# Crazy idea: homey AI

**URL:** <https://community.homey.app/t/crazy-idea-homey-ai/145853>\
**Category:** Ideas & Suggestions\
**Created:** [November 24, 2025, 5:42pm UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853 "2025-11-24T17:42:08Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![Stijn\_vdh](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/stijn_vdh/32/17439_2.png) [@Stijn\_vdh](https://community.homey.app/u/Stijn_vdh)\
**Post date:** [November 24, 2025, 5:42pm UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853/1 "2025-11-24T17:42:08Z")

</div>

Hi,  
maybe a crazy idea…. and I know MCP is still rather new…. also having a payed API access to these big LLM’s is a little to much…. so why not have….  
Homey having an LLM hosted for API access….  
a new Homey AI device, which has a battery and can be charged, that can do a little more then the current homey pro, has camera, etc…. and ofc can see all homey’s and devices in the house….. the homey AI device has a small local GPU but uses the central AI for the real LLM brain. (I know this is possible because we have a similar solution)

the flows get an AI upgrade, meaning that new applications can be connected like your agenda, emails,…..

I would trust this setup over these big commercial players.

---

<div class="post-metadata">

**Author:** ![robertklep](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/robertklep/32/160628_2.png) [@robertklep](https://community.homey.app/u/robertklep)\
**Post date:** [November 25, 2025, 6:13am UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853/2 "2025-11-25T06:13:22Z")

</div>

> [@Stijn\_vdh](#):
>
> Homey having an LLM hosted for API access….

Surely you have seen the multiple apps that have been released recently to connect Homey to LLM’s?

---

<div class="post-metadata">

**Author:** ![Stijn\_vdh](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/stijn_vdh/32/17439_2.png) [@Stijn\_vdh](https://community.homey.app/u/Stijn_vdh)\
**Post date:** [December 1, 2025, 8:07pm UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853/3 "2025-12-01T20:07:49Z")

</div>

sure, but I did not mean only apps…

---

<div class="post-metadata">

**Author:** ![robertklep](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/robertklep/32/160628_2.png) [@robertklep](https://community.homey.app/u/robertklep)\
**Post date:** [December 2, 2025, 5:42am UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853/4 "2025-12-02T05:42:51Z")

</div>

I think I misunderstood, and that with _“Homey having an LLM hosted for API access”_ you mean _Athom_ having a cloud-based LLM.

If you look at Athom’s parent company LG, and what they intend to do with “their Homey” (the [LG ThinQ ON device](https://www.lgcorp.com/media/release/28061)), it sounds like they’re already working on such a device.

However, the question remains if you can trust LG. They already collect loads of user data with their other smart devices, and [Homey is a very good device to collect more](https://www.channelnews.com.au/breaking-news-lg-aquires-home-automation-platform-new-way-to-capture-sell-data-on-customers/).

---

<div class="post-metadata">

**Author:** ![Stijn\_vdh](https://sea1.discourse-cdn.com/flex025/user_avatar/community.homey.app/stijn_vdh/32/17439_2.png) [@Stijn\_vdh](https://community.homey.app/u/Stijn_vdh)\
**Post date:** [December 6, 2025, 1:51pm UTC](https://community.homey.app/t/crazy-idea-homey-ai/145853/5 "2025-12-06T13:51:55Z")

</div>

Yes, indeed athom having a cloud LLM, the homey ai device than would need a small GPU for a distributed setup