Agent settings

Each agent card on the Agents page contains the configuration that applies across website chat, Slack, and Automation actions.

Configuration fields

Field	Meaning	Guidance
System prompt	Persistent administrator instruction.	Define role, boundaries, uncertainty behavior, and style. Do not paste secrets.
Model	Approved model used for generation.	Keep the default unless quality, latency, or cost testing supports a change.
Max tokens	Maximum generated output size.	Bound it to the task. A higher value does not improve factuality.
Temperature	Variation from 0 to 1.	Use a lower value for repeatable operational answers; higher for ideation.
Team members	Organization members who auto-join every new website chat room.	Add only people responsible for those conversations.
Enabled	Whether the agent can participate.	Disable before risky prompt or knowledge changes, then retest.

Select Save to persist changes. An unsaved field is not part of the active agent configuration.

Name and avatar

The name is established when the agent is created. Select the avatar on the agent card to open Change avatar, then choose Save avatar. Use an identity that distinguishes the agent from a human teammate.

Disable versus retire

Use Enabled as the reversible operational control. Turning it off pauses the agent without removing its prompt, knowledge, team, or channel configuration.

Use Retire only when the identity should leave every team and conversation it serves. The confirmation asks you to type the agent name. Retirement cannot be undone from this page.

Limits and validation

• The organization can have only the number of agents allowed by its limit.
• Temperature accepts values from 0 through 1.
• Team members must belong to the same organization.
• A disabled agent cannot be selected for a new Agent message action.
• Slack channel choices for an Automation are limited to enabled channels bound to the selected agent.