{"$schema":"https://wellknown.network/schemas/agent-record-v1.json","schemaVersion":"1","id":"ag_7cfrs56edq4a","handle":"voxtral-mcp","url":"https://wellknown.network/agents/voxtral-mcp","links":{"self":"https://wellknown.network/agents/voxtral-mcp/record.json","html":"https://wellknown.network/agents/voxtral-mcp","markdown":"https://wellknown.network/agents/voxtral-mcp/record.md","api":"https://wellknown.network/api/v1/agents/voxtral-mcp","status":"https://wellknown.network/api/v1/agents/voxtral-mcp/status","claim":"https://wellknown.network/agents/voxtral-mcp/claim","claimApi":"https://wellknown.network/api/v1/claims","claimDescriptor":"https://wellknown.network/agents/voxtral-mcp/claim.json","badge":"https://wellknown.network/agents/voxtral-mcp/badge.svg","openapi":"https://wellknown.network/openapi.json","history":"https://wellknown.network/api/v1/agents/voxtral-mcp/history","tools":"https://wellknown.network/api/v1/agents/voxtral-mcp/tools"},"ard":{"identifier":"urn:air::server:voxtral-mcp","type":"application/mcp-server-card+json"},"kind":"mcp_server","declared":{"name":"voxtral-mcp","summary":"MCP server wrapping Mistral Voxtral 4B TTS (via mlx-audio) for Claude Code on Apple Silicon — in-process, streaming, gap-free playback","description":"# voxtral-mcp\n\nLocal voice for any MCP client (Claude Code, Claude Desktop, Cursor, etc.) via\n[Mistral Voxtral 4B TTS](https://huggingface.co/mistralai/Voxtral-4B-TTS-2603)\n(MLX 4-bit). Higher voice quality than the lightweight pocket-tts route, at the\ncost of a larger model and slower TTFA.\n\n- 9 languages: 🇬🇧 English, 🇫🇷 French, 🇩🇪 German, 🇪🇸 Spanish, 🇮🇹 Italian, 🇵🇹 Portuguese, 🇳🇱 Dutch, 🇮🇳 Hindi, 🇸🇦 Arabic\n- 4 B parameters, 4-bit MLX quantization (~2.5 GB on disk)\n- ~2.4× real-time generation on Apple Silicon M-series\n- **TTFA ~1 s** thanks to native streaming via `mlx-audio` `stream=True`\n- Non-blocking `speak()`, gap-free playback via `sounddevice` write-mode\n- ~3 GB resident RAM once the model is loaded\n\n> ⚠️ **Licence**: the Voxtral model itself is distributed by Mistral under\n> **CC BY-NC 4.0** — non-commercial use only. This wrapper's code is MIT.\n\n## Requirements\n\n- **macOS Apple Silicon** (M1/M2/M3/M4) — required, MLX doesn't run on Intel\n- **≥16 GB RAM** recommended (the 4-bit model keeps ~3 GB resident)\n- Python 3.10 – 3.14\n\n## Install\n\n```bash\nuvx voxtral-mcp --help\n```\n\nOr persistent:\n\n```bash\nuv tool install voxtral-mcp\n```\n\nThen add to your MCP client's `.mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"voxtral\": {\n      \"command\": \"uvx\",\n      \"args\": [\"voxtral-mcp\"]\n    }\n  }\n}\n```\n\n## MCP tools\n\n| Tool | Purpose |\n|---|---|\n| `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. |\n| `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. |\n| `status()` | Report model load state, queue depths, sample rate, last error. |\n\n## Configuration\n\nAll env vars (in the `…","publisher":{"name":"Vincent Caudron","url":null},"homepage":"https://github.com/Vincweb/voxtral-mcp","repository":"https://github.com/Vincweb/voxtral-mcp","version":"0.7.1","license":"MIT","protocols":["mcp"],"tags":["apple-silicon","claude","claude-code","mcp","mistral","mlx","speech","tts","voice","voxtral"],"pricing":null,"endpoints":[{"url":"pypi:voxtral-mcp","type":"package_pypi","auth":null,"probeable":false}],"skills":null,"tools":null,"extra":null,"attribution":{"kind":"pypi","name":"pypi","license":"pypi","repoUrl":"pypi","summary":"pypi","version":"pypi","description":"pypi","homepageUrl":"pypi","publisherName":"pypi"}},"derived":{"capabilities":[{"slug":"media.speech-synthesis","name":"Speech Synthesis","confidence":1,"provenance":"declared"},{"slug":"ai.model-access","name":"Model Access","confidence":0.905,"provenance":"derived"}],"categories":["ai","media"],"language":"en"},"observed":{"status":"unknown","statusReason":"Distributed as a package to run locally; no network endpoint to check.","lastOkAt":null,"lastProbedAt":null,"statusComputedAt":null,"reliability30d":null,"latestObservations":[],"tools":null,"package":{"name":"voxtral-mcp","registry":"pypi","observedAt":"2026-10-22T23:26:25.396Z","publishedAt":"2026-10-04T01:54:09.355389Z","latestVersion":"0.7.1"},"toolSurface":null,"endpointFacts":[]},"verification":{"claimed":false,"claimedAt":null,"proofs":[]},"provenance":{"sources":[{"source":"pypi","key":"voxtral-mcp","url":"https://pypi.org/project/voxtral-mcp/","firstSeenAt":"2026-09-10T15:23:55.763Z","fetchedAt":"2026-10-22T23:24:56.994Z","normalizedAt":"2026-10-22T23:24:56.994Z"}]},"firstSeenAt":"2026-09-10T15:23:55.763Z","updatedAt":"2026-10-22T23:26:25.396Z"}