Skip to main content
Zeptar is a platform for building real-time voice agents. Describe how an agent should behave, preview it by talking to it, and ship versions you can trust. Everything you can do in the dashboard is also available over a clean, versioned REST API.

​
Pick your path

ZeptarAgents

The product: design an agent’s behavior, preview the conversation, and publish versions safely with branches and drafts.

ZeptarAPI

The API: manage agents, start conversations, and automate your workspace from your own backend with an API key.

​
Jump straight in

Build your first agent

Create an agent, write a system prompt, and talk to it in the live preview.

Make your first API request

Create an API key and call the Zeptar API in a couple of minutes.

​
How the pieces fit

  • Agents run a speech pipeline (speech-to-text → language model → text-to-speech → turn-taking). You configure behavior; the runtime orchestrates the call. See the agent overview.
  • Branches, drafts, and versions let you change an agent without touching what’s live. See Versioning.
  • The API mirrors the product surface under /v1, authenticated with a session cookie (browser) or an API key (server-side). See Authentication.