# mcp-voice-summary [Health: Active]

**Category:** 🎧 Text-to-Speech  
**Repository:** https://github.com/eduplopez/mcp-voice-summary  
**GitHub Stars:** 0  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/mcp-voice-summary

## Description
Reads a summary out loud when an AI coding task finishes. Offline or neural voices.

## Tools
Capabilities this server exposes over MCP:

- **language** — an ISO 639-1 code (`es`, `en`, `fr`), a regional variant
- **gender** — `"f"` or `"m"`, or `"any"` to follow the language default. Empty

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "mcp-voice-summary": {
    "command": "uvx",
    "args": ["mcp-voice-summary"]
  }
}
```

## Documentation

## What mcp-voice-summary does

The mcp-voice-summary MCP server lets an MCP-compatible coding assistant read task summaries aloud after it modifies code, creates files, or runs commands. It is intended as an accessibility layer: instead of reading an entire assistant response, you can hear a concise account of what happened.

The assistant decides how much text to send based on the task. Small changes can produce a short sentence, while multi-step work can receive a fuller summary. Playback is queued and handled asynchronously, so the speech operation returns without making the assistant wait for the audio to finish.

## How it works

The server communicates with its MCP client over standard input and output. It exposes `speak_summary` for spoken notifications and `set_language` for changing the language or gender during a session. Language selection can use an ISO 639-1 code, a regional variant, or the system language. Gender can be female, male, or the language default.

Two synthesis engines are available. `edge` generates neural speech through Microsoft's services and requires an internet connection for every utterance. `sapi5` uses installed system voices and works offline. On Linux, SAPI5 relies on espeak-ng and audio playback needs a compatible player such as `ffplay` from FFmpeg or `mpg123`.

## Setup and configuration

Install the package with `pip install mcp-voice-summary`, then configure the installed `mcp-voice-summary` command in an MCP client. `uvx` can launch it without a separate package installation. Claude Desktop, Cursor, and Cline can invoke it through a standard MCP configuration.

Python 3.10 or newer is required, and the server supports Windows, macOS, and Linux. SAPI5 requires the platform-specific extra or dependencies described by the project. Edge needs no additional Python dependency, but it does need network access. No environment block is required: the server uses the operating system language and its defaults when configuration is omitted.

Optional settings control the engine, voice name, language, gender, rate, volume, playback program, queue limits, word limits, timeout, mute mode, redaction, rate limiting, summary age, and full or short notification mode. Restart the MCP client after changing configuration. `VOICE_NAME` takes precedence over other voice selection settings.

## Tools and capabilities

The mcp-voice-summary MCP server provides these documented capabilities:

- Speak a task summary through the selected voice.
- Change language and gender at runtime with `set_language`.
- Select offline or neural synthesis engines.
- Queue notifications without blocking the assistant.
- Redact credentials from spoken summaries by default.
- Limit summary size and notification queue growth.

## Limitations and notes

Automatic language detection is unreliable for short summaries, even when it reports high confidence. A fixed language is more dependable for brief messages. Gender selection applies only to the Edge engine; SAPI5 does not expose voice gender and cannot honor that choice.

The mcp-voice-summary MCP server is audio-output software rather than a transcription or conversation system. It speaks summaries supplied by the assistant and does not describe a separate code-analysis tool. Linux users may need to install an audio player, and Edge users must accept that utterance audio is generated by Microsoft's servers. The project is licensed under MIT.

_Full upstream README: https://allmcps.com/mcp/mcp-voice-summary/readme_