A hands-on walkthrough of connecting a local LLM (via LM Studio and Ollama) to Home Assistant using the HA-MCP Server package. The setup involves installing the uv package manager on Windows, configuring an mcp.json file with the Home Assistant URL and a long-lived token, and then using LM Studio models like Qwen 3.5 (9B) to control smart home devices through natural language prompts. The author notes intentionally disabling high-risk tools such as backup restoration and script triggering to avoid accidental damage to the Home Assistant instance.

3m read timeFrom xda-developers.com
Post cover image
Table of contents
I've used the unofficial HA-MCP Server repoIt didn’t take long to configure the HA-MCP ServerMy LM Studio models now control my smart home

Sort: