Run your home
“Turn off the lights downstairs.”
Lights, locks, scenes and more through Home Assistant — and Kenzy works out which lights, even the seven switches your kitchen somehow has.
// Private · Whole-home · Open source
A voice assistant that lives in your home — not someone else's cloud.
Kenzy puts an assistant in every room of your house — one that knows which room you're in, recognizes who's talking, and can answer the whole house at once. It's powered by the AI you choose — a model on your own machine or a provider you pick — and it runs on hardware you own, so your voice never has to leave the building. No account, no subscription, no company listening in.
$ curl -fsSL https://kenzy.ai/install.sh | bash
See what it can do ↓
// installs on Linux in one command — or step through it manually on Get Started
Every room
One assistant
Private
Your voice stays home
Your choice
Local or cloud AI
Knows you
Recognizes your voice
// What it can do
Just talk — in any room.
No app, no remote, no phone. Say the wake word and ask. Kenzy works out which room you're in and who's asking, then does the thing or answers out loud.
Never forgets
“Remind me to call Mom at 8.”
Timers, alarms, and reminders — spoken into the air, from any room. Ask what's left, cancel by voice, no phone to dig out when it goes off.
Tell the house
“Let everyone know dinner's ready.”
Kenzy speaks your message aloud in every room at once — or calls one room from another, two-way.
Keep the list
“Add milk to the shopping list.”
The list lives in Home Assistant, so it's on everyone's phone at the store — Kenzy just makes adding to it effortless.
Ask anything
“How much sugar goes in sweet tea?”
Your own AI answers — and when it needs current information, Kenzy can search the web and read back what it found.
Instant
“What time is it?”
The everyday stuff — time, lights, timers — answers the moment you finish speaking. No model call, no cloud round-trip.
// In every room
One assistant, all over the house.
Put a small speaker-mic in each room and they all share one AI brain. Kenzy keeps track of which room you spoke in and who you are — so “turn on the lights” means these lights, and the reply comes back in your room.
Every room
Start small, grow anytime
A cheap little node per room, all managed from one place. Begin with one and add rooms whenever.
Room-aware
Knows where you are
Ask in the kitchen and it acts in the kitchen — context follows the room, not a single gadget.
Knows you
Recognizes who's talking
Enrolled voices get personal answers, and sensitive actions can require the right person.
Announce & intercom
Reach the whole house
“Tell everyone dinner's ready” plays in every room at once — or call one room from another for a two-way intercom.
01 Why local
Your voice never has to leave the house.
Most assistants ship every word you say to a server you don't control. Kenzy flips that: because kenzy-llm runs on LiteLLM, you can point it at a model running on your own box — Ollama, LM Studio, vLLM — or a cloud provider if you'd rather. Your call, per service.
Private
Stays on your LAN
Audio, transcripts, and model calls can all run inside your own network. No third party in the loop unless you put one there.
Open model
Bring your own LLM
LiteLLM speaks to local runtimes and every major provider. Swap models with one line of YAML — no rewiring.
No meter
No per-word bills
Run it on hardware you already own and the marginal cost of "what's the weather?" is electricity, not API credits.
Hackable
Yours to take apart
Plain Python, readable configs, and a one-file skill system. Built to be tinkered with, not locked down.
02 What's inside
A real assistant, assembled from parts you can see.
kenzy-node
On-device wake word
openWakeWord runs on every frame locally, with an optional Silero VAD gate to kill false triggers. Train and drop in your own wake word.
kenzy-speaker
Knows who's talking
SpeechBrain ECAPA-TDNN identifies enrolled speakers — so unlocking the front door by voice can require a recognized person.
kenzy-llm
Skills & tool-calling
Drop an async function in skills/, decorate it with @skill, and the model calls it as a tool. No registration, no boilerplate.
fast path
Instant commands
Common phrases like "turn on the lights" resolve deterministically — no model round-trip — so they answer the moment you finish speaking.
v3 · GROUND-UP REWRITE
Rebuilt from the ground up.
Kenzy v3 is a complete redesign — not a refactor. The monolith is gone, replaced by six small services that each do one job and talk over a simple WebSocket protocol.
The result is a system you can spread across the house: a featherweight node on a Raspberry Pi Zero 2 W in each room, the heavy lifting on a server or workstation wherever you've got the horsepower.
- More flexible — run every service on one box or scatter them across the network.
