> For the complete documentation index, see [llms.txt](https://dataqueue.gitbook.io/voicehub-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://dataqueue.gitbook.io/voicehub-docs/getting-started/create-your-first-agent.md).

# Create your first agent

## Step 1: Create Your Voice Agent

1. Click the **"New Agent"** button to get started.
2. Choose a template that fits your use case, or start from a **blank template** if you want to build from scratch.
3. Fill in the **Agent Title** and **Description**.
   * Example: *“Restaurant Assistant” – a voice agent that helps customers book tables and ask menu questions.*

<figure><picture><source srcset="/files/pv89j55EhD0Om4i6whhP" media="(prefers-color-scheme: dark)"><img src="/files/OTfOzGcsyWCNI4lkNGmE" alt=""></picture><figcaption></figcaption></figure>

## Step 2: Agent Configuration

After creating your agent, the first step is to choose how you want to configure it. You have two configuration modes, depending on your needs:

### Option 1: DQ Mode (Arabic/English/Dutch-Optimized Stack)

* This is the default mode when creating a new agent.
* Uses DataQueue’s fully optimized models for speech recognition (STT), conversation logic (LLM), and speech synthesis (TTS).
* Designed for MENA use cases where accuracy, latency, and sentiment detection matter most.
* Locked: you cannot select or override providers manually.
* Voice setup and tuning is handled via the DQ Configs tab.

### Option 2: Custom Mode (Provider Flexibility)

* Offers full flexibility in model selection and configuration.
* Choose from a wide range of supported providers for:
  * STT (e.g., Deepgram, Gladia, Azure)
  * LLM (e.g., OpenAI, Claude, Cohere)
  * TTS (e.g., ElevenLabs, Cartesia, Azure)
* Great for enterprise developers or use cases where custom prompts or models are preferred.

## Switching Between Modes

<figure><picture><source srcset="/files/kR16Xc2qOmGHDyOQtOHw" media="(prefers-color-scheme: dark)"><img src="/files/4NIJqSdYshUpttIbj7Yl" alt=""></picture><figcaption></figcaption></figure>

In the Configuration page (top right):

* Select the language (e.g., Arabic, English)
* Toggle between DQ Mode and Custom Mode

When using Custom Mode, selecting a language automatically filters the models shown by the provider to only include those that support the selected language. This ensures you're always choosing from compatible options.

## Custom Mode Tabs

When in Custom Mode, you’ll have access to the following configuration tabs:

* **Models** – Choose your STT and LLM providers
* **Voices** – Choose your TTS provider. Preview TTS voice options (Includes a field to test voices live text-to-audio samples)
* **Pathway** – Build your agent's logic visually or with a global prompt
* **VoIP** – Assign phone numbers (inbound/outbound)
* **Analysis** – Define summaries, resolution tags, sentiment rules
* **Widget** – Add and customize the embedded web voice interface
* **White Labeling** – Apply brand and domain settings for your team/client

Use [Provider Keys](/voicehub-docs/agent-design-and-usage/provider-keys-custom-mode-only.md) to enable these integrations.

## DQ Mode Tabs

When in DQ Mode, configuration is simplified and optimized:

* **DQ Configs** – Select between Arabic, English, and Dutch voice
  * Customize Speech delivery using:
    * **Stability** – Higher values lead to more uniform and stable speech, though it may lose expressiveness. Lower settings introduce more vocal variety and emotion, but can make the delivery less controlled.
    * **Similarity** – Boosting this parameter sharpens the voice and brings it closer to the intended tone. However, pushing it too far can introduce unwanted glitches, so moderate tuning is key for a natural sound.
    * **Speed** – Changes how fast the voice talks. Less than 1 is slower, more than 1 is faster. Too low or high can sound off.
* **Pathway** – Same as above, build logic or use a global prompt
* **VoIP, Analysis, Widget, White Labeling** – Same as in [Custom Mode](#option-2-custom-mode-provider-flexibility)

You won’t see **Models** or **Voices** tabs here — they’re preconfigured under the hood.

## Prompt vs. Pathway: Two Ways to Build Logic

Once you’ve selected your mode (DQ or Custom), you can define how the agent should behave. VoiceHub supports two approaches to building logic:

### Option 1: Global Prompt

A single prompt that guides the agent’s entire behavior — similar to system prompts in traditional LLM apps.

Use when:

* You want a **quick setup** for straightforward tasks
* Your agent only needs to answer **general questions** or operate reactively
* Your use case is **simple and reactive**

### Option 2: Conversational Pathway

A visual drag-and-drop builder to define complex flows, variables, and decision logic using connected nodes.

Use when:

* You need **branching logic** (e.g., verification → escalate → book → end)
* You want to **extract variables** (e.g., date, location)
* You want fine **control over what the agent says** and when

### Can they be combined?

Yes. You can start with a global prompt, then build out a Conversational Pathway later — or use the pathway as the primary logic and include a global fallback prompt.

This gives you flexibility for both rapid prototyping and scalable complexity.