Skip to main content

What Are AI Agents?

AI Agents are browser automation bots that execute multi-step workflows. They can visit websites, click elements, fill forms, upload files, extract data, run JavaScript, solve CAPTCHAs, and make AI-powered decisions. Agents are used to automate repetitive tasks that would otherwise require manual browser interaction.

Building an Agent

Agents are built using a visual instruction workflow editor. The editor presents a canvas where admins compose a sequence of instructions that the agent follows step by step.

Instruction Types

InstructionDescription
Visit URLNavigate the browser to a specific address
Click ElementClick on a page element identified by a selector
Type InputType text into a form field
Upload FileUpload a file to a designated element
Download FileDownload a file from a URL
Enter IframeSwitch the agent’s context into an embedded iframe
Take ScreenshotCapture a screenshot of the current page
Wait for ElementPause until a specific element appears on the page
Wait for URLPause until the browser URL contains a specific pattern
Solve CAPTCHAAttempt to solve a CAPTCHA challenge
AI InstructionGive the agent a natural-language instruction and let AI determine the actions
Run JavaScriptExecute custom JavaScript code in the browser
Fetch OTPRetrieve a one-time password from a configured source
Loop Over DataIterate over a data array, executing child instructions for each item
Conditional BranchEvaluate a condition and follow the TRUE or FALSE path
TerminateEnd the agent’s execution

Element Selection

When an instruction targets a page element, the agent identifies it using one of these methods:
MethodDescription
CSS SelectorStandard CSS selector syntax
XPathXPath expression
IDElement’s HTML ID attribute
Class NameElement’s CSS class
Tag NameHTML tag name
Link TextFull text of a hyperlink
Partial Link TextPartial match on hyperlink text
Text ContainingMatch by text content

Input Variables

Input variables provide data to the agent before execution. Each variable has:
PropertyDescription
NameVariable identifier
DescriptionWhat the variable represents
TypeText or File
MandatoryWhether the variable must be provided
SensitiveWhether the value should be masked in logs
Dynamic RuntimeWhether the value is resolved at execution time
Input variables can be referenced within instruction configurations (e.g., typing a candidate’s name into a form field).

Output Variables

Output variables capture data extracted during execution. Each variable has a name and description. Output values are collected from instruction results and made available to the system after the agent finishes.

States

Agents can define named states that act as workflow checkpoints. Each state has a name, description, and a default flag. States allow the agent to resume from a known point or branch execution based on the current state.

How Agents Are Triggered

MethodDescription
Automation ActionThe “Run AI Agent” automation action triggers execution with application context
Form ElementAn “AI Agent” form element in a hiring flow triggers the agent when the candidate reaches that step
Manual TriggerAdmins can trigger an agent directly, selecting a specific hiring flow or check-in form and the applications to process
TestAdmins can test the agent from the editor, with options for streaming output and debug mode

Manual Trigger

When manually triggering an agent, admins:
  1. Select the context: Hiring Flow or Check-in Form
  2. Choose the specific flow or form
  3. Select which applications to process
  4. Initiate the execution

Testing

The test interface allows admins to run the agent with:
  • Streaming mode — Watch the execution output in real-time
  • Debug mode — Verbose logging for troubleshooting
  • Input variable values — Provide text or file values for each input
  • Loop data — For agents with loop instructions, provide the data array

Execution History

Every agent execution creates a record that tracks:
  • Execution status and timing
  • Input data provided
  • Output data captured
  • Detailed execution logs
Admins can view execution history from the agent detail view or from dedicated execution log pages.

Scope and Permissions

ScopeDescription
CompanyAvailable only within the company that created it
GlobalAvailable across all companies (managed by platform staff)
Global agents can only be created and deleted by platform staff. Company admins can use global agents but cannot modify or remove them.

Integration with Other Modules

Automations

The “Run AI Agent” action connects agents to automation workflows. When triggered, the automation provides the application context, and the agent’s output variables can feed into subsequent automation actions.

Form Builder

AI Agents can be embedded as form elements. When a candidate reaches the agent step, the system executes the agent with the candidate’s data. The agent’s output variables can be mapped to fill other form fields.

Check-in Forms

Agents can process check-in form data when triggered manually or through automations scoped to check-in forms.