Creating your First Agent

Agents are the core components of your coworker in Wexa. Each agent performs a specific task to streamline your workflow. Here's a breakdown of the Agent Creation interface and what to input:

Step-by-Step Guide to Creating an Agent

1. Agent Type

1.1 Select Agent Type

Skilled Agent: Performs specialized actions based on assigned skills (e.g., sending emails, fetching data).

Decision Agent: Makes branching choices based on outcomes or input values.

1.2 Fill in Agent Details

Title: Name your agent (e.g., “Data Validator”, “Meeting Scheduler”).

Role: Explain what the agent does (e.g., “Processes raw lead data into qualified contacts”).

Description: Expand on how the agent fits into the workflow and what it handles.

2. Task

Tasks are clear, detailed instructions that guide your agent step-by-step. Each Task should specify:

What the agent needs to do

How to perform the task

What output is expected

Tasks are essential for defining each step in a process, whether it's extracting insights, filling out forms, sending messages, or making decisions.

You can use @ to reference or create variables dynamically within a task.

Make sure to include any required inputs, conditions, or logic the agent must follow to perform the task correctly.

3. Actions

In Wexa Studio, Actions empower AI agents to perform tasks by defining how they interact with applications and data. Actions consist of two key components: Skills and Tables.

3.1 Skills

Skills enable AI agents to perform specific operations by connecting with your existing business applications—without the need for custom development.

Automate routine tasks like sending emails, creating tickets, updating records, and more.

Leverage the Connectors Marketplace to quickly integrate with popular tools.

Need a specialized connection? You can request a custom integration directly.

These Skills equip your agents to act intelligently across your workflows, reducing manual effort and increasing efficiency.

3.2 Tables

Tables are internal databases that agents use to retrieve or update structured information. They serve as the backbone of your workflows by providing essential data for informed action.

Store and manage data such as customer records, order details, product inventory, or support logs.

You can enable agents to filter, extract, or update entries based on your defined logic.

With Tables, your agents are not just automating tasks they’re making decisions grounded in data.

Together, Skills and Tables allow your AI agents to seamlessly connect logic, data, and action—turning static processes into dynamic, intelligent workflows.

4. Triggers

These are events that will initiate the agent's workflow. You can add multiple triggers to your coworker for varied responses to different situations. Triggers automate the next steps. When a task is finished, they automatically start the following task, creating a seamless workflow. Dynamic data is inserted into the process using placeholders.

How It Works

Define a Trigger

Name your trigger (e.g., "Trigger 1") to distinguish it within your automation setup.

Select a Process Flow. Choose from the list of available process flows

Execution

Once the trigger condition is met, the selected process flow kicks off immediately—without manual intervention.

5. Knowledge Base

The Knowledge Base in Wexa Studio not only stores critical information but also supports a variety of files such as PDFs, PPTs, Docs, and more. AI agents can refer to these documents while performing tasks such as answering queries, providing customer support, or making decisions.

6. Advanced Settings

6.1 LLM Model Configuration

Select an LLM (e.g., GPT-4, Claude, Gemini) aligned with your task (creativity, accuracy, speed). Balance model size, cost, and latency.

6.2 Output Token Limit

Set token limits to control response length, balancing conciseness and depth (e.g., 200-500 tokens). Higher limits suit detailed tasks but risk inefficiency; lower limits enforce brevity.

6.3 Set Human in Loop

Wexa also integrates human oversight seamlessly into automated workflows for compliance, risk management, or strategic decisions.

Choose one:

Auto Approve: Agent runs tasks automatically.

Ask for Approval: User must approve each action—useful for sensitive tasks.

7. Creating Multi-Step Workflows

You can create multiple agents within a single coworker, essentially building a multi-step workflow. Each agent acts as a step in the process, leveraging its assigned skills (connector permissions) to perform specific tasks. By linking these agents together with triggers, you can design complex automation sequences.

8. Completing Your Coworker

Once you've finished defining and configuring all the necessary agents within your coworker, submit the form to finalize its creation.

With your coworker built and deployed, you're now equipped to automate specific tasks or processes within your workflow. Remember, coworkers can be designed for various functions, and the specific configuration steps might vary depending on the desired functionality.

For more info check this tutorial:

Creating your First Agent

1. Agent Type#

2. Task#

3. Actions#

3.1 Skills#

#

3.2 Tables#

4. Triggers#

5. Knowledge Base#

6. Advanced Settings#

6.1 LLM Model Configuration#

6.2 Output Token Limit#

6.3 Set Human in Loop#

7. Creating Multi-Step Workflows#

8. Completing Your Coworker#