Skip to main contentAgent Settings
Agent Settings centralizes the runtime rules your agents follow during execution, from visual validation to retry logic. Use these controls to keep runs consistent across environments and team members.
Visual Check
Enable Visual Check to analyze screenshots captured during a run. This adds a UI-level validation pass that flags unexpected visual changes, layout shifts, and missing UI states.
Use it when you want fast feedback on UI regressions alongside functional steps.
Retry Configuration
Retry Configuration defines how the agent reacts to transient failures.
- Max failures — up to 5 retryable failures before the run stops.
- Retry duration — wait time between retries, up to 60 seconds.
Set lower values for fast feedback, or raise the duration when you expect slow backends or rate limits.
Default Context (Optional)
Default Context is an optional instruction block that is appended to every agent run. Use it for team-wide guidance such as:
- Preferred tone or execution style.
- Required guardrails (do not delete records, avoid admin actions).
- Common navigation constraints or page expectations.
Leave it blank if you want each workflow prompt to stand alone.
Output Settings
Record GIF controls whether a lightweight GIF is captured for each run.
- Default:
true
- Current status: temporarily disabled (no GIF is generated)
The toggle remains visible so you can see the intended default while the feature is paused.
Element Attributes
Element Attributes defines which HTML attributes the agent records when identifying UI elements. This helps with consistent element naming and selector generation.
Typical attributes to include:
data-testidaria-labelnameidroleplaceholderalttitle
Keep the list focused on stable, semantic attributes to reduce selector churn.
Execution Limits
Execution Limits keep runs bounded so steps do not explode in complexity.
If a step is likely to exceed the limit, break it into smaller, clearer steps in your workflow prompt. 
