Short answer: type /voice in Claude Code, then hold Space while you talk and let go. Your words appear in the prompt as you speak, and you press Enter to send. Prefer to tap? /voice tap starts recording with one tap and sends with the next. It needs a claude.ai sign-in (not an API key) and a microphone on the computer Claude Code runs on.
Everything below is from Anthropic's own documentation, checked on November 9, 2026. (I build Morphy, whose voice typing works with every AI and runs on your own computer; it comes up near the end.)
What Claude Code's voice mode is
It's dictation: you speak your prompt instead of typing it, and Claude still answers in text. The transcription is tuned for code, so words like regex, OAuth, JSON and localhost come out right, and your project's name and git branch are used as hints. You can mix typing and talking in the same message.
- It runs in Anthropic's cloud. Your audio streams to Anthropic for transcription; nothing is processed on your computer.
- It's free to use: transcription doesn't use Claude messages or tokens, and doesn't count toward your plan's limits.
- Where it works: the Claude Code command line on Windows, macOS and Linux, and the VS Code extension. Not in cloud sessions, over SSH, or in VS Code remote sessions.
Turn it on
/voice
Voice mode enabled (hold). Hold space to record. Dictation language: en (/config to change).
| Command | What it does |
|---|---|
/voice | Turn it on or off, in the mode you used last |
/voice hold | Push to talk: hold the key while you speak |
/voice tap | Tap to start, tap again to send |
/voice off | Turn it off |
It stays on across sessions. The first time, Claude Code checks your microphone; on a Mac, macOS asks whether your terminal may use it.
Hold or tap?
Hold (the default) is push-to-talk. Hold Space: after a short warm-up the footer says listening…, and your words appear dimmed until you let go. Then the text is in the prompt, and you press Enter to send, or hold again to add more. A single tap of Space still types a space.
Tap has no warm-up. With an empty prompt, tap Space (the footer says ● REC · tap to send), speak, and tap again: Claude Code sends the message if it's at least three words long. Recording also stops by itself after 15 seconds of silence or two minutes in all.
Esc or Ctrl+C throws a recording away and puts the prompt back as it was.
To send automatically in hold mode too, add "autoSubmit": true to the voice settings:
{
"voice": {
"enabled": true,
"mode": "hold",
"autoSubmit": true
}
}
Use another key
The key is the voice:pushToTalk action in ~/.claude/keybindings.json. A modifier combination like meta+k starts recording on the first press, with no warm-up:
{
"bindings": [
{
"context": "Chat",
"bindings": {
"meta+k": "voice:pushToTalk"
}
}
]
}
In hold mode, don't use a plain letter: the letter would type into the prompt while it warms up.
Talk in another language
Dictation follows Claude Code's language setting (in /config), and defaults to English. It understands Czech, Danish, Dutch, English, French, German, Greek, Hindi, Indonesian, Italian, Japanese, Korean, Norwegian, Polish, Portuguese, Russian, Spanish, Swedish, Turkish and Ukrainian.
{
"language": "dutch"
}
A language that isn't on the list falls back to English for dictation (Claude still answers in your language).
Voice mode not available? The fixes
| What you see | What it means | The fix |
|---|---|---|
Unknown command: /voice | You aren't signed in with claude.ai, or an API key or other provider takes priority | /login; remove ANTHROPIC_API_KEY, ANTHROPIC_AUTH_TOKEN or an apiKeyHelper if one is set, then restart |
Voice mode requires a Claude.ai account | No usable claude.ai sign-in was found | Run /login again |
Voice mode is disabled by your organization's policy | Your admin turned it off | Ask your organization's admin |
Microphone access is denied | Your terminal may not use the microphone | Windows: Settings → Privacy & security → Microphone, allow desktop apps. Mac: System Settings → Privacy & Security → Microphone, turn on your terminal |
No audio detected from microphone | It recorded silence | Pick the right input device in your system's sound settings |
Voice mode requires SoX for audio recording (Linux) | No recorder could start | Install SoX, for example sudo apt-get install sox |
Voice mode could not find a working audio recorder in WSL | WSL's audio needs SoX's PulseAudio part | sudo apt install sox libsox-fmt-pulse (and WSL2 with WSLg; on WSL1, use Claude Code on Windows itself) |
Voice connection failed | The recording didn't reach the transcription service | Check your network, then try again |
No speech detected | Audio arrived, but no words were recognized | Speak closer to the microphone, and check the dictation language |
| Holding Space types spaces | Voice is off, or your terminal doesn't repeat held keys | Run /voice hold, or switch to /voice tap |
On a Mac where your terminal isn't listed under Microphone at all, reset its permission with tccutil reset Microphone com.apple.Terminal (or your terminal's identifier), quit the terminal completely and run /voice again.
Other ways to talk to Claude Code
- Your computer's own dictation types into any terminal: Win+H on Windows, Dictation on a Mac. It works with an API key too.
- Your phone's keyboard dictation, in the Claude app's Code tab, for Claude Code sessions you reach from your phone.
In Morphy
Morphy has voice typing for every chat, on any AI: Claude Code, Codex, Copilot, Gemini and the rest. Click the microphone in a chat's box or hold Ctrl+Shift+Space, and talk. The difference from Claude Code's voice mode: speech recognition runs on your computer (NVIDIA's Parakeet model, a one-time 670 MB download), so nothing you say leaves it, it works offline, it works with an API key or a free model, and it understands 25 languages. On Pro, Clean up with AI takes out the ums and false starts and fixes names like "cloud code" for Claude Code. See Voice typing.
Questions
Does Claude Code have voice mode? Yes: /voice turns on voice dictation in the Claude Code command line and the VS Code extension. You speak your prompt; Claude answers in text.
Why is Claude Code voice mode not available? Usually because you're signed in with an API key or through a cloud provider instead of a claude.ai account. Voice mode needs a claude.ai sign-in: run /login, and remove ANTHROPIC_API_KEY if it's set.
Does Claude Code voice mode cost anything? No. Transcription doesn't use Claude messages or tokens and doesn't count toward your usage limits, though Claude Code itself needs a paid plan.
Can Claude Code talk back? Voice mode only covers what you say: Claude's answers stay text.
Does voice mode work on Windows? Yes, in Claude Code on Windows itself. In WSL it needs WSL2 with WSLg and SoX's PulseAudio backend.
Related guides
- How to use Claude Code on your phone
- Claude Code not working? How to fix it
- How to install and use Claude Code on Windows
- The best Claude Code GUIs
References
- Anthropic: voice dictation, keyboard shortcuts, settings
- Morphy: Voice typing




