Create Agent Project

New

Configure an autonomous in-browser voice copilot with native language support, custom route perception, and static command audio.

1. Core Identity & Language

Define your agent's display name and primary interaction language.

Injects canonical language instructions into the agent's prompt.

2. Google Gemini Models & API Key

Provide your provider credentials and select orchestrator models.

Encrypted securely at rest. Used for real-time model reasoning windows.

Primary model executing DOM tool calls and conversational turns.

Secondary model summarizing chat turns and performing audits.

3. Persona Guidelines & Audio Interaction

Define system prompt behavior and upload pre-recorded introductory audio.

Played instantly to users upon WebSocket connection, eliminating high-latency realtime TTS cold starts.

4. Route Guide & Site Map

Guide the agent's behavior based on the user's active browser URL.

5. Anonymous SDK Testing Token Limit

Allow zero-configuration SDK testing via public endpoint. Set to 0 to require API keys.

Cancel