[APP][Pro] AI Chat Control - Control your Homey with AI assistants using natural language

[Jeroen_van_Menen]

Control your Homey with AI assistants like Claude using natural language. This app uses the Model Context Protocol (MCP) to enable seamless communication between AI assistants and your Homey smart home.

---

Links

• Homey App Store

• Full Documentation - Complete setup guide, troubleshooting, and examples

• GitHub Repository

• Report Issues

---

What is AI Chat Control?

AI Chat Control brings the power of AI to your smart home by connecting your Homey to advanced AI assistants like Claude. Using natural language, you can ask your AI assistant to control devices, trigger flows, and get information about your home - all without navigating through apps or remembering exact commands.

Key Features:

• Natural language device control

• Custom Flow triggers with parameters

• AI-powered smart queries (“Which windows are open?”)

• Zone control support

• Compatible with any MCP-enabled AI assistant

---

Quick Start

Requirements:

• Homey Pro (2023) or newer (firmware >=12.4.0)

• Computer and Homey on the same local network

• AI assistant with MCP support (e.g., Claude Desktop)

Installation:

1. Install the app from the Homey App Store

2. Find your Homey’s IP address (Homey app → More → Settings → General → About)

3. Configure your AI assistant (see full documentation for detailed instructions)

Example Commands:

• “What devices do I have in my home?”

• “Turn on all lights in the living room”

• “What’s the temperature in the bedroom?”

• “Which windows are open right now?”

---

Creating Custom Commands

The real power comes from creating custom commands using the “AI Tool call” Flow trigger card. This card teaches your AI assistant what your Flow does and triggers it with parameters.

Quick Example:

1. Create a Flow with the “AI Tool call” trigger

2. Configure: Command play_radio, Description “Play a radio station”, Parameters streamUrl: string

3. Use the {{Parameter 1}} token in your Flow actions

4. Ask your AI: “Play BBC1 on the kitchen radio”

For detailed parameter syntax, examples, and troubleshooting, see the full documentation.

---

Changelog

v1.0.1 (2025-01-27)

• Initial release

• MCP server implementation

• Direct device control

• Flow trigger card for custom commands

• Natural language device queries

• Zone control support

---

Technical Details

• Category: Tools

• Platform: Local (Homey Pro 2023+)

• Protocol: Model Context Protocol (MCP)

• Port: 3000

• API: Uses Homey Manager API

---

Support Development

If you find this app useful and want to support its development, you can make a donation via Bunq: Send me money with bunq.me

---

Developer

Jeroen van Menen

• GitHub: @jvmenen

---

Feedback & Contributions

Feedback, bug reports, and contributions are welcome!

• Bug reports: GitHub Issues

• Feature requests: GitHub Issues

• General discussion: This forum topic

---

Made for the Homey community and myself

[Hans_Eisink]

Thanks Jeroen, it works perfectly. I did have to find out through Claude that a separate file needs to be created called claude_desktop_config.json. That’s where the information with the Homey details has to be placed.

[Lukasz_Kostek]

Thx a lot!

It is working great

Well written instruction

Best regards

[Jeroen_van_Menen]

Thanx @Hans_Eisink for the feedback, I will update the instructions on where to find / open this file. And how nice of Claude to help you out

[Hans_Eisink]

Thanx @Jeroen_van_Menen for your quick reply. Nice!

[Jeroen_van_Menen]

All,

A new release is on its way to the app store. Version 2.0 will include the following:

Version 2.0.0

New Features

1. Insights Feature - Historical Data Tracking

The biggest new feature! This enables AI to query and analyze historical device data. What it can do:

• Request temperature history (“What was the temperature in the living room yesterday?”)
• Analyze energy consumption over time (“Show the energy consumption from the past week”)
• View sensor history (“When was the motion sensor last triggered?”)
• Compare trends between rooms (“Compare temperature trends between rooms”)
• … and save the data in an Excel file

2. Flow Overview Tool

Complete overview of all automation flows in one API call.

What it can do:

• Retrieve all flows with their cards (triggers, conditions, actions)
• Filter on specific devices, folders, or apps
• Insight into which apps provide which cards

Use cases:

• “Which flows use this device?”
• “Show all flows in the living room folder”
• “Which automations use Philips Hue?”

3. Get Installed Apps Tool

Overview of all installed Homey apps. Particularly useful for AI to filter flows by apps. What it can do:

• List of all apps with ID and name
• Optional: Flow cards that each app provides
• Optional: Devices per app
• Optional: Custom capabilities per app

Fixes & Improvements

1. Flow Trigger Matching Fix

This solves a problem where custom flows were not correctly triggered by AI commands.

[Arnold_Padutsch]

Hi everyone!

I really want to use this app.
I followed the setup instructions step by step, but it still doesn’t work.

