DominionZA/HomeAssistant-ChatGpt-MCP

Home Assistant ChatGPT MCP Server

A Model Context Protocol (MCP) server that allows AI assistants to send commands to ChatGPT in Home Assistant. Compatible with Claude Desktop, Cursor, and any MCP-enabled AI client.

Features

• Send natural language commands to ChatGPT in Home Assistant
• Works with any MCP-compatible AI client
• Secure environment variable configuration
• Real-time device state feedback

Prerequisites

• Home Assistant with OpenAI Conversation integration configured
• Python 3.8+
• Long-lived access token from Home Assistant

Installation

1. Clone the repository:

   git clone https://github.com/DominionZA/HomeAssistant-ChatGpt-MCP.git
   cd HomeAssistant-ChatGpt-MCP
   

2. Install the MCP server:

   pip install -e .
   

3. Set up environment variables:

   Windows (PowerShell):

   $env:HA_TOKEN = "your_long_lived_access_token_here"
   $env:HA_URL = "http://homeassistant.local:8123"  # Optional, defaults to this
   

   Windows (Command Prompt):

   set HA_TOKEN=your_long_lived_access_token_here
   set HA_URL=http://homeassistant.local:8123
   

   Linux/Mac:

   export HA_TOKEN="your_long_lived_access_token_here"
   export HA_URL="http://homeassistant.local:8123"  # Optional
   

   To make environment variables permanent:

   • Windows: Add them to System Environment Variables
   • Linux/Mac: Add the export commands to your ~/.bashrc or ~/.zshrc

Getting Your Home Assistant Token

1. Go to your Home Assistant web interface
2. Click on your profile (bottom left)
3. Scroll down to "Long-Lived Access Tokens"
4. Click "Create Token"
5. Give it a name like "MCP Server"
6. Copy the token and use it as your HA_TOKEN

Configuration for AI Clients

Claude Desktop

Add this to your Claude Desktop config file:

Windows: %APPDATA%\Claude\claude_desktop_config.json Mac: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "homeassistant-chatgpt": {
      "command": "homeassistant-chatgpt-mcp",
      "env": {
        "HA_TOKEN": "your_long_lived_access_token_here",
        "HA_URL": "http://homeassistant.local:8123"
      }
    }
  }
}

Cursor

Add this to your Cursor config file:

Windows: %APPDATA%\Cursor\User\globalStorage\anysphere.cursor\settings\claude_desktop_config.json

{
  "mcpServers": {
    "homeassistant-chatgpt": {
      "command": "homeassistant-chatgpt-mcp",
      "env": {
        "HA_TOKEN": "your_long_lived_access_token_here",
        "HA_URL": "http://homeassistant.local:8123"
      }
    }
  }
}

After adding the configuration, restart your AI client.

Usage

Once configured, you can use natural language commands in your AI client:

• "Turn on the living room lights"
• "Set the bedroom lamp to blue"
• "Turn off all the lights"
• "Set the thermostat to 72 degrees"
• "Turn the Pad Mi Lamp on and to one of the following colors - red, orange, pink, blue, yellow or purple. Don't choose a color it is currently."

The MCP server will:

1. Send your command to ChatGPT in Home Assistant
2. Return success/failure status
3. Show which devices were affected

Environment Variables

Testing

Test the MCP server directly:

homeassistant-chatgpt-mcp

If configured correctly, it will start and wait for input. Press Ctrl+C to exit.

Troubleshooting

"HA_TOKEN environment variable is required"

• Make sure you've set the HA_TOKEN environment variable
• Restart your terminal/AI client after setting variables

"Cannot connect to Home Assistant"

• Check that Home Assistant is running
• Verify your HA_URL is correct
• Ensure your token has the right permissions

"No response from ChatGPT"

• Verify OpenAI Conversation integration is set up in Home Assistant
• Check that ChatGPT has access to control your devices

Security

• Never commit your Home Assistant token to version control
• Use environment variables to keep credentials secure
• Regularly rotate your access tokens

License

MIT License - see LICENSE file for details.