Describe the problem/error/question
I’m currently trying to create an AI Agent in n8n that can communicate with my Home Assistant instance via the MCP Server feature. The goal is to be able to say something like “Turn studio light on” and have the agent automatically identify and trigger the correct switch entity – without having to manually define entities in advance.
I followed the official Home Assistant documentation to set everything up. I’ve added my Home Assistant instance as an MCP Server and made sure the credentials and base URL are correct. However, I haven’t been able to get Home Assistant to work as a target within an Agent. Either the connection doesn’t establish properly or something fails during payload execution.
So, I’m wondering:
- Has anyone successfully integrated Home Assistant via MCP Server into an AI Agent in n8n?
- Are there any working examples, templates or best practices to share?
- How can I create a dynamic prompt that lets me refer to devices naturally, like “turn on living room lamp” or “activate heater,” without needing hard-coded entity names?
What is the error message (if any)?
There is no explicit error message in the UI – the workflow simply fails to execute the intended action in Home Assistant.
The last node (the AI Agent) doesn’t return any meaningful error or result. The execution stops there with no action taken in Home Assistant.
Information on your n8n setup
- n8n version: 1.89.2
- Database (default: SQLite): no
- n8n EXECUTIONS_PROCESS setting (default: own, main):
- Running n8n via (Docker, npm, n8n cloud, desktop app): Docker (Coolify)
- Operating system: Debian