voska/hass-mcp

What it solves

It allows AI assistants (like Claude) to directly interact with a Home Assistant instance. This enables users to control their smart home, query device states, troubleshoot automations, and manage dashboards using natural language instead of manual configuration.

How it works

It implements the Model Context Protocol (MCP) server, acting as a bridge between an LLM and the Home Assistant API. It provides a set of tools that the LLM can call to perform actions (like turning off lights), retrieve data (like sensor history or error logs), and edit Lovelace dashboards via WebSockets, and it supports both stdio and streamable HTTP transports.

Who it’s for

Smart home enthusiasts and Home Assistant users who want to integrate their home automation with advanced AI agents for easier control and guided troubleshooting.

Highlights

• Full Device Control: Query states, control entities, and call any Home Assistant service.
• Dashboard Editing: Read and live-edit Lovelace dashboards with automatic backups and dry-run previews.
• Advanced Diagnostics: Access to Home Assistant error logs and entity state history/statistics.
• Guided Conversations: Built-in prompts for creating automations, debugging issues, and optimizing routines.
• Token Efficient: Uses lean JSON responses to minimize LLM token consumption.
• Flexible Deployment: Supports Docker, uv/uvx, and HTTP transport for network-based tools.