Task voice
Send audio your application already has to an explicit Task over the OneSay Realtime WebSocket. Delivery is acknowledged and resumable, and results arrive as Task output and Artifacts.
OneSay Platform
Voice for applications and devices, with server-owned authority
OneSay gives your product two voice services — Task voice for audio you submit and Realtime RTC for live rooms — behind one Control API, with identity, permissions and resource ownership decided on the server.
Pre-release: the OneSay Platform and its SDKs are not publicly released. Nothing on this site is an offer of service.
Two voice services, one platform
Task voice and Realtime RTC are separate services. Connecting to a room never starts a Task, and a Task never opens a room.
Realtime RTC
Join rooms to publish and subscribe to audio, exchange data and remote procedure calls, and add Agents that work only on the publications you select.
Built for the clients you ship
JavaScript SDK
One client for Task voice in browsers and Node.js, and Realtime RTC in browsers.
ESP32 SDK
An ESP-IDF component for devices on the same server model, with a stable media-port contract; each SDK's capabilities are listed in the compatibility matrix.
Agents
Create, read and stop Agents explicitly, with an input set you choose and an audit trail on the server.
Console
Manage projects, environments, applications and keys through the same public API your code uses.
Authority stays on the server
Clients are assumed to be public and untrusted. Every decision that matters is made and recorded by OneSay.
- Server-owned scope, Grants and deny-first authorization
- Short-lived tokens for browsers and devices; API keys shown once
- Explicit Consent before a sensitive action, recorded with its decision
- Task usage metered on the server from what actually ran; RTC and Agent usage settlement is still in progress
Start with the Quickstarts
Follow the JavaScript or ESP32 Quickstart in the Developer Portal.