✨ Homey MCP Server | Megathread

[Emile]

That’s a good idea. I’m planning to add a few new tools soon.

• rename_device
• move_device_to_zone
• get_standard_flow
• create_standard_flow
• update_standard_flow
• get_advanced_flow
• create_advanced_flow
• update_advanced_flow
• get_flow_trigger_cards
• get_flow_condition_cards
• get_flow_action_cards

Let’s hope the LLMs understand these

[pendojoshua]

Can’t wait! If you’re ready and need some testing just let me know. Happy to help!

[HenriDO]

Great news !

[Emile]

I’ve just deployed a new version of the Homey MCP Server with those new tools. I managed to create a few Flows & Advanced Flows!

But, based on your AI agent or Homey, your results might be very different

It’s the cutting edge after all!

Note: you need to re-authorize to grant your AI agent the new permissions to manage Flows.

[pendojoshua]

Thanks! I’ll test tomorrow. Just updated the permissions on https://claude.ai/settings/connectors. But for now it tells me the scope is still missing. Might take some time to puck up the changes in the key.

[Emile]

Hm, maybe remove & re-add the connection to enforce a new sign-in.

[pendojoshua]

Yes, that works!

1. Go to https://claude.ai/settings/connectors
2. Click configure
3. For both read-only tools and other tools either review manually or select Always allow
4. Go back (to all connectors)
5. Click disconnect and wait a bit
6. Click (re)connect
7. Test in Claude

I was able to rename al my lights using “I want to rename some lights. Can you turn on the kitchen lights one by one and ask me how I want to name that lamp?”.

So much easier than via the app, thanks!

[chrb]

Hi, how does it work with Open.ai and what are Common use cases? Regards

[pendojoshua]

Integration is explained in the post mentioned in the first topic:

Introducing the Homey MCP Server – News | Homey

Use cases are simple: use a chatwindow from ChatGTP (in your case) and use natural language to control Homey.

• “Turn on all lights in the living room”
• “Set the heater to 22 degrees in the master bedroom”
• “Rename all devices in the kitchen, turn them on one by one and ask me what to name them”
• Or create flow based on your description

The sky is the limit, just another way to control your Homey setup!

[Jamie]

So, I was not able to get the MCP running with ChatGPT,

No matter what I did it never really worked

1. It had read acccess, but I was ever able to get write access to my home to update devices

2. It hit threshold and then would update

3. it got out of sync (maybe from above)

4. There was a massive difference in what “worked” between the webbrowser (the most) vs the iphone app (never worked at all).

I did some digging the different devices having different access levels is a known thing, and it chatGPT was gating my requests - even after setting allow access which stoppped the MCP from even getting them to pass on.

---

I gave up and tried it on Claude today, the difference was night and day, claud allows you to easily define what it has access to do vs what it needs to ask for on the connection screen - and everything has been working really well since then.

TLDNR: If your making a choice between providers, I found Claude to work well and ChatGPT to be really problematic.

[Jamie]

I’ve been play with this for a little while now, and while it has been used for for cleaning up flows and devices. Its a bit limited in what it can do mainly due to the fact that it does not have access to when a capability was last changed. (insights access would be the same).

If it doesn’t understand when a zone, motion, presence, contact, temp last changed then we cant fully leverage the LLM for what it does best ……

---

BUT what I was able to do was hook it into the api from the audit app, then teach claude how to pull hot and cold data

curl -u api:key \
  -H "Content-Type: text/plain" \
  -X POST "https://eu-fsn-3-connect.betterstackdata.com" \
  -d "SELECT
    dt,
    JSON_VALUE(raw, '$.detail.code') AS code,
    JSON_VALUE(raw, '$.detail.value') AS value,
    JSON_VALUE(raw, '$.message') AS message
  FROM remote(t236097_xavier_logs)
  WHERE JSON_VALUE(raw, '$.detail.device') = 'Lounge Back Multisensor'
    AND JSON_VALUE(raw, '$.detail.code') = 'alarm_motion'
  ORDER BY dt DESC
  LIMIT 1
  FORMAT JSON"

And with that we can start doing bayesian probability and ask questions like

”Hey Bro-emy - report”

Home Status Report

Time: 10:16 PM, December 23rd, 2025

Quick Status

• Climate: Lounge cooling (24.7°C → 16°C target), Nursery cooling (25.1°C → 24°C target), Bedroom/Study off and warm (25.2°C / 28°C)

• Humidity: Nursery 69.9% (high), Bedroom 56%, Study 45.5%, Lounge 45.9%

• Lighting: Lounge, Kitchen, Bathroom, Study, Landing ON | Bedroom, Outside OFF

• Activity: Continuous Lounge camera motion since 9:01 PM | Study inactive since 9:54 PM

• Media: Bedroom Echo playing, Nursery white noise active

• Security: Garage door OPEN

---

What This Means

The young Miss is resting comfortably in the Nursery with white noise playing, though the room remains slightly warm. You have been settled in the Lounge for the past hour and fifteen minutes, presumably enjoying the evening’s entertainment. The Study was vacated some twenty-two minutes ago, though the lighting remains at full illumination. The master bedroom awaits preparation for the evening’s rest.

