Skip to main content
The Agent Tab is the primary identity configuration surface for an agent. It controls the name visible in the dashboard, the agent type, the operating language, the voice model, and quick access to the prompt and welcome message. These settings define what the agent is and how it presents itself.

​
Settings Reference

​
Agent Name

The internal name for this agent. Displayed in:
  • The dashboard agent list
  • Call History records (as “Agent Name”)
  • API responses that include agent metadata
  • A/B test result labels
The name is not read aloud to callers. Choose a name that clearly identifies the agent’s purpose for your team (e.g., Appointment Booking - US, Support Tier 1 - EN).

​
Agent Type

The agent type determines the base behavioral mode:
TypeDescription
InboundOptimized for answering incoming calls. Default greeting behavior is configured in the Inbound Tab.
OutboundOptimized for making outbound calls. Begins speaking immediately when the call connects.
BothCan handle inbound and outbound calls. Settings from both modes apply.
Setting the agent type to Inbound or Outbound does not prevent you from assigning the agent to both inbound and outbound routes. The type primarily affects default greeting and response initiation behavior.

​
Language

The primary language the agent operates in. This setting affects:
  • The transcriber (ASR) language model selected for the call
  • The TTS voice options available (voice models are language-specific)
  • The LLM language context (the prompt should be written in the selected language for best results)
Supported languages include English (US, UK, AU), Spanish, French, Portuguese, Hindi, and others. See the dashboard for the current full language list.
Changing the language after the agent is built may require updating the prompt and welcome message to match the new language. Variable names are not language-specific and do not need to change.

​
Voice

The voice model used for text-to-speech output. Each language has a set of available voice models. Click the voice name to preview a sample. Detailed voice configuration (speed, stability, denoising) is in the Audio Tab.

​
Model

The LLM used to generate agent responses. This is a shortcut to the primary model selection; full LLM configuration is in the Engine Tab and LLM Tab.

​
Prompt

A shortcut to the agent’s system prompt. Click Edit Prompt to open the prompt editor. The full prompt is the core instruction set for the agent’s behavior, persona, and task scope.
The LLM Playground is the best place to iterate on prompt changes. Open the Playground from the testing section after editing the prompt here.

​
Welcome Message

The first thing the agent says when a call connects. For inbound calls, this is the greeting the caller hears immediately. For outbound calls, this is what the agent says when the recipient answers. Keep the welcome message concise. Callers disengage if the opening is too long before they can speak.

​
Version Selector

The Agent Tab displays the current version being viewed. Use the version dropdown to switch between published versions and the current draft. Switching versions loads that version’s settings across all Agent Setup tabs.
Editing any setting while viewing a published version creates a new draft based on that version. You must publish the draft to make changes live.

​
Save Changes

Click Save to save changes to the current draft. Changes are not applied to live calls until you publish the draft as a new version.