key-based access

Other

Varies by service

Connections required for

Trigger apps AI models Tools

Select an AI model Required

Choose an AI model for your agent. Ensure the required connection for the model provider is already configured. The model determines how well the agent performs.

Understanding

Interprets inputs accurately

Reasoning

Applies logic step by step

Output quality

Generates precise results

Define instructions Required

Instructions define what the agent should do and how it should behave. Well-written instructions directly determine the quality of the agent's output.

Clearly define the agent's role

Specify step-by-step logic

Define rules and constraints

Use / to reference and call tools directly within the instructions

Use Enhance to improve clarity and effectiveness

Example

Analyze incoming tickets, classify them into category, sub-category, and item, and update the ticket using the appropriate tools.

Add tools Required

Tools allow the agent to perform actions in connected systems. You can add multiple tools depending on your workflow. Ensure connections are set up for each tool.

To add a tool

Click Add Tool

Select the app and action

Configure the tool

Get data from a system Update records Send emails Add labels or tags

Configure tool fields Required

Each tool requires field configuration to define how data is passed. Proper configuration ensures accurate and consistent execution.

Field options

Let agent select

Dynamically determined based on instructions and context

Choose a value

Set a fixed value manually

Ignore

Skip field if not required

Let agent choose from allowed fields

Restricted to predefined values — useful for category, status, priority

Examples

Ticket ID

Let agent select

Category

Let agent choose from allowed fields

Priority

Choose a value

Save and activate the agent Activate

After configuring all components, save and activate the agent. Only active agents will run when triggered.

Click Save

Set the agent status to Active

Test your AI Agent Final step

Before using the agent in production, test it with sample data to verify everything works correctly.

Trigger the agent with sample data

Verify outputs and actions

Ensure tools execute correctly

Refine instructions if needed

Tip

Use the Activity section to monitor executions and debug issues after the agent is live.

Using Copilot (Optional)

Copilot helps you build and configure agents faster. Describe your use case in plain language and let Copilot guide the setup.

Natural language setup

Describe your automation in plain words

Step-by-step guidance

Guided configuration for complex workflows

Instruction generation

Generate and refine agent instructions automatically

Faster configuration

Set up your agent more efficiently

You are ready

Your first AI Agent is set up.

Monitor runs in the Activity section and refine instructions as needed.

Get started free