Open VSX Registry
Video could not be loaded. Watch the demo on GitHub.
Say "Hey Claude" to open Claude Code
Works Across Every Major OS
Native support for Windows, macOS, and Linux. On every platform, a lightweight local model handles detection.
Available for Your Editor
Install on VS Code or any compatible editor.
Install from the VS Code Marketplace or the Open VSX Registry. For other editors, download the .vsix from GitHub Releases.
How It Works
From speech to action in milliseconds. All processing happens on your machine.
by Analytics in Motion
Default Wake Phrases
Works out of the box with these three routes. No configuration needed.
Add your own phrases in settings.json. Map any spoken phrase to any VS Code command.
Features
Zero config
No API keys. No accounts. Nothing to install alongside it. Install and go.
Fully local
All speech processing runs on your machine. No audio is recorded, stored, or transmitted.
Custom wake phrases
Map any spoken phrase to any VS Code command. Support for multiple aliases per route.
Smart handoff
The extension releases the microphone when a phrase is detected so the assistant can use it. A configurable cooldown resumes listening automatically.
Cross-platform
Windows, macOS, and Linux. Native audio capture through decibri handles platform differences.
Open source
Apache 2.0 licensed. Inspect the code, contribute, or fork it. Built in the open on GitHub.
Built for Accessibility
Voice control makes your editor more accessible. Wake Word is designed with inclusivity in mind.
Hands-free operation
Control your editor entirely by voice. No keyboard or mouse required to trigger commands, open assistants, or focus the terminal.
Privacy first
No cloud processing means no recordings, no third-party access, and no data retention. Your voice stays on your machine.
Consent-driven
Microphone access requires explicit approval. A clear modal explains exactly what happens before any listening begins. Decline at any time.
Smart mic management
Wake Word automatically releases the microphone when handing off to an assistant, preventing conflicts. Status bar feedback keeps you informed.
FAQ
No. All speech recognition runs locally on your machine. Nothing is recorded, stored, or transmitted.
Yes. Wake Word uses sherpa-onnx with a local model (~17 MB, downloaded on first use) on Windows, macOS, and Linux. The speech engine is packaged with the extension; on Windows it needs the Microsoft Visual C++ Redistributable, which most machines already have.
Yes. Add entries to wakeWord.routes in settings.json. Each entry maps a spoken phrase to any VS Code command.
Yes. Install via the Open VSX Registry or download the .vsix from GitHub Releases.
When a wake phrase is detected, the extension tells its speech engine to close the mic and waits for it to confirm. Only then does the target assistant open and take over. After a configurable cooldown (default 30s), listening resumes.