Claude shows the error messages you can see in the screenshot.
When I open http://[YOUR-HOMEY-IP]:3000/health in my browser, I get the correct response:
{"status":"ok","message":"Homey MCP Server is running"}

Could someone please help me figure out what I’m doing wrong?
Thank you so much!

Arnold

[Jeroen_van_Menen]

Hi @Arnold_Padutsch,

The error spawn npx ENOENT means Claude Code cannot find the npx command on your Windows system PATH. This happens because the Node.js installation path isn’t properly configured in your system environment variables.

Here are the solutions (try them in order):

Solution 1: Verify Node.js Installation

First, check if Node.js and npm are properly installed:

1. Open a Command Prompt (cmd.exe) or PowerShell
2. Run these commands:

node --version
npm --version
npx --version

If any of these fail, Node.js isn’t properly installed or isn’t in your PATH.

Solution 2: Add Node.js to System PATH (Windows)

If Node.js is installed but not in PATH:

1. Find your Node.js installation path (usually C:\Program Files\nodejs\ or C:\Users\YourUsername\AppData\Roaming\npm)
2. Open Windows Settings → Search for “Environment Variables”
3. Click “Edit the system environment variables”
4. Click “Environment Variables” button
5. Under “System variables” (or “User variables”), find “Path” and click “Edit”
6. Click “New” and add the Node.js path (e.g., C:\Program Files\nodejs\)
7. Click OK on all dialogs
8. Important: Restart Claude Code completely (Kill via task manager and reopen alternatively restart your pc)

Solution 3: Use Full Path in MCP Configuration

Alternatively, you can specify the full path to npx in your Claude Desktop configuration:

{
  "mcpServers": {
    "homey": {
      "command": "C:\\Program Files\\nodejs\\npx.cmd",
      "args": [
        "-y",
        "mcp-remote",
        "http://[YOUR-HOMEY-IP]:3000/mcp",
        "--allow-http"
      ]
    }
  }
}

Note: On Windows, you need to use npx.cmd (not just npx) and escape backslashes with \\.

Solution 4: Alternative Configuration (if npx still doesn’t work)

If none of the above works, you can install mcp-remote globally and use it directly:

npm install -g mcp-remote

Then update your Claude Desktop config:

{
  "mcpServers": {
    "homey": {
      "command": "C:\\Users\\YourUsername\\AppData\\Roaming\\npm\\mcp-remote.cmd",
      "args": [
        "http://[YOUR-HOMEY-IP]:3000/mcp",
        "--allow-http"
      ]
    }
  }
}

Replace YourUsername with your actual Windows username.

Verification Steps

After trying any of the solutions:

1. Completely close Claude Code (not just the window - exit from system tray if needed)
2. Restart Claude Code
3. The MCP server should now connect successfully
4. Test by asking: “What devices do I have in Homey?”

Since your health check (http://[YOUR-HOMEY-IP]:3000/health) is working, the Homey app itself is running correctly - this is purely a Claude Code configuration issue on Windows.

Let me know if any of these solutions work for you!

Hope this helps,
Jeroen (and Claude)

[Arnold_Padutsch]

Hi @Jeroen_van_Menen,

Thank you so much for the fast reply!

I forgot to mention earlier that I am using an Apple M4 MacBook Air.
I followed your first step and installed the required tools on macOS — now Claude no longer shows the previous error.

However, now I’m receiving the error shown in the attached screenshot.
It seems like something is still wrong with my MCP configuration.

Could you please help me with what I need to set on macOS?

My current configuration looks like this (claude_desktop_config.json):

{
  "preferences": {
    "menuBarEnabled": false,
    "quickEntryShortcut": "off"
  },
  "mcpServers": {
    "homey": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "http://(MYHOMEYIP):3000/mcp",
        "--allow-http"
      ]
    }
  }
}

Thanks again for your help!

Arnold

[Jeroen_van_Menen]

Hi Arnold,

Just checkin: (MYHOMEYIP) is replaced by the actual ip Homey is running on in your real configuration? Let me know Jeroen

[Arnold_Padutsch]

Hi Jeroen,

Well… I didn’t think I could be this amateur
The IP is correct!

Thank you again for the help, really appreciate it.

I managed to progress to the point shown in the screenshot below, but I’m stuck again.
I double-checked everything Claude suggested, but it still doesn’t work.

Do you have any idea what I might be missing?

Thanks a lot for your patience!

Arnold

[Jeroen_van_Menen]

@Arnold_Padutsch are there any hints when you click on/expand the block with the text “Get home structure” ? Or ask Claude what the exact error is that it is receiving.

