Configure AI agent settings to optimize automated test creation, execution, and maintenance across your testing workflows.

​
Default Auto-heal Sensitivity

Set the default sensitivity level for auto-healing behavior when UI changes are detected.

  • Locators changes: Adapts only to html changes that affect locators
  • Moderate UI changes: Handles element relocations and structural modifications
  • Major user flow changes: Manages significant app changes that affect the user flow

Individual test steps can override this global setting as needed.

​
Auto-heal Exploration Cycle Limit

Configure how many exploration cycles the auto-heal agent uses to fix failing steps.

  • Default: 6 cycles
  • Purpose: Limits the AI exploration to fit your preferences
  • Behavior: When set to 0, auto-heal is effectively disabled

Only actions that weren’t part of the original test step are considered an exploration cycle.

​
Custom AI Prompt

Add a global AI prompt that will be applied to every AI testing agent call across your project.

This prompt provides consistent instructions and context to ensure reliable behavior across all automated testing activities.

Benefits:

  • Standardize AI behavior across your entire project
  • Provide domain-specific context about your application
  • Ensure consistent testing approaches and expectations
  • Reduce the need to repeat instructions in individual tests

Example Custom Prompts:

  • Application-specific context: “This is an e-commerce platform with a multi-step checkout process”
  • Testing guidelines: “Always verify form validation messages appear before proceeding”
  • Domain knowledge: “User accounts have three roles: customer, admin, and moderator”

The custom AI prompt will be prepended to all AI agent instructions. Keep it concise and focused on the most important context for your testing scenarios.

​
Agent Cycle Limit

Control how many cycles AI agents can run for each test step. This setting acts as a safety limit to over-exploration while giving AI enough room to solve complex problems.

Recommended Setting: Keep the default value of 70 cycles for most use cases. AI agents are designed to terminate efficiently once they complete their task, so AI rarely reaches the cycle limit.

​
When to Adjust

Increase the limit (80-100 cycles) if you have:

  • Very complex, multi-step user workflows
  • Applications with heavy async operations or slow loading times
  • Complex form validations that require multiple attempts

Decrease the limit (30-50 cycles) if you want:

  • Faster test execution with simpler workflows
  • Stricter boundaries for AI problem-solving time
  • To quickly identify problematic test steps that might be hanging