{"id":45149482,"url":"https://github.com/shreyaskarnik/voice-mcp","last_synced_at":"2026-02-20T02:30:25.696Z","repository":{"id":337569421,"uuid":"1154041508","full_name":"shreyaskarnik/voice-mcp","owner":"shreyaskarnik","description":"Bidirectional voice MCP server for Claude Code — listen (STT) and speak (TTS) on Apple Silicon via mlx-audio","archived":false,"fork":false,"pushed_at":"2026-02-10T06:44:58.000Z","size":4635,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-02-10T12:52:49.228Z","etag":null,"topics":[],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"apache-2.0","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/shreyaskarnik.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-02-10T00:24:22.000Z","updated_at":"2026-02-10T06:45:02.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/shreyaskarnik/voice-mcp","commit_stats":null,"previous_names":["shreyaskarnik/voice-mcp"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/shreyaskarnik/voice-mcp","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shreyaskarnik%2Fvoice-mcp","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shreyaskarnik%2Fvoice-mcp/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shreyaskarnik%2Fvoice-mcp/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shreyaskarnik%2Fvoice-mcp/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/shreyaskarnik","download_url":"https://codeload.github.com/shreyaskarnik/voice-mcp/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/shreyaskarnik%2Fvoice-mcp/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":29639107,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-02-19T22:32:43.237Z","status":"online","status_checked_at":"2026-02-20T02:00:07.535Z","response_time":59,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":[],"created_at":"2026-02-20T02:30:25.218Z","updated_at":"2026-02-20T02:30:25.691Z","avatar_url":"https://github.com/shreyaskarnik.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# voice-mcp\n\nBidirectional voice MCP server for [Claude Code](https://docs.anthropic.com/en/docs/claude-code). Adds `listen()` and `speak()` tools so Claude can hear you and talk back — all running locally on Apple Silicon via [mlx-audio](https://github.com/Blaizzy/mlx-audio).\n\n```\nUser speaks → listen() → mic + VAD → Voxtral Realtime STT → text to Claude\nClaude responds → speak(text) → Kokoro TTS → AudioPlayer → speakers\n```\n\n## Requirements\n\n- macOS with Apple Silicon (M1+)\n- Python 3.11+\n- [uv](https://docs.astral.sh/uv/)\n- A working microphone and speakers\n\n## Setup\n\n```bash\ngit clone \u003cthis-repo\u003e \u0026\u0026 cd voice-mcp\nuv sync\n```\n\nFirst run downloads the models from HuggingFace (~2.5 GB total):\n\n- **STT**: `mlx-community/Voxtral-Mini-4B-Realtime-2602-int4` (Voxtral Realtime, 4-bit quantized)\n- **TTS**: `mlx-community/Kokoro-82M-bf16` (Kokoro, 82M params)\n\n### Configure Claude Code\n\nThe repo includes `.mcp.json` so Claude Code auto-discovers the server. Just open Claude Code from the project directory:\n\n```bash\ncd voice-mcp\nclaude\n```\n\nOn first use, Claude Code will prompt you to approve the MCP server.\n\nTo use from a **different project**, add to that project's `.mcp.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"voice\": {\n      \"type\": \"stdio\",\n      \"command\": \"uv\",\n      \"args\": [\"--directory\", \"/path/to/voice-mcp\", \"run\", \"server.py\"]\n    }\n  }\n}\n```\n\n## Usage\n\n### Voice input\n\nTell Claude \"listen to me\" or use the `/listen` slash command. You'll hear a rising chime when the mic is active — speak naturally, and recording stops automatically after 1.5s of silence (falling chime).\n\n### Voice output\n\nClaude will speak conversational responses automatically. Code, data, and technical details stay in the terminal.\n\n### Change voice\n\nUse `/voice` to browse all 54 available voices across 9 languages, or `/voice am_echo` to switch directly.\n\n## Tools\n\n### `listen(duration?)`\n\nRecords audio from the microphone and returns a transcription.\n\n- **Default**: VAD-based — waits for speech, stops after silence\n- **With `duration`**: fixed-length recording (seconds)\n- **STT languages**: ar, de, en, es, fr, hi, it, ja, ko, nl, pt, ru, zh\n\n### `speak(text, voice?, speed?, lang?)`\n\nSpeaks text aloud through the computer's speakers.\n\n- **voice**: Kokoro voice ID (default: `af_heart`). See `/voice` for options\n- **speed**: playback speed multiplier (default: 1.0)\n- **lang**: language code — `a` American English, `b` British English, `e` Spanish, `f` French, `h` Hindi, `i` Italian, `j` Japanese, `p` Portuguese, `z` Mandarin\n\n## How it works\n\nThe server runs as a stdio subprocess managed by Claude Code. Audio I/O (mic + speakers) happens in the server process; only text crosses the MCP protocol.\n\n- **STT**: Voxtral Realtime (4B params, int4) — streams audio through a causal encoder-decoder with adaptive RMS normalization\n- **TTS**: Kokoro (82M params, bf16) — ALBERT text encoder → prosody predictor → iSTFTNet vocoder\n- **VAD**: webrtcvad (mode 3) with energy-based fallback\n- **Audio cues**: synthesized tones via sounddevice — rising chime = listening, falling chime = done\n- **Notifications**: macOS banners via hooks when listening starts/stops and when speaking\n\nModels are pre-loaded at server startup via FastMCP's lifespan hook, so the first tool call is fast.\n\n## Demos\n\n\u003e Turn your sound on — this demo has voice!\n\n### Claude Code\n\n\u003chttps://github.com/user-attachments/assets/22493ae6-2da9-4a06-b8ef-be203c98b35d\u003e\n\n### OpenAI Codex\n\n\u003chttps://github.com/user-attachments/assets/09b6dceb-d83d-46a9-a2fd-d66fa3bc1186\u003e\n\n### Mistral Vibe\n\n\u003chttps://github.com/user-attachments/assets/7530373f-719d-46c5-8a35-2337e8121356\u003e\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshreyaskarnik%2Fvoice-mcp","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fshreyaskarnik%2Fvoice-mcp","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fshreyaskarnik%2Fvoice-mcp/lists"}