Transcription MCP Sever

An MCP (Model-Context-Protocol) server for transcribing MP3 audio files using the AssemblyAI API. This server provides command-line and Docker-based tools to transcribe audio. this application can be use by Podcasters, Content Creator, Educators and Business Teams making thier task more easier.

Features

• Accepts .mp3 files via command-line interface using Fire
• Accepts .mp3 file path provided via an AI assistant such as Claude or Cursor.
• Transcribe MP3 audio files into JSON transcripts using AssemblyAI API
• Speaker diarization (speaker labels) enabled
• Save transcripts in a local "transcripts/" directory
• Easy-to-use command-line interface
• ready to run in a Docker container for portability and deployment

System Architecture Overview

This application is a microservice designed to transcribe audio files using AssemblyAI.

Architecture Breakdown

• Uvicorn as MCP Server Runtime: Uvicorn runs the server process, acting as the entry point for executing transcription jobs.
• AssemblyAI Client: This component handles all communication with the AssemblyAI API.
• CLI via Google Fire: The application includes a robust command-line interface built with Google Fire. Users can transcribe audio files directly from the terminal
• Dockerized Environment: While the server can be run directly on a local machine, it also includes a lightweight Docker configuration for users who prefer containerized deployment.

Getting started

Prerequisites

• Python 3.10+ (its Stable, popular, well-supported)
• Uvicorn (Lightweight ASGI server )
• Google Fire ( CLI Framework, Makes CLI creation simple and powerful)
• AssemblyAI API key (API Interface, sign up at https://www.assemblyai.com/)
• Docker (optional, for containerized usage. Ensures consistent, portable deployment)

Installation

1. Clone the repository:

git clone <your-repo-url>
cd <repo-folder> 

2. Install dependencies:

pip install -r requirements.txt

3. Set your AssemblyAI API key as an environment variable:

export API_KEY="your_assemblyai_api_key"   # Linux / macOS
set API_KEY="your_assemblyai_api_key"      # Windows CMD

Usage

Inspec With MCP Dev

Run This Command To Inspect And Test The Fuctionality Of Your Tool on a Web Ui.

uv run mcp dev mcp\server_transcription.py

connect and test server_transcription

Run the command to connect with claude

uv run mcp install mcp\server_transcription.py

Json Format

when connecting with an AI assitant like claude or cursor, the config should be in this jason format.

{
  "mcpServers": {
    "Audio Transcription Service": {
      "command": "C:\\Users\\HomePC\\Desktop\\mcp_task2\\.venv\\Scripts\\uv.EXE",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "mcp",
        "run",
        "C:\\Users\\HomePC\\Desktop\\mcp_task2\\mcp\\server_transcription.py"
      ],
      "env": {
        "API_KEY": "<your-api-key-here>"
      }
    }
  }
}

test client_transcription

Run the command

python  mcp\client_transcription.py "path/to/audio/file"

Docker Usage

Build Docker Image

docker build -t transcription-service -f run_with_docker/Dockerfile .

Run Docker Container

set API_KEY="your-api-key-here"

docker run -d -p 8050:8050 -e API_KEY=%API_KEY% -v C:\Users:/mnt/users -v %CD%\transcripts:/app/transcripts -v %CD%\logs:/app/logs --dns 8.8.8.8 --name transcription-server transcription-service

run client

python run_with_docker\client.py "path/to/audio/file"