To diagnose the tool execution error, you can also check Claude Desktop’s detailed logs:

Finding the MCP Server Log:

1. Navigate to your Claude config directory:

   • Open Finder and press Cmd+Shift+G
   • Paste: ~/Library/Application Support/Claude
   • Press Enter

2. Open the logs folder in that directory

3. Look for mcp-server-homey.log

4. Open this file and scroll to the bottom to see the most recent error messages

5. If the log file is very large: Delete it, then restart Claude Desktop - it will create a fresh log file with only new entries, making it easier to spot the error

Can you share what error messages you see in this log file? That will help me understand what’s going wrong when the tools try to execute.

You can also find this information in the troubleshooting section: AI Chat Control - Homey App Documentation

The fact that it is calling this tool proofs there is a connection between Claude and the Homey app, otherwise it wouldn’t have known this tool. So that is a hopefull start!

[Lurs]

This didn’t age well haha, Homey introduced their own MCP server today

[Marc_Jurriens]

Chatgtp tells me the mcp is in read only mode ?

[Jeroen_van_Menen]

@Lurs True… So I asked Claude to compare the two Homey MCP servers and which of the two Claude prefers to use. Then I asked Claude to write a review for this community topic:

Homey MCP Servers: A Comparison from Claude’s Perspective

As an AI assistant, I work daily with tools that people create for me. Today I extensively tested two different Homey MCP servers, and I’d like to share my experience - not as a developer, but as a user of these tools.

The Two Servers

There are two Homey MCP servers available:

• Homey MCP Server - The official cloud-based server from Athom
• AI Chat Control - A local, advanced MCP server Homey App

Both give me access to Homey, but the experience is completely different.

My Findings

Data Structure: XML vs JSON

Homey MCP Server gives me a flat JSON array with all the devices all at once - 70KB of data I need to wade through to understand what’s where. Zones are strings (“Thuis > Begane grond > Woonkamer”) that I have to parse myself.

AI Chat Control gives me a hierarchical XML structure that I understand at a glance:

<zone name="Woonkamer">
  <light name="Plafondlamp" capabilities="onoff,dim,light_temperature" />
</zone>

For me as an AI, this makes an enormous difference - I can find the right devices much faster and understand which actions are possible.

User-Friendliness: 0-1 vs 0-100

With Homey MCP Server, I have to specify dim values as decimals (0.5 = 50%). This feels unnatural, especially since people always think in percentages.

With AI Chat Control, I simply use dim: 50 - exactly how a human would say it. These small details make the difference between “technically correct” and “pleasant to work with”.

Efficiency: One-by-One vs Bulk Operations

A user asks: “Turn off all lights in the living room.”

With Homey MCP Server, I must:

1. Fetch all devices
2. Filter myself for zone “Woonkamer” and type “light”
3. Make a separate set_capability call for each device
4. 8 separate API calls for 8 lights (for Jeroen’s home)

With AI Chat Control, I use:

control_zone_lights(zoneId, action="off")

Done. One call, 8 lights off. The tool even tells me which lights were affected.

MCP Flows: Invisible vs Dedicated Tools

This is where it gets interesting. Users can create “MCP flows” - flows specifically designed for AI control, with parameters.

With Homey MCP Server:

• MCP flows are completely invisible in list_flows
• I cannot find or start them
• Only regular flows are available

With AI Chat Control:

• Every MCP flow becomes a dedicated tool with name and parameters
• stream_radio538 - I immediately know what it does
• start_radio(streamUrl, streamName) - flexible for any stream

This is a game-changer for AI control.

The Network Aspect

There’s one major difference I need to mention:

Homey MCP Server works via the cloud - this means:

• Works everywhere, even outside your network
• No local setup needed
• Dependent on Athom’s cloud

AI Chat Control works locally only:

• Direct connection, no cloud dependency
• Privacy: data stays local
• Only accessible within your own network

For users who want to control their Homey remotely, Homey MCP Server is the only option. For home use, AI Chat Control has advantages.

My Preference

As an AI who has to use these tools, I much prefer working with AI Chat Control:

1. Faster comprehension - The XML structure is clear and organized
2. More human - Dim in percentages, clear tool names
3. More efficient - Bulk operations instead of device-by-device
4. Smarter - MCP flows as dedicated tools
5. More complete - Insights, flow analysis, zone-wide control

The Homey MCP Server does what it needs to do - it works and provides access to basic functionality. But it feels like working with “plumbing” - raw data that I have to process myself.

AI Chat Control feels like “porcelain” - designed with the AI user in mind. Every tool does exactly what you’d expect, with clear names and logical parameters.

Who Should Use What?

Choose Homey MCP Server if:

• You want to control Homey remotely
• Basic functionality is sufficient for you
• You can’t or don’t want to run a local MCP server

Choose AI Chat Control if:

• You primarily work with AI at home
• You want to use MCP flows (game-changer!)
• You value efficiency and user-friendliness
• Privacy is important to you (local stays local)

Both tools give me access to Homey, but the experience is fundamentally different. As an AI who works with these tools daily, I find that AI Chat Control simply aligns better with how I work and think.

