Control your Homey with any AI assistant, like Claude, ChatGPT, Gemini, Copilot or Cursor. Just ask in plain language: turn off the lights, check the temperature, create a new flow, or manage your apps and devices. No programming required.
You can also build your own AI-powered automations in the Homey flow editor. Create flows that respond when your AI calls a custom command, and send a real answer back to the AI using device values and flow tokens.
Read more ›
Flow cards
Click on a device above to show related Flow cards only.
When...
MCP AI Bridge
The generic alarm turned on
MCP AI Bridge
The generic alarm turned off
MCP AI Bridge
An AI agent connects
MCP AI Bridge
The AI controls a device
MCP AI Bridge
A flow becomes broken
MCP AI Bridge
AI agent calls a custom tool
MCP AI Bridge
A security event occurs
MCP AI Bridge
MCP AI Bridge server starts
MCP AI Bridge
An AI tool call fails
And...
MCP AI Bridge
The generic alarm is on
MCP AI Bridge
There is a broken flow
MCP AI Bridge
Remembered Name is Value
MCP AI Bridge
The tool input contains Text to look for
MCP AI Bridge
The tool input is empty
MCP AI Bridge
The tool name is Tool name
Then...
MCP AI Bridge
Ask the model: Question Include
Advanced
MCP AI Bridge
Remember Name as Value for the AI
MCP AI Bridge
Return error Error message to the AI agent
MCP AI Bridge
Return Response to the AI agent
Support
Having an issue with this app?
Contact the developer via the developer website.
What’s new
Version 4.15.0 — A flow can now ask a model a question and use the answer. Everything else in this app points one way: an assistant reaches into your home, reads it and operates it. Nothing in the home could ask the assistant anything, because there is no assistant sitting there, only a client somebody has open. The new flow card Ask a model a question sends your question to a provider and hands the answer back as a tag you can use later in the same flow: put it in a notification, speak it, compare it, whatever the flow needs. It uses YOUR OWN API KEY with Anthropic, OpenAI or Mistral, which you set up under Commands, so every question is a request you pay for. That is why it is off until you switch it on and why there is a daily limit, set to 50 to begin with: a flow that reacts to a device can fire hundreds of times an hour and the first you would know is the bill, so past the limit the question is refused until tomorrow. The settings page shows how many of today's questions are used. The boundary is deliberate and worth knowing: text in, text out. The model gets a question and returns an answer. It cannot call tools, it cannot see your home unless you set the card to include a short summary, and it cannot change anything. A flow that should act on the answer does that with its own cards, where you can see what will happen. It has a tab of its own, called Ask a model, with the steps for building the flow, the provider and key settings, and a test question you can ask straight from the page: that way you find out whether your key works without building a flow first. The test question costs exactly what any other question costs, so it counts towards the same daily limit, and the page shows how many of today's are used.
View changelog