Kuika's AI Agents feature allows you to integrate more interactive and scenario-based artificial intelligence experiences into your applications. With this feature, you can create artificial intelligence agents that perform different tasks, have specific roles, and can provide dynamic responses based on user input.
Steps to Create an AI Agent
Log in to the Kuika platform and open the project you want to work on.
Go to the Datasources module.
Go to the “AI Agents” section in the left panel.
Click the + icon to create a new AI Agent.
In the window that opens, you will need to fill in the following information:
Agent Name
Give the agent a descriptive name (for example, “JobPostGeneratorAgent”).
Model Selection
There are two different model types available:
All: Used for versatile, general-purpose tasks.
CrewAI: Provides an advanced scenario infrastructure that enables multiple AI agents to work as a team.
AI Agent Templates
You can create an agent using ready-made templates or define a structure from scratch using the Custom option:
Custom
Definition: Allows you to create your own agent from scratch.
Use Case: For developers who want to use unique task definitions, custom user inputs, or different AI models.
Features: You have full control over role prompts, user prompts, models, and parameters.
Sample Input: “Generate a title and description for a podcast episode.”
Employee Researcher
Definition: Gather information about a specific employee from the internet or a database.
Use Case: Human resources, internal communications, and executive dashboards.
Data Source: Typically supported by Serper Tool.
Input: Employee name, position, email.
Output: Structured text or JSON about the employee's history, competencies, and projects.
Company Researcher
Description: Conducts online research on a specified company and provides summary information.
AI Agents can be adapted to different scenarios with their customizable template structure. The system offers a flexible solution thanks to input definitions, user interaction, and background tools.
How to Use?
Define Agent Inputs:
Click the “Agent Inputs” button in the upper right corner.
Define the required inputs according to their types:
String: Text information (e.g., candidate name, product name)
Boolean: True/False options (e.g., was the reference check performed?)
Integer: Whole numbers (e.g., years of experience)
Float: Decimal numbers (e.g., expected salary)
Interact with the Chatbot:
The inputs you defined become active in the chatbot interface.
The user enters data into the Chatbot using natural language.
Sample message:
“Candidate name: Ayşe Demir, Was the reference check performed? Yes, Years of experience: 6, Expected salary: 85000.”
AI Agent Generates Response:
Creates an automatic response based on the defined inputs.
If necessary, analysis, calculation, or data query operations are performed.
Results are presented to the user in real time.
Perform Actions Using Tools:
Thanks to the tools you add to the Agent, you can create files, extract data, and perform analysis.
Example Scenario: Candidate Evaluation
Objective: Obtain candidate information, perform suitability assessment, and save the result to a file.
Steps:
Define the following inputs from Agent Inputs:
candidateName (String)
hasReferenceChecked (Boolean)
yearsOfExperience (Integer)
expectedSalary (Float)
Write to the chatbot as follows:
“Candidate name: John Doe, Has the reference check been completed? Yes, Years of experience: 6, Expected salary: 85,000.”
The agent generates the following response:
John Doe
Reference check completed.
With 6 years of experience, he is suitable for the position.
Expected salary: 85,000 TL – This value appears to be in line with the market average.
Evaluation report is being generated..."
FileWriterTool is activated and creates a file named john_doe_evaluation.txt.
What is Agent Response Builder?
It is used to configure what kind of response the agent will give to the user. It enables AI responses to be generated in a dynamic, meaningful, and structured way. This structure allows you to:
Write custom messages based on inputs,
Embed tool outputs into the response,
Generate flexible content with conditional statements (e.g., show if reference check is done, warn if not).
It can be accessed by clicking the view icon (Response) in the Agent panel.
Here, you can call all inputs using the input name in the text and define conditional statements with Jinja-like structures.
Supported Tool Descriptions
Settings for MCP-based integrations (Brave Search, Slack, Github, GoogleWorkspace, GoogleMaps, Youtube, Trello, Airbnb, Office365, Dropbox) are configured through the Configuration Manager module. Once API keys, access permissions, and connection details are defined for the relevant MCP in this module, they become available for use within the AI Agent.
ScrapeWebsiteTool
Purpose: To extract data or content from a specified website.
Usage: Parses the HTML content of the site when a URL is entered. Specific tags, titles, or product information can be obtained.
Example Scenario:
Comparing product prices on an e-commerce site.
Extracting headlines and content from a news site.
CodeInterpreterTool
Purpose: To execute Python code, perform calculations, analyze data, and create graphs.
Usage: Enables analysis based on numerical inputs received from the user. Performs complex operations (averaging, comparison, visualization).
Example Scenario:
Comparing salary expectations with the market average.
