Claude Code on Windows — voice, dictation and hands-free setup

Most of the people installing voice tooling for the Claude Code CLI are on Windows, and most of the guides are written on a Mac. This one isn't. Here is how talking to the Claude Code CLI works on Windows 10 and 11: the dictation that ships with the CLI, the WSL caveats, the permission toggle that trips everyone up, and how to go fully hands-free with speech recognition that runs on your own machine.

The built-in dictation on Windows

The Claude Code CLI includes push-to-talk dictation. Type /voice to enable it, hold Space while you speak and release to drop the transcript into the prompt, or switch to tap mode: tap to start, tap again to send. It works in the terminal and in Anthropic's own VS Code extension. On Windows there are three things to know before you rely on it:

• Native Windows works; WSL needs WSLg. If you run the CLI inside WSL, dictation needs WSLg, which ships with WSL2 from the Microsoft Store on Windows 10 and 11. On WSL1, or a distribution without WSLg, Anthropic's advice is to run the CLI natively on Windows instead.
• It needs a Claude.ai sign-in. With an API key, Amazon Bedrock, Google's Agent Platform or Microsoft Foundry configured, /voice isn't available.
• Your audio goes to Anthropic's servers for transcription. It isn't processed on your machine.

It also needs a local microphone, so it doesn't work over SSH or in VS Code Remote sessions such as Dev Containers.

The permission toggle that breaks everything

Windows has a system-wide microphone switch that silently blocks apps. If dictation records nothing, open Settings → Privacy & security → Microphone and check three things: microphone access is on, apps are allowed to use it, and the toggle for desktop apps is on as well. VS Code and terminal apps count as desktop apps, so that last one is usually the culprit. A Bluetooth headset gives you a cleaner signal than a laptop's built-in array, which matters more for wake words than for dictation.

Where dictation stops

Dictation is input. Everything else about working with an agent is still a keyboard job: nothing is spoken back, permission prompts are still typed, and a key has to be held or tapped, so your hands stay near the keyboard by design. If you just want to type less, the built-in mode is enough. If you want to leave the keyboard, you need the rest of the loop.

Fully hands-free on Windows

Verbative is a VS Code extension for Windows 10 and 11 (x64) and macOS that puts a voice loop around the Claude Code CLI. Put on a headset and it goes like this:

1. Say "listen". The wake word is detected on your machine by a small whisper.cpp model, so nothing records until you ask.
2. Dictate your prompt. It's transcribed on your machine with whisper.cpp — on Windows it uses your GPU through Vulkan when there is one, and Verbative picks a dictation model your machine can handle. Speak in any of roughly 99 languages; it's translated to English before it reaches the agent.
3. Say "stop listening". The Claude Code CLI runs the prompt against your codebase.
4. Hear the result. A short summary is spoken back in a natural neural voice, generated locally. Permission prompts are read aloud too, and you answer "approve", "deny" or "explain" out loud.

Say "update" for a spoken status check while the agent works, "stop claude" to abort, and switch between several agents by voice. The Agent Workspace, where each agent runs in its own terminal tile with a working indicator, runs on Windows as well.

Two details for the comparison above: nothing is uploaded — speech is recognized and generated on your machine, and our servers never see audio, prompts or code — and it runs on your existing Claude Pro or Max subscription through the Claude Code CLI. No API key, no SDK, and the free tier activates itself.

Setting it up on Windows

1. Install the Claude Code CLI and sign in, following Anthropic's Windows instructions.
2. Install Verbative from the VS Code Marketplace and click the Verbative "V" in the activity bar. The first time you start the listener, it tells you what it is about to download — the speech engines and models, about 1.3 GB — and keeps them on your machine.
3. Allow the microphone for desktop apps as described above.
4. Click Start listener, wait for "Ready. Say listen to begin.", and say "listen".

The install docs have the current requirements and the troubleshooting page covers microphone and audio issues.

Frequently asked

Does Claude Code voice mode work on Windows?+

Yes. The built-in /voice dictation works natively on Windows 10 and 11 in the terminal and in Anthropic's VS Code extension. Inside WSL it needs WSLg, which comes with WSL2 from the Microsoft Store; on WSL1 run the CLI natively instead.

Why does /voice say it requires a Claude.ai account?+

The built-in dictation only works when a Claude.ai sign-in is the active authentication. If an API key, Amazon Bedrock, Google's Agent Platform or Microsoft Foundry is configured, the command is unavailable. Run /login with a Claude.ai account to use it.

Does Verbative run on Windows?+

Yes. The Verbative VS Code extension supports Windows 10 and 11 on x64 as well as macOS, including the wake word, on-device dictation, spoken replies, voice-answered permissions and the Agent Workspace. The separate Verbative Voice desktop app is macOS-only.

Is dictation on Windows processed locally?+

The Claude Code CLI's built-in dictation streams your audio to Anthropic's servers. Verbative transcribes on your machine with whisper.cpp by default — on Windows through Vulkan on your GPU when there is one — and generates its spoken replies locally too, so no audio leaves the machine.

Which languages can I dictate in on Windows?+

Roughly 99 with Verbative. Whatever you dictate is transcribed and translated to English on your machine before it reaches the Claude Code CLI, with nothing to switch on. The voice commands themselves stay English so they are never ambiguous.