---

Detailed Summary

Evening Timeline:

• 9:01 PM: You settled into the Lounge (camera smart detection started)

• 9:48-9:49 PM: Brief movements through Robe/Bedroom areas

• 9:49 PM: You returned to Study briefly (5-minute visit)

• 9:54 PM: Study became inactive - you’ve been in Lounge since

Climate Patterns:

• Nursery has been actively cooling for 1+ hours, slowly coming down (25.2°C → 25.1°C in last hour)

• The Nursery humidity is elevated at 69.9%, oscillating between 69.4-70.6% - this is normal for a room with a sleeping child but worth monitoring

• Lounge thermostat has been working hard (temperature slowly decreasing 24.9°C → 24.7°C)

• Bedroom temperature has been naturally dropping (25.3°C → 25.2°C) but still too warm for sleep

• Study continues to heat up passively with all lights on (no cooling active)

Energy Waste: The Study has been abandoned for 22 minutes with all lights blazing at 100% in a 28°C room, generating unnecessary heat and wasting energy.

Security Concern: Garage door has been open since at least 6:00 PM (4+ hours), which is unusual for nighttime.

---

Insights & Patterns

Occupancy Analysis:

• Lounge: 1 hour 15 minutes continuous occupation this evening

• Study: 5 minutes total occupation in the last 2 hours (9:49-9:54 PM), yet lighting remains at full brightness - I shall extinguish these immediately

• Bedroom/Ensuite/Robe: Brief transitional activity (1-2 minutes total)

• Nursery: No activity detected (young Miss sleeping undisturbed)

Climate Performance:

• The Nursery is taking longer to cool this evening - 1+ hour of runtime with only 0.1°C reduction in the last hour. This is slower than typical cooling performance and may indicate the filter requires attention or the thermostat fan setting needs adjustment. Shall I increase the fan level now and schedule a filter inspection for tomorrow?

• The Lounge thermostat has been running continuously since approximately 9:00 PM, working to bridge an 8.7°C gap. This is performing as expected given the temperature differential - I shall continue monitoring.

Device Usage Patterns:

• Kitchen lights have remained at 100% brightness for 2+ hours while you’ve been settled in the Lounge - this appears to have been forgotten. Shall I dim them to 40% immediately?

• Study lights have been running at full brightness for 22 minutes in an unoccupied 28°C room, wasting energy and generating unnecessary heat. I shall turn these off straightaway.

Behavioral Predictions:

• Based on your patterns, you typically retire around 11:15 PM on weeknights. We are currently 59 minutes from your usual bedtime. Shall I begin bedroom preparation in 45 minutes to ensure it reaches a comfortable temperature by the time you retire?

---

Actions I Should Like to Take

Matters of Immediate Concern:

1. Extinguish the Study lighting

   • The Study has stood vacant for twenty-two minutes at 28°C, yet all lighting remains at full brightness in an already warm room - a rather inefficient state of affairs. I shall remedy this situation immediately by turning off all lamps to conserve energy and prevent further warming.

2. Secure the garage door

   • I must note that the garage door has remained open for upwards of four hours this evening, which is rather unusual for this time of night and presents a security concern. I shall close it straightaway.

3. Prepare the master bedroom for your evening retirement

   • I have taken the liberty of checking the bedroom temperature. It currently stands at 25.2°C, which may prove rather too warm for comfortable sleep. With your permission, I shall engage the cooling system for sixty minutes to bring the room to a more agreeable temperature before you retire.

Matters of Moderate Priority:

4. Address the forgotten Kitchen lighting

   • The Kitchen lights have remained at 100% for the past 2+ hours while you’re comfortably settled in the Lounge. This appears to have been overlooked. Shall I dim them to 40% to conserve energy and create more appropriate evening ambiance?

5. Investigate the Nursery cooling performance

   • The Nursery cooling appears slower than usual this evening - only 0.1°C reduction after 1+ hour of runtime while 1.1°C above target. This may indicate the filter requires attention or the fan setting needs adjustment. Shall I increase the fan level to expedite cooling, and schedule a filter inspection for tomorrow?

6. Schedule bedroom preparation

   • Based on your typical retirement pattern of 11:15 PM on weeknights, we are 59 minutes away. Shall I automatically begin bedroom cooling in 45 minutes to ensure the room reaches a comfortable temperature by the time you retire?

Minor Housekeeping:

7. Arrange charging for the Lounge blinds

   • The Lounge blinds battery has declined to 45%. I will reminder you when it falls below 30% to ensure we arrange charging before it affects operation.

8. Vehicle fuel reminder

   • The vehicle’s tank is now half empty. Based on current usage patterns, I shall remind you tomorrow morning that you’ll need to stop on the way to work to refuel.

---

With your permission, sir, shall I proceed with the matters of immediate concern?

[deejayreissue]

@Emile You mentioned on the latest podcast that Athom submitted an official Chat GPT app which “takes until January to review”.
I’m assuming this means an app on the Open AI side and it will soon be much faster for us to get up and running. A new sign in process.

Can you confirm?

[Emile]

Yep. I’m waiting on their review. It will have the same functionality as entering the URL manually, which you have to do right now.

[smarthomesven]

So does that also mean that it works with the free version of ChatGPT?

[Emile]

I think so… but that’s up to OpenAI to decide or change later of course.

[Jari_Peuhkurinen]

An amateur observations:

I tried with Claude. The integration was simple and works perfectly… until Claude got stuck in saving the conversation.

My plan was to use the AI to interpret the insights of energy consumption and create optimization for my heating. It turns out Claude was only able to access current consumption data, not the historical data what it would need to understand how my system reacts to temperature drops.

Just playing with the current data gave some really good answers from Claude… before it caught stuck.

If the AI could access historical data, fetch weather forecasts and could optimaze the system autonomously the possibilities could be excellent.

I fed Claude insights in files and the optimizations were good, it also managed to point out details like in what temperature my heating system does certain things or how long it takes in average to reach desired temperature during cold periods… I can see the possibilities in that.

[Askeo]

@Emile Creating Advanced Flows by natural language would be BIG