An open standard for voice-first applications.
Device-agnostic · Model-agnostic · Platform-agnostic
Visual interfaces have the DOM. Accessibility has ARIA. Voice has nothing.
VIP is building that missing layer — an open specification that standardizes how any application exposes itself to any voice agent, on any device, without bespoke integration work per app or per model. The screen becomes secondary. The voice becomes primary. Any device with a mic, a speaker, and a network connection is a valid client.
| Repository | Description |
|---|---|
| docs | The specification source and |
| documentation site |
More repositories are coming as the protocol matures — reference implementations, SDK libraries, and conformance test suites.
The v0.1 draft covers the full core interaction model: session lifecycle, the Voice Interaction Tree, Action Registry, interaction primitives, and the state machine. Message format schemas and transport layer specs are being written next.
Read the spec → voiceinteractionprotocol.io
This is an open effort. A standard only works if it is built with the people who will implement it.
- Discuss → GitHub Discussions
- Contribute to the spec → docs repository
- Report an issue → open an issue
Copyright 2026 Voice Interaction Protocol Contributors · Spec licensed under CC BY 4.0 · Code licensed under Apache 2.0