KENZY.
Whole-home voice · yours to own

A voice assistant that lives in your home.

One assistant in every room — it knows where you are and who's talking, runs the AI you choose, and keeps your voice inside your own four walls. No account. No subscription. No one listening in.

Get started See what it does
$ curl -fsSL https://kenzy.ai/install.sh | bash
Listening · Kitchen
on-device
YOU"Hey Kenzy — turn off the lights downstairs."
KENZYDone — the downstairs is dark.
The wake word, your words, and Kenzy's reply all stayed on your own network — nothing was sent to the cloud.
Listens on-device
Stays on your LAN
Knows who's talking
Answers out loud
In every room

Every room. One brain.

A small speaker-mic per room, all managed from one place. Kenzy tracks which room you spoke in and who you are — so "turn on the lights" means these lights, and the reply comes back to you.

Kitchen
◍ online · enrolled: 2
Living room
◍ online · intercom ready
Bedroom
◍ online · alarms armed
Office
◍ online · fast-path on
Just talk

No app. No remote. Say the wake word and ask.

Run your home
"Turn off the lights downstairs."
Control lights, locks, and scenes through Home Assistant. Kenzy figures out which lights you mean.
Never forgets
"Remind me to call Mom at 8."
Set timers, alarms, and reminders by voice from any room, and cancel them the same way.
Knows you
"Remember that the gate code is 4312."
Per-person memory. She knows who's talking, and a private fact is only ever spoken back to you — "everyone should know…" shares one with the house.
Tell the house
"Let everyone know dinner's ready."
Speak in every room at once, or call one room from another. Automations can announce to the whole house.
Keep the list
"Add milk to the shopping list."
The list lives in Home Assistant, so it shows up on everyone's phone at the store.
In your pocket
"Is the garage door open?"
Ask from your phone through the Home Assistant app — the same Kenzy, your memory, your identity, answering in her own voice.
Ask anything
"How much sugar goes in sweet tea?"
Your own AI answers. When it needs current information, Kenzy can search the web and read back what it finds.
Instant
"What time is it?"
Everyday questions answer the moment you finish speaking, with no cloud round-trip.
Private by design

Your voice never has to leave the house.

Most assistants ship every word to a server you don't control. Kenzy flips that — audio, transcripts and model calls can all run inside your own network. No third party unless you put one there.

Nothing leaves
Audio, transcripts and model calls can all run inside your own network.
Bring your own LLM
LiteLLM speaks to local runtimes and every major provider.
No usage bills
Run it on hardware you already own. There is no per-question fee.
Yours to hack
Plain Python, readable configs, and a simple skill system.
Yours to take apart

Your model. Your skills. Your rules.

Because kenzy-llm runs on LiteLLM, point it at a model on your own box — or a cloud provider. Plain Python, readable configs, a one-file skill system.

configs/llm.yamlchoose your model
# a model on your own machine
model: "ollama/llama3.1"
base_url: "http://localhost:11434"
# or a cloud provider
# model: "gpt-5-mini"
# model: "anthropic/claude-sonnet-5"
skills/lights.pyadd your own
@skill
async def lights_off(room: str):
"""Turn off the lights in a room."""
await ha.call(
"light.turn_off", area=room)
return f"Done — {room} is dark."
Write one for real — the skills guide ↗
What's inside

Six small services. One pipeline.

A ground-up rewrite. Run them on one box, or scatter a featherweight node onto a Pi Zero in each room.

kenzy-node—
Room device: wake word, audio capture, playback.
kenzy-server:8765
The hub: connects rooms and runs the pipeline.
kenzy-stt:8767
Speech-to-text — local faster-whisper or cloud.
kenzy-llm:8766
The language model + skills, any LiteLLM provider.
kenzy-tts:8769
Text-to-speech — local Kokoro or cloud.
kenzy-speaker:8768
Voice identification — always runs locally.
How it all fits together ↗
The hardware

Any speakerphone. Any small board.

A room node runs on a cheap single-board computer with a USB speakerphone that has echo cancellation — that's what lets Kenzy hear you over her own voice. Use a separate more capable machine for the brain, or point the AI at your favorite cloud provider.

// pictured: Anker PowerConf S330 + 3D-printed cradle, Orange Pi Zero 2W inside [download the STL file]

A Kenzy room node: USB speakerphone in a 3D-printed cradle
Kenzy room node hardware, detail
Kenzy room node hardware, detail
From the maker
"She's not smarter than the AI you already know — she runs the same models. She's just there in the room, ready when you are."

I've wanted this since I watched Tony Stark talk to J.A.R.V.I.S. It's taken years for the technology to catch up, but we're genuinely close to that reality thanks to the work of a lot of brilliant makers.  Kenzy is built on top of their great foundational work to bring the vision to reality.

— John, USA · lnxusr1

Questions, answered.

Privacy
Does it work without the internet?
Yes. The wake word, speech, and AI can all run on your own hardware. You can choose a cloud model for the AI if you prefer, but the everyday commands never need the internet.
Cost
Do I need an account or subscription?
No. There's no Kenzy account and no monthly fee — you run it on hardware you already own.
Smart home
Does it work with my smart home?
Yes. Through Home Assistant, you can control lights, locks, scenes, and more by voice from any room — and Kenzy can answer from the HA companion app on your phone, too.
Setup
Is it hard to set up?
One command installs it on Linux — it even asks where you want the AI to run, cloud or your own hardware. A full manual path is available too, along with a web dashboard for managing every room.
Hardware
What do I need to run it?
A small single-board computer and a speakerphone with echo cancellation in each room, plus one more capable machine for the server and AI. You can also point the AI at a cloud model.
Privacy
What actually leaves my network?
As little as you want, including nothing at all. The wake word and voice recognition always stay local, and speech-to-text is local by default. Skills that reach the internet only run when you use them.

Your home finally has a voice.

Install it, run the services you need, point it at your choice of hardware, and start talking.

$ curl -fsSL https://kenzy.ai/install.sh | bash
Get Started ↗ View on GitHub ↗