Skip to main content
Agent behavior settings configure context-driven overrides, alerting, human handoff, and real-time agent assistance.

Behavior Profiles

The Behavior Profiles page lets you define context-driven overrides for agent conversation behavior. Profiles layer on top of any agent baseline when their conditions match, without modifying the core agent definition. Navigation: ProjectSettingsBehavior Profiles

Create a Behavior Profile

Click Create Profile to open the structured editor. The profile header includes the following fields:
FieldDescription
Profile NameA unique identifier for the profile (for example, voice_support).
PriorityNumeric priority. Lower values take precedence when multiple profiles match.
WHEN ExpressionA boolean expression that determines when this profile activates (for example, true for always-on, or a context variable check).
Below the header, the profile editor organizes settings into three behavior sections: Speaking
SettingDescription
StyleResponse style (for example, warm and concise).
ToneResponse tone (for example, reassuring).
EmotionEmotional register (for example, calm).
PaceResponse pacing (for example, steady).
Language PolicyLanguage selection policy (Inherit, Fixed, or Detect).
Fixed LanguageThe language code when Language Policy is Fixed (for example, en-US).
Max SentencesMaximum number of sentences per response.
One Thing at a TimeWhether the agent addresses one topic per turn.
Tool Lead-inHow the agent introduces tool invocations.
Tool Results StyleHow the agent presents tool results to the user.
Tool Result PointsNumber of key points to extract from tool results.
Internal HandoffsBehavior when handing off to another agent.
Human HandoffsBehavior when handing off to a human operator.
Listening
SettingDescription
Barge-InHow the agent handles user interruptions mid-response.
On PauseBehavior when the user pauses mid-utterance.
On OverlapBehavior when user and agent speak simultaneously.
Unclear AudioHow the agent handles unclear or noisy input.
Self-CorrectionHow the agent detects and handles user self-corrections.
Interaction
SettingDescription
Answer ShapeThe structure of responses (for example, direct, exploratory).
DetailLevel of detail in responses.
InitiativeHow proactively the agent guides the conversation.
GroundingHow the agent establishes shared understanding.
ClarificationWhen and how the agent asks clarifying questions.
Max Clarification QuestionsMaximum number of clarification questions per turn.
Assume When Low RiskWhether the agent assumes intent when the risk of misunderstanding is low.
Confirm ParametersWhether the agent confirms parameters before executing actions.
Confirm ActionsWhether the agent confirms actions before performing them.
UncertaintyHow the agent communicates uncertainty.
Offer Next StepsWhether the agent proactively suggests next steps.
EmpathyLevel of empathetic language in responses.
On CorrectionBehavior when the user corrects the agent.
On ConfusionBehavior when the user appears confused.
On MisheardBehavior when the system misrecognizes audio input.
Repair AttemptsMaximum number of repair attempts before escalation.
Avoid Re-askingWhether the agent avoids repeating questions already answered.
Remember ConstraintsWhether the agent tracks and respects stated constraints across turns.
ClosureHow the agent closes a conversation or topic.
Authoring Note: A free-text field for documenting when and why this profile applies. All settings default to Inherit, meaning the agent’s base configuration applies. Override any setting to customize behavior for the profile’s conditions. Click Save to create the profile, or switch to Raw DSL to edit the profile definition as code.

Alert Settings

The Alerts page serves as an inbox for workflow approvals and proactive notifications. Navigation: ProjectSettingsAlert Settings Tabs
TabPurpose
ApprovalsReview and act on pending workflow and agent approvals, data entry requests, and completed items.
Alert RulesConfigure rules that trigger proactive notifications based on system events or thresholds.
Filter the Approvals tab by source (Workflow or Agent) and type (All, Approvals, Data Entry). Toggle Include completed to view historical approvals alongside pending items.
The Approvals tab requires the human_task:read permission. If your role lacks this permission, the system displays a “Forbidden: missing required permission” message.

Agent Transfer

The Agent Transfer Settings page configures defaults for agent transfer sessions, routing, voice gateway, and PII handling. Navigation: ProjectSettingsAgent Transfer The page organizes content into four collapsible sections:
SectionDescription
Session LifecycleTTL and concurrency limits for transfer sessions by channel.
Default RoutingDefault connection, queue, and priority for agent transfers.
Voice GatewayVoice-specific transfer settings for SIP-based handoffs.
PII HandlingControl how the system handles PII during transfers.

Session Lifecycle

SettingDescriptionDefault
Chat TTLSession timeout for chat channels (minutes).0.5
Email TTLSession timeout for email channels (minutes).4
Voice TTLSession timeout for voice channels (minutes). 0 = end with call.0
Messaging TTLSession timeout for messaging channels (minutes).0.5

Default Routing

Configure the default connection, queue, and priority that the system uses when an agent initiates a transfer. These defaults apply unless the agent or workflow overrides them at runtime.

Voice Gateway

Configure voice-specific transfer settings for SIP-based handoffs, including SIP headers, DTMF relay, and call recording behavior during transfers.

PII Handling

Control how the system handles personally identifiable information during transfers. Configure which PII fields pass through and which the platform masks or redacts when the conversation hands off to a human operator. Click Save to apply changes or Reset to restore all values to defaults.

Agent Assist

The Agent Assist page manages Agent Assist connections for this project. Navigation: ProjectSettingsAgent Assist Project Settings: The Enable Agent Assist for this project toggle controls whether Agent Assist is active for this project. When enabled, human agents receive AI-powered suggestions, knowledge lookups, and response drafts in real time. When disabled, Agent Assist is unavailable for all sessions in the project. Connections Click Add connection to configure a new Agent Assist integration with endpoint and authentication credentials.