- homestead, our free app (the door opens soon)
- step-by-step guides
- pick-a-model cheatsheet
- no account. no catch.
we put private ai on
your machine.
and keep it running.
you don't have to learn anything or touch a terminal. we set up a private, local ai on the computer you already own. pay once, own it, $0 a month after. no cloud. no spying. no water bill. what it learns about your work stays yours.
pay once. own it.
the ai itself is free and runs on your machine without us. you can do it yourself, or pay us once to set it up right. the monthly add-on is optional, and nothing breaks if you cancel it. that's the whole model.
- we remote in (60–90 min)
- install the engine + a clean chat app
- the best current model for your machine
- privacy + telemetry locked down
- 30 days of "did it stick" support
- money back if your machine can't run it
- your ai's memory: local, readable, rot-checked. MEMORY INTACT, or we tell you.
- model upgrades as they ship (they ship monthly now)
- quarterly "is this still optimal" checkup
- priority help from a human who knows your setup
- a monthly memory rot-report: proof it still remembers you right
- the mechanic, not the rent
prices in usd. managed requires a setup first. cancel managed anytime and your ai keeps working. it's your machine. already set up? add managed alone → · prefer to talk first? book a call →
same deal, more machines. private ai on every laptop, a shared playbook for the team, one person to call.
the guides are free. so why pay?
honest answer: you can absolutely do this yourself. most people just don't want to spend the weekend. same result, two ways.
- read the guides + install homestead-memory (free, mit)
- you install ollama, pick a model, wire up the chat app, tune the privacy settings
- a weekend if you're comfortable in a terminal
- you own the result, forever
- same result, we remote in and do it in about an hour
- we pick the best model for your exact machine and lock every privacy setting
- homestead-memory set up so it remembers you, and it catches rot before you trust it
- 30 days of support. $0/mo after. you still own it.
rent vs own.
the cloud charges you forever and can shut you off whenever. ownership costs once. here's three years, no funny accounting:
it just works.
the install
ollama as the engine, a friendly chat app (open webui) as the face. no command line required after we leave.
the right model, not a brand
we're not locked to any one company. we pick the best open model that fits your machine and your work — llama, qwen, deepseek, mistral, gemma — and as better ones ship, we swap you onto them.
privacy, locked
telemetry off, no accounts, no phone-home. the only place your prompts go is your own disk.
kept current
open models improve every few weeks. on managed, we swap you onto the better one so you never fall behind.
a human on call
when something acts up, you message a person who knows your setup. not a bot, not a queue.
your data, yours
optional encrypted backup of your config and notes. never your prompts, never your conversations.
your memory, verified
your ai's memory lives in plain files on your disk. you can open them, read them, keep them. and it checks itself for rot: contradictions, dead sources, silent drift. MEMORY INTACT or ROT DETECTED. no black box.
probably yes.
the thing nobody tells you: you don't need a gaming rig. a normal laptop from the last few years runs a genuinely useful model. we check before you pay, and if your machine can't handle it, we tell you and you owe nothing.
three steps. no jargon.
you book
pick a time. tell us what machine you have and what you want ai for. two minutes.
we set it up
we remote in, install everything, pick your model, lock it down. you watch if you want.
you own it
it lives on your machine. you use it like chatgpt, except private. on managed, we keep it healthy.
the honest faq.
wait. you hate subscriptions, and you're selling one?
fair hit. here's the difference: rent is when they can shut you off. nobody can shut you off. the ai is free, lives on your machine, and runs without us forever. setup is a one-time job, like paying a plumber. managed is an optional human mechanic on call, and if you cancel, nothing breaks. that's not rent. that's a mechanic.
is it actually private?
yes. the model runs on your computer. your prompts and answers never touch our servers or anyone else's. we set it up so it can't phone home. no agents, no plugins, no remote code either. just you and your ai.
is it as good as chatgpt?
for most everyday things, yes: writing, summarizing, drafting, answering, light coding. the open models crossed that line this year. for frontier-hard reasoning the big clouds still edge it. we're honest about where the line is and help you keep a cloud option for the rare case if you want.
what's homestead?
our free app. a private chatgpt for your own laptop that remembers you on YOUR machine and knows exactly what your hardware can run. it's almost ready, and the waitlist is on the homepage. the studio setup works today, with or without it, and homestead will slot right in when it ships.
what if my computer is slow?
most laptops from the last few years run a capable model. if yours genuinely can't, we tell you before you pay and refund the setup. no surprises.
what if it breaks?
setup comes with 30 days of support to make sure everything's solid. after that, managed means you message us and we fix it. no managed? we'll still quote you a one-off fix, honestly.
can i cancel managed?
anytime. it's your machine and your model. cancel and it keeps working. you just stop getting upgrades and a human on call.
do you work with small teams?
yes. $499 sets up to 5 machines, +$99/mo if you want us maintaining them. bigger or regulated org? see the audit path below.
your own ai.
set up this week.
book a call. we'll tell you in five minutes whether your machine's ready and exactly what it'll cost. no pressure, no sales script.
regulated team? run a free ai-egress audit →