Deepgram MCP Server

This is a Deepgram MCP (Modular Communication Protocol) server that integrates with Claude Desktop to provide real-time audio generation from text input. It leverages the Deepgram API for audio processing and plays back the generated audio within your MCP-compatible client.

Demo

Features

• Converts text to audio using Deepgram
• Plays audio in real-time
• Easy to configure and run locally

🚀 Getting Started

Prerequisites

• Python 3.13 or higher installed on your machine
• A Deepgram API key (Get one here)
• Claude Desktop or any other MCP client like Cursor app installed

🛠️ Installation

Step 1: Clone the Repository

git clone https://github.com/your-username/deepgram-mcp.git
cd deepgram-mcp

Step 2: Install Dependencies

This project uses Poetry for dependency management.

Install Poetry (if not already installed):

curl -sSL https://install.python-poetry.org | python3 -

Install project dependencies:

poetry install

Alternatively, you can manually install from pyproject.toml:

pip install -r requirements.txt

🧩 MCP Server Setup in Claude

Step 3: Register MCP Server in Claude Desktop

1. Open Claude Desktop.
2. Go to Settings → Developer Settings.
3. Click Edit Config to modify the configuration JSON.
4. If the config is empty, add the following block. Otherwise, merge this configuration under the mcpServers section.

⚠️ Replace:

• //path to python with the full path to your Python interpreter.
• /absolute/path/to/server.py with the absolute path to the server.py file in this repository.
• DEEPGRAM_API_KEY with your actual Deepgram API key.

{
  "mcpServers": {
    "deepgram-mcp": {
      "command": "/usr/bin/python3",
      "args": [
        "/Users/your-username/path/to/deepgram-mcp/server.py"
      ],
      "env": {
        "DEEPGRAM_API_KEY": "your-deepgram-api-key"
      },
      "description": "Deepgram MCP server to transcribe audio. It is a STT server that uses the Deepgram API to transcribe audio."
    }
  }
}

🧪 Testing the Integration

1. Restart Claude Desktop after saving the configuration.
2. Trigger the MCP client to use the deepgram-mcp server.
3. You should hear audio generated from your text input.

📁 Project Structure

deepgram-mcp/
│
├── server.py               # Entry point for the MCP server
├── pyproject.toml          # Dependency & package management
└── README.md               # Documentation

🙋‍♂️ Support

For issues or questions, feel free to open an issue or contact the maintainer.