claude-speak

claude-speak

Give your Claude Code agent a voice. 🗣️

claude-speak reads Claude's replies aloud and plays subtle sound cues as it works — so you can listen to a long session instead of watching the terminal.

Features

• 🗣️ Spoken replies — Claude's responses read aloud, in real time
• 🔊 Sound cues — gentle audio for sending, tool calls, permission prompts, and exit
• ⏯️ Playback controls — skip, replay, pause, mute, and switch voices on the fly
• 🎙️ Great voices — ElevenLabs or OpenAI, with a built-in macOS voice as fallback
• 🧵 Per-session — each session has its own queue; two terminals never talk over each other

Install

/plugin marketplace add mmsaki/claude-speak
/plugin install claude-speak@mmsaki

Restart Claude Code — that's it. It speaks with the built-in macOS voice out of the box (no API key, no cost).

Want premium voices? (optional) Add a key, then switch the engine:

printf 'YOUR_ELEVENLABS_KEY' > ~/.claude/.elevenlabs_key && chmod 600 ~/.claude/.elevenlabs_key
# then, in Claude:   /claude-speak:engine elevenlabs
# (OpenAI works too: ~/.claude/.openai_key + /claude-speak:engine openai)

Commands

Command	What it does
/claude-speak:next · :prev	skip forward / back
/claude-speak:replay	replay the current segment
/claude-speak:stop · :play	pause / resume
/claude-speak:mute	silence the voice (cues keep playing)
/claude-speak:voice <name>	switch voice
/claude-speak:voices	list available voices
/claude-speak:list	show what's queued

---

macOS · MIT · full docs · report an issue