Transcript Server

An MCP App Server for live speech transcription using the Web Speech API.

MCP Client Configuration

Add to your MCP client configuration (stdio transport):

{
  "mcpServers": {
    "transcript": {
      "command": "npx",
      "args": [
        "-y",
        "--silent",
        "--registry=https://registry.npmjs.org/",
        "@modelcontextprotocol/server-transcript",
        "--stdio"
      ]
    }
  }
}

Local Development

To test local modifications, use this configuration (replace ~/code/ext-apps with your clone path):

{
  "mcpServers": {
    "transcript": {
      "command": "bash",
      "args": [
        "-c",
        "cd ~/code/ext-apps/examples/transcript-server && npm run build >&2 && node dist/index.js --stdio"
      ]
    }
  }
}

Features

• Live Transcription: Real-time speech-to-text using browser's Web Speech API
• Transitional Model Context: Streams interim transcriptions to the model via ui/update-model-context, allowing the model to see what the user is saying as they speak
• Audio Level Indicator: Visual feedback showing microphone input levels
• Send to Host: Button to send completed transcriptions as a ui/message to the MCP host
• Start/Stop Control: Toggle listening on and off
• Clear Transcript: Reset the transcript area

Setup

Prerequisites

• Node.js 18+
• Chrome, Edge, or Safari (Web Speech API support)

Installation

npm install

Running

# Development mode (with hot reload)
npm run dev

# Production build and serve
npm run start

Usage

The server exposes a single tool:

transcribe

Opens a live speech transcription interface.

Parameters: None

Example:

{
  "name": "transcribe",
  "arguments": {}
}

How It Works

1. Click Start to begin listening
2. Speak into your microphone
3. Watch your speech appear as text in real-time (interim text is streamed to model context via ui/update-model-context)
4. Click Send to send the transcript as a ui/message to the host (clears the model context)
5. Click Clear to reset the transcript

Architecture

transcript-server/
├── server.ts          # MCP server with transcribe tool
├── server-utils.ts    # HTTP transport utilities
├── mcp-app.html       # Transcript UI entry point
├── src/
│   ├── mcp-app.ts     # App logic, Web Speech API integration
│   ├── mcp-app.css    # Transcript UI styles
│   └── global.css     # Base styles
└── dist/              # Built output (single HTML file)

Notes

• Microphone Permission: Requires allow="microphone" on the sandbox iframe (configured via permissions: { microphone: {} } in the resource _meta.ui)
• Browser Support: Web Speech API is well-supported in Chrome/Edge, with Safari support. Firefox has limited support.
• Continuous Mode: Recognition automatically restarts when it ends, for seamless transcription

Future Enhancements

• Language selection dropdown
• Whisper-based offline transcription (see TRANSCRIPTION.md)
• Export transcript to file
• Timestamps toggle