speaklna mcp
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
speaklna
A thin command-line shell that lets AI coding tools on a Mac talk through the Speaklna app (local text-to-speech).
Supported: Claude Code, Codex, OpenCode, Grok Build (grok), OpenClaw.
It does three things:
speaklna mcp— a stdio MCP server with five tools:speak,save,stop,status,voices.speaklna hook <tool>— a turn-end hook. The model ends each reply with one line starting with🔊; the hook reads that line and sends it to the app. Secrets are redacted, the hook never prints and always exits 0.speaklna setup/speaklna uninstall— write (and cleanly remove) the MCP entry, the hook and a short rules file for each tool.
AI tools can only speak, save speech as a WAV file, stop, check status and list voices. They cannot import voices or change settings.
Usage
npx -y speaklna setup # detect installed tools, show every change, ask before writing
npx -y speaklna setup claude codex --yes
npx -y speaklna uninstall
speaklna say "Hello" # also: stop, status, voices (add --json)
speaklna save "Hello" -o hello.wav # speech to a WAV file, nothing is playedsave reads the text the same way as ⌥T (24 kHz mono 16-bit WAV). Without -o the file stays in the folder you chose in Speaklna (Connect AI › Saving Audio) and its path is printed. The first time, Speaklna asks you to choose that folder and nothing is saved; run it again after choosing. Text can come from stdin with -. It needs Speaklna 1.1 (4) or later, runs only when the app is not already reading, and stops if you press ⌥T meanwhile; Ctrl-C cancels it. Exit code 1 when nothing was saved.
Notes:
setupshows exactly which files and keys it will change and asks y/N. Originals are saved once as<file>.speaklna-bak.Hooks call the installed
speaklnaby absolute path. For a stable path runnpm i -g speaklnafirst.Codex: after setup, open
/hooksonce inside Codex and trust the hook.OpenClaw: restart the Gateway if it is running. Its "can't verify where this plugin came from" warning is expected.
Needs macOS with the Speaklna app. On other systems the commands print a message and exit 0.
Protocol between this CLI and the app: PROTOCOL.md. Tests: npm test (Node 18+, no dependencies).
License: MIT. Author: Aivmelna LLC.
Related MCP server: low-latency-tts-api-server-mcp
中文
给 Mac 上的 AI 编程工具用的 Speaklna(本地朗读)命令行薄壳,支持 Claude Code、Codex、OpenCode、Grok Build、OpenClaw。
speaklna mcp:MCP 服务,工具speak/save/stop/status/voices。speaklna save "文字" -o 文件.wav:念法和 ⌥T 一样,存成 wav、不出声;不给-o就留在你在 Speaklna 里选的文件夹(接入 AI › 存成音频),打印路径。第一次会先弹框让你选文件夹、这次不存,选好再跑一次。speaklna hook <工具>:每轮结束的钩子,只念回复最后一行以🔊开头的报告句;密钥类内容会先抹掉;钩子不打印、不阻塞、退出码恒为 0。speaklna setup/uninstall:把 MCP、钩子和“每轮结尾写一行 🔊”的规矩写进各家配置,卸载时删干净。
AI 只能念、存成 wav、停、看状态、列音色,改不了设置,也不能导入音色。
npx -y speaklna setup # 自动发现已装的工具,先列出要改的文件,同意才写
npx -y speaklna uninstall写入前会备份为 <文件>.speaklna-bak。Codex 装完要在 Codex 里打开 /hooks 点一次信任;OpenClaw 的 Gateway 开着就重启一次(会提示无法确认插件来源,正常)。
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Remote text-to-speech MCP server with text_to_speech and list_voices tools. Free plan.
MCP server for Text-to-Speech
MCP server exposing the AceDataCloud Fish Audio API (text-to-speech with voice conditioning)
Nifty's MCP server — exposes tasks, projects, messages, and files as tools for AI agents.
Related MCP Servers
- AlicenseAqualityDmaintenanceMCP server for text-to-speech using macOS say command, enabling speech synthesis, audio file generation, and voice management.521 npm1MIT
- FlicenseNot gradedqualityCmaintenanceA low-latency text-to-speech MCP server that uses local Kokoro GGUF inference via TTS.cpp, providing say, get_voices, and get_status tools for AI agents to synthesize speech and manage playback queues.-
- FlicenseNot gradedqualityCmaintenanceA native macOS MCP server for voice input and output, allowing Claude to listen and speak with offline speech recognition and TTS.1-
- AlicenseNot gradedqualityCmaintenanceEnables AI agents to narrate internal thoughts, announce tool actions, and speak final responses aloud through local cross-platform text-to-speech. It exposes dedicated audio tools over stdio JSON-RPC for MCP-compatible clients.MIT