Skip to main content

​
What you can do

The Zeptar API exposes the same primitives the Zeptar product surface uses. With it, you can:
  • Manage agents — create, list, update, duplicate, and delete conversational AI agents.
  • Start preview conversations — dispatch a LiveKit room and join an agent for a voice call.
  • Manage workspaces — list the organizations a user belongs to, look up a workspace by slug, add and remove members.
  • Browse voices — list the voice models available for speech synthesis.

​
Base URL

For local development, the API runs on http://localhost:3001 (start with pnpm dev from the repo root).

​
Conventions

  • REST + JSON. Every endpoint accepts and returns JSON.
  • Versioned paths. All endpoints are prefixed with /v1. Breaking changes will bump to /v2.
  • Snake-case in payloads, camelCase in URLs. Path parameters use camelCase ({agentId}, {orgId}); response bodies use snake_case (agent_id, system_prompt) to match the Python agent runtime.
  • ISO-8601 timestamps. All createdAt / expiresAt fields are ISO strings, not Unix timestamps.
  • Idempotent reads. GET calls never mutate state. Repeating a GET always yields the same response (modulo concurrent writes).

​
Organization scope

Most resources (agents, conversations, members) are scoped to the requester’s active organization, resolved from the session cookie. To act on a specific organization, either set the active workspace via the web UI or pass orgId explicitly in the path (e.g. /v1/organizations/{orgId}/members).

​
OpenAPI

The full OpenAPI 3.0 specification is generated from the live NestJS controllers and updated on every release.

Zeptar API OpenAPI spec

Download the machine-readable spec