pretty please. That would be so awesome 
There are several open-sourced community versions available already. They work fine for everyday use but could definiately do with some more specialized attention of a true programmer.
Custom Apps and Drivers / Hubitat Package Manager - HubitatI am aware of those. They seem to be experiments or hobby projects. no offense to the creators but I have no idea if theyāll be maintained or not even if theyāre working or not and a few of them are lacking features.
If you are actually using Ai you can very very easily maintain or fork any of the existing community mcp servers. Easy to maintain with Ai.
Drop this perception of barriers.
1 Like
I am VERY much involved in MCP topics professionally so have some context in this answer beyond Hubitat. BUT...MCP is heading in the right direction, but itās still early.
The protocol is solid, but most of the ecosystem around feels like prototypes, fragmented implementations, and unclear long term support. The real momentum right now is happening in higher level agent frameworks and tool layers, which are moving faster and abstracting MCP away. Until MCP settles and gets more consistent, enterprise-grade adoption end to end, itās hard to call it as the ring to rule them all. So to your thought, it is an enterprise hobby project. Everyone is learning.
1 Like
Some, Iām sure. āEverybodyā sounds like hyperbole to me.
4 Likes
Oh no, everybody wants to learn to use Yaml. It's true. 
1 Like
Sorry, couldnāt resist. Iām actually a fan of Home Assistant too. And yeah, obviously AIā¦ but still funny.
# Enterprise-grade automation platform (not a weekend Raspberry Pi hobby)
platform:
name: "Actually Reliable Automation"
version: "1.0"
description: >
A bold attempt at running automation without YAML-induced emotional damage,
random restarts, or explaining to guests why the lights stopped working again.
standards:
protocol: "MCP"
message_format: "json"
reliability: "predictable (wild, I know)"
features:
- name: "Lights turn on"
behavior: "consistently"
note: "No 47-line YAML dependency chain required"
- name: "Automation logic"
behavior: "human-readable"
note: "No need to summon a forum thread from 2018 to debug it"
- name: "System uptime"
behavior: "stable"
note: "Does not depend on a Zigbee device named 'lamp_3_final_final_v2'"
- name: "Integrations"
behavior: "maintained"
note: "Not abandoned by someone who lost interest mid-commit"
- name: "Error handling"
behavior: "explicit"
note: "Not āworks unless it doesnāt, then good luckā"
anti_patterns:
- name: "Home Assistant mode"
symptoms:
- "Restart required to apply a typo fix"
- "Automations silently failing since last Tuesday"
- "Dashboard works perfectly until someone actually needs it"
- "Forum says ājust use Node-REDā"
resolution: "Seek adult supervision or replace entire stack"
scaling:
approach: "intentional architecture"
vs_home_assistant: "hope and YAML"
closing_statement: >
Turns out, when you treat automation like software instead of a weekend experiment,
it stops acting like one.
2 Likes
I did that once, never again! I suggest everyone try it out, kick the tires... you'll be back and have a much greater appreciation for what we have here. Solid software/hardware, great official support who regularly come on here to help. Amazing community. The list goes on.
Could it be better, of course. But right now - this is the place to be and it's only getting better.
BTW, YAML isn't that bad once you get used to it. But I swear they over complicate everything on purpose. Think of Hubitat as Windows and HA as Linux. Sure one is more powerful but the other is so easy to use!
2 Likes
I also been burned by HA many years ago. Currently, like many others, I run it alongside HE - to bring in specific devices into HE.
The leaps and bounds that the HE platform has progressed by in the last few months is under appreciated. And the next platform releases promises major additional integrations.
3 Likes
Maybe Iām the dumb one, but Iām still struggling to find a way to make AI useful with Home automations HE or HA.
I have siri for voice, she works most of the time and is SUPER easy to setup. Anything i have found that works better than Siri/Alexa/Gemini for voice remote control, requires a considerably more difficult set up.
I know AI can be used for code, but there are SO many community apps and drivers anything left seams super niche but maintaing code that already exist seems to be a good fit.
So there is the possibility of AI writing automations, but at some point.. the things that can be automatedā¦ will be. Unless i find some new weird project iām like 95% of the way there.
I legit have a bit of FOMO over this and iām struggling to see what all the hype is in the context of making home automation cooler and easier. What am i missing?
4 Likes
For what HA/HE actually do today, youāve probably already squeezed out 95% of the value and AI doesnāt add much beyond being a science project.
Where it gets interesting isnāt better rules or voice. It is when the system can handle intent and context on its own, like āmake mornings smootherā and it adjusts lights, temp, and routines automatically with out actual defining rules like today. And it maintains that smoother and learns how we like it based upon environmental factors (like you are not home, but the dog is).
We are obviously NOT there yet.
2 Likes
This is kinda one of my sticking points with AI. Iām doing this already with my ~20w brain and my ~5w HE hub and i have no choice but to keep my 20w brain running and mildly entertained. My morning automations are so smooth iāve have had previous co-workers comment that i come in to work like iāve walked off a cloudā¦ on ~25w. Yes, i know iām special.
With HA, itās relatively easy to get AI and LLMās setup and be performant, but after i got it all setup i was left with ānow what?ā and my mini is consuming 100w.
Maybe you are right AI is still to new to be useful in the home automation space for most users. I was really hoping to teach this old dog some new tricks and i had just missed something.
I still think AI needs to get to Black Mirror's "White Christmas" episode of quality to be useful to most consumers and everything will need to communicate to everything to pull it off.
2 Likes
The value of AI and MCP server for Hubitat is not to "Setup some AI managed lighting automation"
It is to do stuff like:
- Make a rule to turn the porch lights on at dusk and off at dawn
- I have christmas lights up on Device "Exterior_Switch", make them turn on at dusk and off at dawn...and pause the porch light rule for now
- Tell me how often the backyard light turned on in the last 30 days
- Tell me which devices havent had comms in more than 2 days
- Read the logs from my RoomLighting and tell me why the lights might be blinking when when I walk in the room instead of just turning on
- Find me all the devices that report low battery
- Write a rule to alert me whenever any battery goes below 20%
And eventually stuff like:
- Whenever batteries fall below 30%, begin watching amazon and order replacements when the price is 20% or more below normal
