Configuring Your Agent

The Agent Settings panel sits on the right of an Agent’s conversation view. It lets you adjust the Agent while you are talking to it, so you can hear the effect of a change straight away. Use the panel toggle in the top right of the conversation to show or hide it. The panel has three collapsible sections, plus the Agent’s avatar, name, and tags at the top, and icons for Share Profile, Edit Agent, and Delete History.

​

Conversation Setting

Controls how the Agent handles being interrupted.

• Interactive Interruption: lets you cut in while the Agent is speaking.
• Auto Interruption: lets the Agent stop itself and respond when it picks up conversational cues.
• Interruption Sensitivity: how readily an interruption is triggered. Drag right and shorter pauses will interrupt.

​

Behavior Controls

Defines what the Agent is, through its two prompts.

• Starting Message: the greeting spoken at the beginning of every conversation. It sets the tone.
• Description prompt: the longer prompt defining traits, behavior, and interaction style. This is the main lever on how the Agent responds.

​

Identity Controls

Changes how the Agent sounds.

• Agent Voice: pick from your voices and the OpenHome library.
• Agent Language: the language the Agent speaks.

​

Dynamic Agent Updates

Changes in Behavior Controls and Identity Controls apply to the live conversation immediately, with no restart. That makes the panel the fastest way to iterate: adjust the prompt or swap the voice, say something, and hear the result.

​

Conversation Modes

The conversation view supports voice and text together, and you can switch mid-conversation.

• Voice: the microphone button captures speech, and the Audio toggle controls whether the Agent speaks its replies aloud. The waveform button generates an interrupt.
• Text: type into Send a message and press the send arrow. Replies appear in the transcript exactly as in a voice call.

The + button uploads an image into the conversation.

​

Next Steps

• Personality Design: write a compelling character prompt
• Voice and Model Configuration: choose your STT, TTT, and TTS providers
• Agent Memory and Context: inject persistent context into your Agent