As usual when doing a passion project, I discovered these apps only after I created and published my own MCP server running on Homey Pro!
I wanted to switch radio stations from Claude and ended up building a full MCP server that runs directly on Homey. The result is AI Chat Control - now available in the Homey App Store.
Key differences with the solutions here:
Runs on Homey Pro (no separate server needed)
Custom Flow trigger card - create your own AI tools that trigger Flows with parameters
Works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible AI assistant
Started as a simple radio control, grew into something much bigger!
They refer to the Homey Community Forum at the bottom of the article, and not to any specific thread, so it’s likely that many readers then will land at this article. Can’t find any other MCP related thread as this one.
Just trying to add the MCP connector for Homey-MCP (the Athom one). It seems it’s not possible with a personal (free) account even if I habe a subscription and API keys.
Has anyone more details what’s needed? ChatGPT itself says I would need a Pro or Enterprise account
Hmm, I think API key is something different. PRO (sorry, PLUS) subscription is separate and API access as well. I do believe that for MCP integration, you actually need PLUS subscription “only”, so 23€/month. But there is much more than MCP to be honest.
Super useful feature, and it’s genuinely great that Athom decided to build its own MCN connector all of that is fantastic.
The only issue is that it doesn’t yet do the one thing it really should: let us create advanced, complex flows using AI. Let me explain…..
I believe I raised this in another thread, but it didn’t get much traction.
Still, Im convinced that if there’s one place where LLMs should truly shine, its here: simplifying the process of building automations.
Most of us already know exactly what we want our smart home to do we just dont always have the time or patience to spend hours constructing flows, wiring conditions, managing variables, and testing triggers.
Yes, the MCP connector is impressive, and it enables some very smart interactions but to what end?
To ask things we can already see at a glance?
“Is the kitchen light on?” “Dim the corridor by 20%.”
Nice, sure, but not transformative. In many cases, it’s just an extra step.
What would actually be a breakthrough is this: Describe an Advanced Flow in natural language and have it built directly in Homey.
No dragging blocks around, no hunting through logic expressions.
Just: “When X happens and Y conditions apply, do Z.”
That’s real simplification, that’s where time is saved, that’s when the smart home finally feels smart.
And of course, anyone who enjoys crafting flows by hand can keep doing so nothing needs to be taken away.
But giving people the choice matters. Some of us want our evenings free for family, hobbies, rest not debugging conditional flows after a long day.
As far as I know, this isn’t possible today because Athom doesn’t expose the APIs needed to create or import Advanced Flows.
It appears to be an intentional product decision rather than a technical limitation.
I genuinely hope Athom reconsiders this……
This would align Homey with where the smart home is clearly heading and push it ahead in true usability ( not just power).
This is the kind of innovation that makes Homey not only capable, but genuinely effortless to live with.
You can have AI create advanced scripts using the built in HomeyScript.
I gave Cursor AI a list of my devices and had it create all my actions in one big script, running in a loop. So now all my buttons support any number of clicks, the time, light, temperature and precence sensors controls the lights, ovens, etc.
It’s not as readable as an advanced flow, but javascript supports classes for organizing.
It can’t create flows or read what button was pressed, so every action has to have a flow to call the big script with a parameter, but it still is far less work.
I still have to copy the updated script to Homey, I haven’t found a good way to automate this.
I’ve been thinkering with this, when tried to do this with their local api, the workflows page on the browser keeps on crashing that i had to factory reset my homey device and run a backup(thank for that)! So i found out that the webapi is able to do this via the developers page. I’m investigating how to implement this….
It would be nice to have an Alexa integration / skill to interact via voice and let the AI with MCP do the job. E.g. ask Alexa if a window is open act a device without specifying the correct device name.
I’m curious if Alexa+ will be open to use other LLMs or use MCP…
More info & discussion: [APP][Pro] AI Chat Control - Control your Homey with AI assistants using natural language
App Store: AI Chat Control | Homey