Calculating the average of numbers received from the user.
SQLTableInfoTool
Purpose: To discover table structures in the connected database.
Usage: When connected to the database, it lists table names, column names, and data types.
Example Scenario:
What fields are in the table named “sales”?
In what format is the date of birth stored in the “customers” table?
SQLDataQueryTool
Purpose: To retrieve data directly by running SQL queries.
Usage: Queries such as SELECT, JOIN, and WHERE can be run. Results are presented in tabular form.
Example Scenario:
List orders from the last 7 days.
Show all transactions for a specific customer.
Retrieve products with low stock levels.
ExcelReadTool
Purpose: To read Excel files and analyze the data within them.
Usage: Sheets, cells, and data sets in the loaded .xlsx file can be examined.
Example Scenario:
Find total sales from a sales report in Excel.
List orders for a specific date range.
Compare sales by product.
FileWriterTool
Purpose: Create or write files in formats such as text, JSON, CSV.
Usage: An evaluation, report, or output is prepared using the information provided by the user and written to a file.
Example Scenario:
Candidate evaluation results are stored in a .txt file.
An analysis result is exported in CSV format.
Chatbot output is written to a JSON file and archived in the system.
FileReadTool
Purpose: To read and analyze the content of uploaded files.
Usage: When a user uploads a file, its content can be examined line by line or based on its structure.
Example Scenario:
To analyze a past sales report uploaded by the user.
To summarize a .txt file containing customer feedback.
Extracting specific fields from a JSON-formatted structure.
Brave Search MCP
Purpose: To perform internet searches via the Brave search engine.
Usage: Retrieves web results on the specified topic.
Example Scenarios:
Searching for current news.
Finding content on the internet about a specific company.
Slack MCP
Purpose: To interact with Slack channels.
Usage: Message contents can be read, summarized, or searched.
Example Scenarios:
Summarize channel conversations from the last week.
Find discussions on Slack about a specific project.
Github MCP
Purpose: To interact with GitHub repos and issues.
Usage: Open PRs, issues, and commit messages can be read.
Example Scenarios:
List open PRs.
Review the latest commits on a specific branch.
GoogleWorkspace MCP
Purpose: Integration with Google Docs, Sheets, and Drive content.
Usage: Documents can be searched and read.
Example Scenarios:
Read the latest report in Google Sheets.
Find a specific document in Drive.
GoogleMaps MCP
Purpose: Obtain location and place information.
Usage: Provides details about a specified address or location.
Example Scenarios:
List nearby restaurants.
Retrieve map information for a specific address.
YouTube MCP
Purpose: Search YouTube videos and retrieve content information.
Usage: Title, description, and channel information can be retrieved.
Example Scenarios:
List videos on a specific topic.
Analyze links in video descriptions.
Trello MCP
Purpose: Interact with Trello boards.
Usage: Cards and lists can be read.
Example Scenarios:
List all tasks on a board.
Read the description of a specific card.
Airbnb MCP
Purpose: Get accommodation information via Airbnb.
Usage: Lists suitable homes based on location and date information.
Example Scenarios:
Search for a 3-day stay in Paris.
Find homes within a specific price range.
Office365 MCP
Purpose: Integration with Office 365 files and emails.
Usage: Documents can be read, emails can be analyzed.
Example Scenarios:
List the latest emails in Outlook.
Reading a report in OneDrive.
Dropbox MCP
Purpose: To interact with Dropbox files.
Usage: Documents can be read, files can be listed.
Example Scenarios:
Opening a presentation in Dropbox.
Finding the most recently uploaded files.
Tips
Any changes made to inputs automatically update the AI response.
Tool combinations allow you to create different usage scenarios.
Each input defined in the prompt can be used directly as {{inputName}}.
How to Configure Serper Tool Settings?
1. Add Configuration to the Application
Log in to the Kuika platform.
Open the project you will be working on from the Apps screen.
Click on the Configuration Manager module.
On the screen that opens, give the configuration a name and click the CREATE button.
2. Configure Serper Tool Settings
After creating a new configuration, open the App Settings screen.
Go to the AI Settings section.
Click on the Serper Tool option from the drop-down menu.
Click the ADD NEW button.
Configure the following settings in the pop-up window that opens:
Name: Name to be given to the configuration.
API Key: Access key obtained via the Serper API.
Once the Serper Tool becomes available for use by the Agent, it is automatically detected in AI Agent configurations and actively used in tasks involving external data sources.
Using the AI Agent
Go to the UI Design module.
Open the screen where you want to add the Agent.
Click on the +ADD ACTION > AI Agents tab and select the agent you created.
Connect the relevant user inputs and activate the AI-powered scenarios.