Introduction

The Agentic Key component in wolkvox Studio allows you to create autonomous processes in which an artificial intelligence agent acts as a central orchestrator, analyzes a goal, maintains execution context, makes decisions, and coordinates other sub-agents or tools to complete a task.

Unlike traditional automation, where every step must be connected in a strictly sequential manner, Agentic Key can determine which tool it needs to use, evaluate the obtained result, and proceed with the next action or rethink the process when necessary. It can perceive, maintain context, reason, act, and validate task outcomes, even executing new cycles when an action does not initially produce the expected result.

For example, an Agentic Key can be given the goal of processing quotation requests and, based on this goal, coordinate one sub-agent to check emails, another to search for products in a database, another to reply to the customer, and another to log the activity performed. 

The component is located within the Cognitive group in wolkvox Studio and is available for Agentic Engine routing points.

What can you use Agentic Key for?

Some scenarios where you can use this component include:

• Automating business processes that require consulting different sources before executing an action.
• Processing requests received by email and responding to them automatically.
• Looking up product information, inventory, customer data, or other databases.
• Coordinating multiple specialized sub-agents within the same process.
• Creating quotes or proposals from information received in natural language.
• Automatically logging management results in a database.
• Executing processes that require verifying whether a task was actually completed before proceeding.
• Designing automations that can react to errors or incomplete results without relying solely on a linear flow.

How to configure and use the Agentic Key component?

Open component configuration

1. Access the Agentic Engine routing point you wish to configure. Within the component panel, select the "Cognitive" group, locate the "Agentic Key" component, and drag it onto the flow canvas. Connect the component according to the logic you need to implement. Double-left-click "Agentic Key" to open its configuration panel.
2. From this panel, you can define the artificial intelligence model it will use, available tools, sub-agents it can coordinate, its goal, the intents and variables you need to extract, and the variable where its response will be stored.

Select the AI Model

In "AI Model", select the model that Agentic Key will use to analyze context, reason, and determine the actions it must execute.

The available options are:

• Fast: prioritizes speed and is suitable for simple processes that require faster response times.
• Smart: prioritizes higher accuracy in understanding context and decision-making.
• Reasoner: uses greater reasoning capabilities for scenarios with more conditions or complex processes.
• Automatic: allows wolkvox to dynamically select the most appropriate model based on the complexity of each situation.

Configure available tools

The Tools section allows you to define the sources, functions, and sub-agents that Agentic Key can use to fulfill the defined goal.

wvx Copilot

In the "wvx copilot" tab, you can select a previously configured knowledge domain.

If you select a domain, Agentic Key can use that knowledge as additional context during execution.

If you select "none", no knowledge from wvx Copilot will be used.

wvx search

Go to the "wvx search" tab to provide URLs that can be used as additional sources of information.

Add one URL per line.

These URLs must be public and must not require authentication. The quality of the information retrieved will depend on the content available on the registered pages.

Functions

The "Functions" tab contains a table with the following columns:

• Name
• Description
• Variables

From this section, you can review the available functions previously created with the "Functions (With AI)" component.

Subagents

The "Subagents" tab is one of the main configurations in Agentic Key.

Here, available sub-agents are displayed alongside their descriptions. Check the checkbox for those you wish to make available to the Agentic Key.

Each sub-agent must have a specific responsibility. Agentic Key will use these agents as specialized tools and decide when to delegate a task to them based on the defined goal and configured instructions.

In the provided example, four sub-agents are used:

• One responsible for checking and reading emails.
• One responsible for querying product information in a database.
• One responsible for replying to the email with a quotation.
• One responsible for logging management results in a database.

The example instructions also establish each sub-agent's responsibility and the logical order of the process.

Agentic Key does not simply execute these agents as a fixed sequence. It can verify their outcomes and perform new cycles when it needs to complete or correct a task. In the demonstration, for example, it is explained that the agent can revisit an action when it is unsure that it finished correctly.

Define the goal and rules of the Agentic Key

Go to the "Goal, conversation tone" tab.

This field contains the primary instructions that Agentic Key will use to know what it must achieve and how to act to achieve it.

Here you can define, among other aspects:

• The role the agent will have.
• Its main goal.
• The tasks it must complete.
• The available sub-agents.
• The purpose of each sub-agent.
• The logical order in which they should be used when the process requires it.
• The validations it must perform.
• The rules it must not violate.
• The expected behavior during errors.
• The conditions to consider a task completed.
• The tone and style of responses when the process involves customer communication.

Recommendation

The prompt must clearly define:

Role → goal → available tools → responsibilities → rules → validations → error handling → completion condition.

It is not necessary to turn each step into an independent block on the canvas. Precisely one of the advantages of Agentic Key is that it can concentrate much of that logic within its configuration and later delegate it to the corresponding sub-agents.

Define the order for the Key Agent

Below the goal field, you will find "Order for the Key Agent".

Use this space when you need to provide additional instructions regarding order, priority, or behavior that Agentic Key should follow during execution.

This field can complement the general instructions defined previously when the process requires establishing a specific sequence or priority.

Configure intents and variable extraction

Go to the "Intents and variable extraction" tab.

This section allows you to identify specific intents and store information obtained during execution in variables that can later be used within the flow.

The table contains:

• Intention: configured intent.
• words: words related to that intent.
• Variables: associated variables.

Click the "+" button to create a new intent.

Configure the intent

Define:

Intent: set the name that will identify it.

Comma-separated words for intents: write the words or expressions that will help relate processed information to that intent.

You can then add associated variables using the "+" button.

When adding a variable, you must configure:

Variable name: the name under which the retrieved data will be stored.

Variable prompt: instruction explaining to the artificial intelligence what information to identify and save.

For example:

Variable name: customer_name

Variable prompt: Identify and save the customer name found during the process.

Click “Add variable” to add it.

Extraction prompts must be specific and focus on the data you want to store, without adding instructions unrelated to the extraction.

When you finish configuring the intent and its variables, click “Add intent”.

Configure AI Response

In the "AI Response" field, define the variable where the result generated by the artificial intelligence will be stored.

In the interface shown, for example, appears: $var_conver 

This variable can later be used within the flow to retrieve the result generated by Agentic Key.

The Conversations component uses this same mechanism to store the AI response.

Save the configuration

Once you have finished configuring:

• ASR.
• AI Model.
• Tools.
• Subagents.
• Goal and instructions.
• Intents and variables, if needed.
• Response variable.

Click “Save”.