Skip to main content
Three core settings define each agent’s identity and behavior. These settings apply across all tools and conversations.

Agent Goal

A clear description of what the agent is designed to achieve. The agent goal influences the model’s decisions across all tools and interactions.
Example: “Help customers manage their credit card accounts, including viewing transactions, filing disputes, and freezing cards.”

Guardrails

Agent-level restrictions that apply across all tools. Guardrails are verbal descriptions that block specific unwanted behaviors, acting as universal constraints.
Example: “Never provide specific financial advice or investment recommendations.”

Tone of Voice

The communication style the agent adopts in all responses. Tone of voice ensures consistent personality across every interaction.
Example: “Professional but friendly. Use clear, jargon-free language. Be concise but thorough when explaining account details.”