# Configure Agent

Customize your AI voice agent's behavior, voice, and capabilities

## Overview

After creating an agent, you can configure it to match your specific needs. The agent configuration is divided into three main pages:

1. **Configuration:** General settings, voice selection, and advanced options.
2. **Prompt:** Agent description, greeting, instructions, and closing message.
3. **Tools:** Built-in tools and workspace integrations.

## Configuration Page

The Configuration page contains all the core settings for your agent, organized into three sections: General, Voice, and Advanced.

### General Section

#### Agent Name

The name your agent will go by during conversations. This helps identify the agent in your dashboard and can be referenced in prompts.

- Maximum 50 characters
- Choose a name that reflects the agent's purpose

#### AI Model

Select the AI model that powers your agent's responses. Different models offer trade-offs between speed and capability.

| Provider | Models | Notes |
|---|---|---|
| OpenAI | GPT 5.2, GPT 5.1, GPT 5, GPT 5 Mini, GPT 5 Nano, GPT 4.1, GPT 4.1 Mini, GPT 4.1 Nano, GPT 4o, GPT 4o Mini, GPT 4 Turbo, GPT 3.5 Turbo | Wide range of options from fastest to most capable |
| Google | Gemini 3 Pro Preview, Gemini 3 Flash Preview, Gemini 2.5 Flash, Gemini 2.5 Flash Lite | Gemini 2.5 Flash is the fastest option |
| Anthropic | Claude Sonnet 4.5, Claude Sonnet 4, Claude Haiku 4.5, Claude 3.7 Sonnet, Claude 3.5 Sonnet, Claude 3 Haiku | Claude Sonnet 4.5 is the most accurate option |

**Tip:** For most use cases, Gemini 2.5 Flash offers the best balance of speed and quality. Use Claude Sonnet 4.5 when accuracy is critical.

#### Language

The language your agent will speak and understand. You can select from a wide range of supported languages including English, Spanish, French, German, Portuguese, Italian, Dutch, Russian, Chinese, Japanese, Korean, Arabic, Hindi, Turkish, Polish, Swedish, Finnish, Danish, Norwegian, Greek, Czech, Romanian, Hungarian, Bulgarian, Croatian, Slovak, Ukrainian, Vietnamese, Indonesian, Malay, Filipino, Tamil and Urdu.

#### Phone Number

Assign a phone number to this agent for handling inbound calls. When a call comes in to the assigned number, this agent will automatically handle the conversation.

**Note:** Phone number assignment is not available on the free plan. [Learn how to add phone numbers](/how-to/phone-number).

#### Timezone

Set the timezone for your agent. This affects time-based greetings and scheduling functionality.

#### Temperature

Temperature controls the creativity or randomness of responses generated by the AI model. Choose from preset options:

- **Deterministic:** More predictable, consistent responses
- **Creative:** Balanced creativity and consistency
- **More Creative:** Higher variability and creative responses

#### Knowledgebase

Select a knowledgebase to give your agent access to specific information about your business, products, or services. [Learn how to create a knowledgebase](/how-to/knowledgebase).

#### Custom Vocabulary

Add business-specific terms and phrases to improve speech recognition accuracy. This is particularly useful for:

1. Product names and brand terms
2. Technical jargon specific to your industry
3. Proper nouns that may be uncommon

#### Filter Words

Specify words or characters that the agent should not speak. This helps ensure your agent maintains appropriate language. Special characters are allowed.

#### Use Realistic Filler Words

When enabled, the agent will include natural filler words like "uh" and "um" to make conversations sound more human-like and natural.

#### Delete Agent

Permanently delete the agent. This action will erase all personalized data, voice profiles, and integrations associated with the agent.

**Warning:** Deleting an agent is irreversible. Make sure to export any important data before proceeding.

### Voice Section

#### Voice

Choose the voice for your agent from a library of pre-built voices, or clone your own voice. Each voice has characteristics including:

1. **Accent:** Regional accent (e.g., American, British)
2. **Gender:** Male or Female
3. **Languages:** Supported languages for the voice

#### Speech Model

Select the speech synthesis model for voice output. Each model is optimized for different use cases:

| Model | Best For | Features |
|---|---|---|
| Flash | Low latency applications | Fastest response time |
| Turbo | Quality-focused use cases | Better voice quality |
| Conversational | Natural dialogue | Expressive Mode toggle available |
| Multilingual | Multi-language support | Supports multiple languages |

#### Stability

Controls how consistent the voice sounds across utterances.

- **Lower values:** More expressive but less predictable
- **Higher values:** Steadier but less emotional

**Note:** This setting is replaced by Expressive Mode when using the Conversational speech model.

#### Speed

Controls the speed of generated speech.

- **Below 1.0:** Slows down the speech
- **Above 1.0:** Speeds up the speech

Extreme values may affect the quality of generated speech.

#### Similarity

Determines how closely the AI matches the original voice. Higher values provide closer matching but may include unwanted noise from the original recording.

#### Optimize Streaming Latency

Controls the latency optimization level for voice streaming.

- **Lower values:** Faster response but may reduce quality
- **Higher values:** Better quality but increased latency

#### Turn Timeout

The time in seconds to wait for the user to speak before the agent takes another turn. Set to `-1` to disable timeout.

#### Expressive Mode

Available with the Conversational speech model. Enhances your agent with emotionally intelligent speech and improved turn-taking.

When enabled, the agent uses expressive audio tags like [happy], [sad], or [excited] to add emotional nuance to responses.

