Voice Agents

Build agents around the conversation you actually need.

Configure prompt, first message, language, model boundaries, voice, knowledge and tools in a focused builder designed for real-time voice applications.

Start Building Read the Docs

Support AgentDraft configuration

System Prompt

You are a concise voice assistant. Ask one question at a time, use connected tools when needed, and hand off when the caller requests a person.

First Message

Language LLM Provider

STT Provider TTS Provider

Agent Builder

Keep behavior, voice and connected context in one place.

Bring prompts, language, model configuration, voice, knowledge and tools into one focused workspace.

Agent Builder

Shape the agent experience around reusable configuration rather than scattered infrastructure.

System Prompt

Define role, boundaries and conversational behavior in a clear configuration surface.

First Message

Control how the agent opens a conversation and establishes context.

Language

Make language an explicit part of agent configuration.

LLM Configuration

Choose and tune a provider-neutral model layer as supported by your deployment.

Voice Configuration

Configure speech input and output through provider-neutral STT and TTS boundaries.

Knowledge + Tools

Give agents context and controlled actions.

Knowledge augments the conversation with retrieval context. Tools connect business actions through explicit interfaces with scoped inputs and outputs.

Knowledge

Attach relevant knowledge sources for retrieval-augmented conversations.

Tools

Give agents controlled access to application actions and external systems.

AI AgentConversation context

KnowledgeRetrieve relevant context

ToolsExecute controlled actions

Test + Deploy

Validate the experience before it reaches a live channel.

Use a clear testing surface, then connect deployment targets when the underlying runtime integrations are configured.

Testing

Exercise conversational paths before publishing changes.

Deployment

Move tested agent configuration toward connected channels and runtime environments.

Observability

Inspect transcripts, events and execution context after conversations occur.

Caller speaking

TranscriptLIVE DEMO

Caller

I need to update the delivery window for my order.

Agent

I can help with that. I’ll check the order details and available windows.

Caller

Tomorrow afternoon would work best.

Technical eventsDemo

Speech Detected+0 ms

Model Request+148 ms

Tool Called+382 ms

Response Generated+612 ms

TTS Started+744 ms

Interactive product simulation with representative transcript and runtime events.

Voice Configuration

Provider-neutral by default.

Choose provider-neutral model, speech recognition and speech synthesis layers through clear configuration boundaries.

LLM ProviderConfigured providerReasoning boundary

STT ProviderConfigured providerSpeech input boundary

TTS ProviderConfigured providerSpeech output boundary

Configure the conversation. Connect the system.

Start with an agent, then add workflow, telephony, tools and observability as the application grows.

Start Building