Provides audio transcription capabilities through OpenAI's Whisper model, supporting various models (base, small, medium, large, large-v2, large-v3) and output formats.
Enables execution of shell commands with basic security validation, allowing to run commands in specified working directories while blocking potentially dangerous operations.
Whisper CLI MCP Server
An MCP server that provides shell command execution and OpenAI Whisper transcription capabilities.
Features
whisper_transcribe: Transcribe audio files using OpenAI Whisper
shell_command: Execute shell commands safely with basic security validation
Related MCP server: Audio Transcriber MCP Server
Installation
Install dependencies:
Make the server executable:
Usage
Running the Server
Tools Available
whisper_transcribe
Transcribe audio files using whisper-cli.
Parameters:
audio_file(required): Path to the audio filemodel(optional): Whisper model (base, small, medium, large, large-v2, large-v3)language(optional): Language code for transcriptionoutput_format(optional): Output format (txt, vtt, srt, json)
shell_command
Execute shell commands with basic security validation.
Parameters:
command(required): Shell command to executeworking_directory(optional): Working directory for the command
Security
The shell_command tool includes basic security validation to prevent execution of potentially dangerous commands. Commands containing the following patterns are blocked:
rm -rfsudochmod 777dd if=> /dev/
Configuration
To use this server with Claude Desktop, add the following to your claude_desktop_config.json: