voxtral-mcp 0.5.0

MCP server wrapping Mistral Voxtral 4B TTS (via mlx-audio) for Claude Code on Apple Silicon — in-process, streaming, gap-free playback

voxtral-mcp

Local voice for any MCP client (Claude Code, Claude Desktop, Cursor, etc.) via Mistral Voxtral 4B TTS (MLX 4-bit). Higher voice quality than the lightweight pocket-tts route, at the cost of a larger model and slower TTFA.

• 9 languages: 🇬🇧 English, 🇫🇷 French, 🇩🇪 German, 🇪🇸 Spanish, 🇮🇹 Italian, 🇵🇹 Portuguese, 🇳🇱 Dutch, 🇮🇳 Hindi, 🇸🇦 Arabic
• 4 B parameters, 4-bit MLX quantization (~2.5 GB on disk)
• ~2.4× real-time generation on Apple Silicon M-series
• TTFA ~2 s thanks to native streaming via mlx-audio stream=True
• Non-blocking speak(), gap-free playback via sounddevice write-mode
• ~3 GB resident RAM once the model is loaded

⚠️ Licence: the Voxtral model itself is distributed by Mistral under CC BY-NC 4.0 — non-commercial use only. This wrapper's code is MIT.

Requirements

• macOS Apple Silicon (M1/M2/M3/M4) — required, MLX doesn't run on Intel
• ≥16 GB RAM recommended (the 4-bit model keeps ~3 GB resident)
• Python 3.10 – 3.13

Install

uvx voxtral-mcp --help

Or persistent:

uv tool install voxtral-mcp

Then add to your MCP client's .mcp.json:

{
  "mcpServers": {
    "voxtral": {
      "command": "uvx",
      "args": ["voxtral-mcp"]
    }
  }
}

MCP tools

Tool	Purpose
speak(text, voice?, interrupt?)	Generate audio for text and queue it for background playback. Returns immediately, streaming generation. By default, calls queue and play sequentially (including across turns). Pass interrupt=True to abort current playback and clear the queue first.
stop_speaking()	Stop current playback, drop queue, cancel in-flight generation. Use for explicit "mute" requests. For mid-turn interruption + new speech, use speak(..., interrupt=True) instead.
status()	Report model load state, queue depths, sample rate, last error.

Configuration

All env vars (in the env block of .mcp.json):

Variable	Default	Notes
VOXTRAL_MODEL	mlx-community/Voxtral-4B-TTS-2603-mlx-4bit	Any Voxtral MLX model on HF (4-bit / 6-bit / bf16)
VOXTRAL_STREAMING_INTERVAL	2.0	Approx. seconds of audio per streaming chunk
VOXTRAL_MAX_TOKENS	4096	Generation cap (in audio tokens, not characters)
VOXTRAL_SAMPLE_RATE	24000	Output sample rate

Voices

Voxtral ships with 20 preset voices across 9 languages. A non-exhaustive sample:

Language	Voices
English	casual_male, casual_female, cheerful_female, neutral_male, neutral_female
French	fr_male, fr_female
Spanish	es_male, es_female
German	de_male, de_female
Italian	it_male, it_female
Portuguese	pt_male, pt_female
Dutch	nl_male, nl_female
Arabic	ar_male
Hindi	hi_male, hi_female

Pass voice="fr_female" in your speak() call to switch.

Claude Code users

If you're on Claude Code (CLI, desktop, or via Cursor), install the bundled plugin (MCP wiring + /voice-mode skill) in one shot:

/plugin marketplace add Vincweb/voxtral-mcp
/plugin install voxtral@vincweb-tools

See the main repo for architecture details and a side-by-side comparison with kyutai-tts-mcp (smaller / faster TTFA / permissive licence).

License

MIT for this wrapper. The underlying Voxtral model is CC BY-NC 4.0 (non-commercial) — see Mistral.