Creating and Configuring Agents¶

An agent is the core building block of SpeakNode. It defines the personality, behavior, voice, and capabilities of your voice AI assistant.

Creating a New Agent¶

1. Navigate to Agents in the sidebar.
2. Click + Create Agent.
3. Enter a name for your agent (e.g., "Customer Support", "Appointment Scheduler").
4. Click Create.

You will be taken to the agent configuration page.

Configuring the System Prompt¶

The system prompt defines your agent's personality, goals, and behavior rules. This is the most important configuration for your agent.

1. In the System Prompt field, write instructions for your agent.
2. Use markdown headings to structure your prompt (e.g., # Personality, # Goal, # Guardrails).

Selecting the LLM Model¶

The LLM model powers the reasoning and response generation of your agent.

1. Open the LLM section.
2. Select a model from the dropdown (e.g., GPT-4o, GPT-4o Mini, Claude Sonnet).
3. Adjust the Temperature slider:
   • Lower values (0.0 - 0.3): More deterministic, consistent responses. Good for data collection and structured tasks.
   • Medium values (0.3 - 0.7): Balanced creativity and consistency. Good for customer support.
   • Higher values (0.7 - 1.0): More creative, varied responses. Good for casual conversation.

Adding Languages and First Messages¶

Each agent can support multiple languages. For each language, you configure a greeting message and a TTS voice.

1. Open the Languages section.
2. Click + Add Language.
3. Select a language from the dropdown (e.g., English, Spanish, German).
4. Enter the First Message -- this is what the agent says when a conversation starts.
5. Repeat for each language you want to support.

Configuring TTS Voice per Language¶

For each language, you select a text-to-speech voice and configure its settings.

1. Click on a language entry to expand its settings.
2. Select a Voice from the available voices for that language.
3. Adjust Speed (0.5x - 2.0x). A speed of 0.9x - 1.1x is recommended for natural conversation.
4. Optionally select a Background Sound (e.g., office ambience, call center) to make the conversation feel more natural.

Configuring the STT Model¶

The Speech-to-Text (STT) model determines how your agent transcribes the caller's speech.

1. Open the STT section.
2. Select an STT model from the dropdown.
3. Configure language detection settings if your agent supports multiple languages.

Configuring Voice Activity Detection (VAD)¶

VAD controls how the agent detects when the caller starts and stops speaking.

1. Open the VAD section.
2. Adjust the settings:
   • Turn Timeout: How long to wait after the caller stops speaking before the agent responds. Use 5-10 seconds for support, 10-15 seconds for data collection.
   • Silence Threshold: The silence duration that signals the end of a turn.

Adding Tools¶

Tools let your agent perform actions during a conversation, such as looking up data or booking appointments.

1. Open the Tools section.
2. Click + Add Tool to attach an existing tool, or create a new one.
3. Toggle System Tools on or off as needed (e.g., end call, transfer).
4. Configure Transfer Rules if you want the agent to transfer calls to another agent under certain conditions.

See Tools & Webhook Tools for a full guide on creating and configuring tools.

Publishing the Agent¶

Once your agent is fully configured, you need to publish it to make it available for phone calls, the web widget, and the API.

1. Review all your configuration sections.
2. Click the Publish button in the top-right corner.
3. The agent is now live and can receive calls or be embedded on your website.

Summary¶