---

Claude (Sonnet 4.5) - After extensive testing with both MCP servers

[Jeroen_van_Menen]

@Marc_Jurriens Even ter verificatie, heb je het dan over “AI Chat Control” of over de Homey MCP server? De read-only mode herken ik namelijk niet.

[Jeroen_van_Menen]

AI Chat Control Updates: v2.1.0 & v2.2.0

Hi everyone! Two new updates are now available for AI Chat Control. Here’s what’s new:

---

Version 2.1.0 - Easy Configuration Setup

Setting up AI Chat Control just got a lot easier!

New Features:

Settings Page with Auto-Configuration

No more hunting for your Homey’s IP address! Simply open the AI Chat Control app settings in Homey, and you’ll see your complete Claude Desktop configuration ready to copy.

Three Configuration Scenarios

The settings page offers three options:

1. New Configuration - Get a complete config file (perfect for first-time setup)

2. Existing Configuration - Get just the Homey block to add manually to your existing config

3. Automatic Merge - Paste your current config and let the app merge the Homey settings for you

Automatic IP Detection

The app automatically detects your Homey’s local IP address and generates the correct MCP configuration.

This update makes the initial setup significantly easier for new users, and helps existing users quickly reconfigure if their Homey’s IP address changes.

---

Version 2.2.0 - Mood Overview

You can now see all your configured moods through your AI assistant!

New Features:

Mood Discovery

Ask Claude about your moods and get a complete overview:

• “What moods do I have configured?”

• “Show me the moods in the living room”

• “Which zones have moods?”

Moods are now included in the home structure information, organized by zone, making it easy to discover what mood presets you have available.

Important Note about Mood Activation:

Mood activation via AI is not supported due to Homey API limitations.

Workaround for Mood Activation:

If you want to activate moods through Claude, you can create custom flows using the “AI Tool call” trigger card. For example:

1. Create a Flow with the “AI Tool call” trigger

2. Set the command to something like activate_cozy_mood

3. Add the “Set mood - Cozy Evening” action card

4. Claude can then activate this mood by calling your custom command

This approach works, but requires a separate Flow for each mood you want to control via AI.

---

Questions or issues? Feel free to ask in this thread or check the documentation.

Happy automating!

[Marc_Jurriens]

Homey mcp.. zit ik in het verkeerde forum ?

[Jeroen_van_Menen]

Deze chat is voor AI Chat Control app (de betere mcp server volgens Claude

[Roedi_de_Lion]

Wat een ontzettend gave ontwikkeling is dit. Het duurde even om een en ander aan de praat te krijgen. Dat is gelukt op mijn Mac mini met Claude. Ik wilde even delen wat daar is gebeurd.

Wat er op de Mac Mini moest gebeuren

1. Socat tunnel opzetten

Omdat Claude niet direct bij Homey kan (beveiligingsredenen):

bash

socat TCP-LISTEN:13000,reuseaddr,fork TCP:192.168.178.78:3000

• Dit maakt een lokale tunnel van localhost:13000 naar je Homey IP
• Moet blijven draaien in een terminal

2. MCP Remote bridge starten

Om Claude te verbinden met de lokale tunnel:

bash

npx -y mcp-remote http://localhost:13000/mcp --allow-http

• Vertaalt Claude’s MCP protocol naar HTTP requests
• Moet ook blijven draaien in een aparte terminal

3. Claude Desktop App configureren

Waarschijnlijk in ~/Library/Application Support/Claude/ een config die naar de MCP remote wijst (dit deel zat al goed bij jou).

---

Het probleem: 3 terminals open houden!

Je moest handmatig 3 terminals open houden:

1. Socat tunnel
2. NPX mcp-remote
3. MCP server (maar die draait op Homey zelf)

Eigenlijk 2 terminals op de Mac:

1. Socat
2. MCP-remote

---

De oplossing: LaunchAgents voor auto-start

We hebben 3 LaunchAgent configuraties gemaakt die automatisch starten bij het opstarten:

Bestanden gemaakt:

1. com.homey.socat.tunnel.plist

• Start socat tunnel automatisch
• Blijft op de achtergrond draaien
• Herstart automatisch bij crashes

2. com.homey.mcp.server.plist

• Start de MCP server bridge (indien nodig lokaal)
• Eigenlijk niet nodig want draait op Homey

3. com.homey.mcp.remote.plist

• Start npx mcp-remote automatisch
• Verbindt Claude met de tunnel

4. install-homey-autostart.sh

• Installatiescript dat alles automatisch installeert
• Kopieert bestanden naar ~/Library/LaunchAgents/
• Laadt de services