### Advanced Section

#### Advanced Noise Cancellation

When enabled, advanced mode removes background noise more effectively—like chatter in a café or keyboard typing—to make the user's voice clearer and easier to understand.

#### TTS Output Format

Select the output format for text-to-speech audio. Available options:

| Format | Notes |
|---|---|
| PCM 8000 Hz | Basic quality |
| PCM 16000 Hz | Recommended |
| PCM 22050 Hz | Higher quality |
| PCM 24000 Hz | Higher quality |
| PCM 44100 Hz | Highest Quality |
| PCM 48000 Hz | Professional quality |
| μ-law 8000 Hz | Optimized for telephony |

#### Silence End Call Timeout

The maximum number of seconds since the user last spoke. If exceeded, the call will terminate automatically. Set to `-1` to disable this feature.

#### Limit Call Duration

Sets a maximum duration for calls in minutes. Useful for controlling costs and ensuring calls don't run indefinitely.

#### Enable Recordings

Toggle to record calls for playback and easy review in the call logs. Recordings help with quality assurance and training.

#### Conversations Retention Period

Set the number of days to keep transcriptions and recordings. Set to `-1` for unlimited retention.

## Prompt Page

The Prompt page defines how your agent communicates, including its personality and the messages it uses during conversations.

### Agent Description

A brief description of your agent that helps define its role and purpose. This is used internally and can help you organize multiple agents.

- Maximum 250 characters
- Example: "A customer support agent for urvo"

### Greeting Message

The first message your agent will say when answering calls. This sets the tone for the conversation.

- Maximum 500 characters
- Example: "Hello! How can I help you today?"

**Best Practice:** Keep greetings concise and welcoming. Include the company name or agent purpose when relevant.

### Agent Instructions

Detailed instructions that define your agent's personality, role, and behavior. This is the core prompt that guides how your agent responds.

Example: "You are a helpful AI assistant. Please be friendly and professional in all interactions."

**Generate with AI:** Click the "Generate with AI" button to automatically create instructions based on your agent description.

**Note:** For guidance on writing effective agent instructions, see the [Agent Instructions](/guides/agent-instructions) guide.

### Closing Message

The final message your agent will say before ending calls. Use this to thank callers and provide any final information.

## Tools Page

Tools extend your agent's capabilities by enabling specific actions during conversations. You can add two types of tools: Built-in Tools and Workspace Tools.

### Built-in Tools

Built-in tools are pre-configured actions that your agent can perform. Each tool has configurable settings:

1. **Name:** Custom name for the tool
2. **Description:** What the tool does
3. **Disable Interruptions:** Prevent user from interrupting while tool executes

#### End Conversation

Allows the agent to end the conversation when appropriate. Configure when and how the agent should conclude calls.

#### Detect Language

Enables the agent to detect the language being spoken and respond accordingly. Useful for multilingual support scenarios.

#### Skip Turn

Allows the agent to skip its turn and wait for more user input. Useful when the user is still speaking or gathering information.

#### Transfer to Agent

Transfer the call to another AI agent based on specific conditions. Configure multiple transfer rules with these settings:

1. **Agent:** Select the target agent
2. **Condition:** Define when to transfer (e.g., "user asks about billing")
3. **Delay before transfer:** Milliseconds to wait before transferring
4. **Transfer Message:** Message to play during transfer
5. **Enable First Message:** Play the transferred agent's greeting after transfer

#### Transfer to Number

Transfer the call to a phone number or SIP URI. Two transfer types are available:

1. **Conference:** Creates a conference call with the destination. Supports phone number or SIP URI destination with condition for transfer.
2. **SIP REFER:** Transfers using SIP REFER protocol. Supports phone number or SIP URI destination with condition for transfer.

#### Play Keypad Touch Tone

Allows the agent to play DTMF (touch tone) sounds. Useful for navigating IVR systems or entering codes during outbound calls.

#### Voicemail Detection

Detects when a call reaches voicemail and can take appropriate action. Configure behavior for when voicemail is detected.

### Workspace Tools

In addition to built-in tools, you can add Workspace Tools that connect to external services and APIs. These tools are created in the main dashboard's Tools page and can be assigned to any agent in your workspace.

Workspace tools enable integrations such as:

- CRM lookups and updates
- Appointment scheduling
- Database queries
- Custom API calls
- Third-party service integrations

**Learn More:** See [Add Integrations](/how-to/integrations) for information on creating workspace tools and connecting external services.

## Configuration Best Practices

### Voice Settings

1. Start with recommended defaults
2. Test different voices for your use case
3. Use Conversational model for natural dialogue
4. Enable Expressive Mode for emotional responses

### Prompt Writing

1. Be specific about the agent's role
2. Include examples of desired responses
3. Define boundaries and limitations
4. Keep greetings concise

### Tool Configuration

1. Only enable tools you need
2. Write clear conditions for transfers
3. Test transfer flows thoroughly
4. Use delays to smooth transitions

### Performance

1. Use faster models for simple tasks
2. Enable recordings for quality review
3. Set appropriate timeout values
4. Add custom vocabulary for accuracy

## Next Steps

With your agent configured, you're ready to:

1. [Add a phone number](/how-to/phone-number) to receive calls
2. [Create a knowledgebase](/how-to/knowledgebase) for your agent
3. [Analyze calls](/how-to/analyze-calls) to monitor performance
4. [Review best practices](/guides/best-practices) for optimal results