What is Wirevox?
Wirevox is a platform for building and operating AI agents that communicate with people through voice calls or website chat. A Wirevox agent combines conversational instructions with a model, a voice when applicable, business knowledge, and controlled actions.
Wirevox is designed for teams that need more than a standalone chatbot. It provides one place to configure agents, connect business systems, review conversations, manage contacts, and track usage.
What you can build
Wirevox supports two agent channels:
- Voice agents handle spoken conversations. Depending on their configuration and deployment, they can receive or place calls.
- Chat agents handle text conversations through a website deployment or supported chat experience.
Channel selection affects which configuration and functions are valid. For example, a voice agent requires a voice configuration, while a chat agent does not.
The main building blocks
Every agent begins with a core configuration and can be extended with other workspace resources.
| Component | Purpose |
|---|---|
| System prompt | Defines the agent's identity, objective, conversation behavior, boundaries, and instructions for using available functions. |
| Variables | Store reusable values that can be resolved inside supported prompt fields. |
| Model | Generates the agent's conversational responses and tool decisions. |
| Voice | Converts responses into speech for voice agents. |
| Knowledge base | Provides retrieved information from approved documents, text, or website sources. |
| Function | Gives an agent a bounded action such as gathering information, transferring a call, running a workflow, or using a connected integration. |
| Integration | Authorizes Wirevox to communicate with an external service. An integration connection and an agent function are separate configurations. |
These resources are owned by a workspace. Access to view or change them depends on the current user's workspace permissions.
Drafts and published configuration
Agent configuration changes are made in a draft. A draft can differ from the configuration currently serving live conversations.
Publishing promotes the reviewed draft into the agent's active configuration. Editing a draft does not automatically publish it. This separation allows a team to prepare and test changes before they affect customers.
Always distinguish between a successful draft update and a successful publish. A saved draft can contain unpublished changes.
How knowledge and actions differ
Knowledge and functions solve different problems:
- Use a knowledge base when an agent needs to find and explain information.
- Use a function when an agent needs to perform a controlled operation.
- Use an integration when that operation requires access to an external provider.
Attaching a function makes it available to the agent runtime. The system prompt must still contain appropriate instructions explaining when and how the agent should use that function.
What Vox does
Vox is the internal Wirevox platform assistant. It can explain the platform, inspect workspace information the current user is authorized to access, and prepare supported changes.
Vox follows the same workspace boundary as the user. Read operations can run automatically when permitted. Changes that require approval are presented for review before execution. An approval request does not mean the change has completed.
Usage and credits
Wirevox measures billable platform activity in credits. Voice conversations, supported tools, website chat, and Vox can consume credits using their applicable rates.
Credit pricing and usage records are separate from agent configuration. Changing an agent does not itself imply that it has been published or that a conversation has occurred.
A typical workflow
- Create or select a workspace.
- Create an agent and choose its channel.
- Define the agent's purpose and complete system prompt.
- Select a compatible model and, for voice agents, a voice.
- Add only the knowledge and functions the agent needs.
- Test the draft configuration.
- Review and publish the agent.
- Monitor conversations, outcomes, errors, and credit usage.
- Improve the draft and publish another reviewed version when needed.
Continue learning
- Read How Wirevox is organized for the relationship between workspaces, agents, and shared resources.
- Read Core terminology for the exact meaning of terms used throughout this documentation.