- More responsive — a deterministic fast path skips the LLM for everyday commands.
- Edge-ready — nodes do only wake word + capture, so they fit tiny hardware.
- Hard to knock out — if a cloud model fails, Kenzy can quietly retry a local one; and if something truly breaks, she says so instead of going silent.
# point Kenzy at a model on your own machine model: "ollama/llama3.1" base_url: "http://localhost:11434" # ...or a cloud provider, same two lines # model: "gpt-4o" # model: "claude-opus-4-8"
Ollama
LM Studio
vLLM
OpenAI
Anthropic
any LiteLLM provider
// How it works
Listens, thinks, answers.
Step 1 · Hears you
Wakes on your word
A small speaker-mic in the room catches the wake word on the device itself — nothing is sent anywhere until you actually call on it.
Step 2 · Thinks
Works out what you meant
Your words become text and your chosen AI figures out the request — and which room and which person it came from.
Step 3 · Acts
Does it, and answers
It runs the command — your lights, a question, an announcement — and speaks the reply back in your room.
// From the maker
Why I built this.
I've wanted this since I watched Tony Stark talk to J.A.R.V.I.S. — an assistant so woven into his life it was basically an extension of him. For a long time that was wishful thinking: the speech recognition wasn't good enough, the synthetic voices sounded like robots, and the "brain" didn't exist yet. Then, one piece at a time, the technology caught up. I started building Kenzy about seven years ago and shipped the first release in 2020, during the lockdown — for my own house first, and now for yours.
Honestly, most of the technology inside isn't mine. Kenzy is assembled from the work of a lot of brilliant makers — openWakeWord, Whisper, SpeechBrain, Kokoro, the open model ecosystem — and my job was mostly to put the pieces together so they disappear into the walls. She's not smarter than the AI you already know; she runs the same models. She's just there — in the room, already listening, no phone to dig out.
The other day my wife Nicki was making sweet tea and asked me how much sugar to use. I just said "Hey Kenzy" — and she answered, out loud, in the kitchen: a cup per gallon, more if you want it southern sweet. Nobody touched a screen. That's the whole idea, and it's why the reminders, the lists, the timers exist too: Nicki asked, so I built them.
— John somewhere in the Southern US · lnxusr1 on GitHub
nicki: how much sugar goes in the tea? john: hey kenzy — how much sugar should we use? kenzy: One cup per gallon for sweet tea — a cup and a half if you want it southern sweet.
// a real exchange. no app was opened in the making of this tea.
// Common questions
The honest answers.
Privacy
Does it work without the internet?
Yes — wake word, speech, and the AI can all run on your own hardware. Prefer a cloud model for the "brain"? Opt in piece by piece — and if the cloud fails, Kenzy can quietly fall back to a local model, and the everyday commands never needed the internet at all.
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 — control lights, locks, scenes and more by voice, from any room.
Setup
Is it hard to set up?
One command installs it on Linux. There's a full manual path too, and a web dashboard to manage every room once it's running.
Hardware
What do I need to run it?
The room nodes are designed to run on small single-board computers (SBCs) like the Raspberry Pi, paired with a speakerphone that has echo cancellation — that's what lets Kenzy hear you over her own voice, and it's what powers the two-way intercom and self-silencing alarms. (On a speaker without it, Kenzy adapts honestly: those two features politely decline rather than misbehave, and everything else works.) You'll also want one more capable machine for the server and AI — or point the AI at a cloud model. See the hardware guide in the docs for tested boards.
Privacy
What actually leaves my network?
As little as you want — including nothing at all: every stage can run on your own hardware, and the Running Fully Local guide shows exactly how. Out of the box, the wake word and voice recognition never leave (no cloud version of them exists), speech-to-text is local by default, and the AI "brain" and speaking voice start on a provider for the easiest setup — swap in local models whenever you like. Skills that fetch the outside world (weather, news, web search) only make requests when you use them, and you can turn any of them off.
Music
Can it play music in every room?
It's on the roadmap — multi-room music is a feature we're working on. For now, you can pair Kenzy with Home Assistant and Music Assistant for whole-home audio.
Languages
What languages does it speak?
English, today. Multi-language support is on the roadmap but not here yet — we'd rather tell you now than after you've set it up.
// Bring it home
Your home finally has a voice.
Install it, run the services you need, point it at your hardware, and start talking. The docs walk you through every step.