Meet Galileo OS
Most AI today lives somewhere you don't control. Your prompts, your documents, and your automations travel to a provider's servers, get processed on hardware you'll never see, and come back with a bill attached to every request. That works — until the prices move, the terms change, the model you depend on gets deprecated, or the data simply shouldn't leave your building in the first place.
Galileo OS is built on the opposite premise: your AI should run on hardware you own.
What it is
Galileo OS is a full Linux operating system for personal AI computers. It isn't an app you install on top of Windows or macOS, and it isn't a cloud service. It's the operating system itself, imaged onto a dedicated machine. Once installed, that machine becomes an always-on AI appliance for a person, a household, a team, or an office.
You don't sit at it. You don't plug in a keyboard and monitor. You manage it the way you manage your router: open a browser on any device on your network and reach Galileo's web interface. The machine quietly runs in the background and serves everyone connected to it.
That's the cleanest way to picture the whole thing. A router is a box you plug in once, that's always on, that you configure from any device's browser, and that does its job for the whole network without anyone thinking about it. Galileo is that, for AI.
What's inside
Galileo composes a handful of building blocks into working AI systems.
Models run locally. Galileo runs open large language models on the machine's own compute through Ollama. You pull models from Ollama's registry and run them on hardware you own — no API calls, nothing leaving the device.
Agents do the work. An agent is a model plus its configuration: its instructions, its permissions, its memory. Agents are the actors you put to work inside automations.
Workflows chain it together. A visual, drag-and-drop workflow builder lets you connect agents, app actions, and logic into multi-step automations. A workflow can be triggered by chat, on a schedule, or by a webhook — one automation, many ways to start it.
Apps connect to the outside. Integrations link Galileo to external services like Drive, GitHub, and ClickUp, each contributing its own actions to the workflow builder.
Channels meet you where you already are. You reach your agents through Discord, WhatsApp, Telegram, email, or direct chat. There's no new app to install — you talk to your AI from the surfaces you already use.
A knowledge base keeps your documents private. Galileo includes built-in retrieval. You index your own documents into a local knowledge base, and agents draw on it while they work. The documents are embedded, stored, and queried entirely on your machine — nothing is uploaded to a cloud vector store.
When an agent runs, it works inside a temporary container that's created for the task and destroyed afterward. Nothing carries over between runs unless you explicitly store it, which keeps each task clean and isolated. All of it runs on the machine you own, and all of it works offline.
Why it exists
Three things make local AI worth owning rather than renting.
The first is speed. You could assemble something similar yourself — a model runner, a chat interface, a workflow engine, a vector database, a handful of channel bots. But that's weeks of configuration and a stack of tools that break independently and need ongoing maintenance. Galileo ships the integration already done. The honest version of this claim: the do-it-yourself route can eventually get you to a similar place, in weeks of plumbing. Galileo gets you there in days because the work of wiring it together is already finished.
The second is economics. Cloud AI charges per token or per request, and that bill scales with usage indefinitely. Galileo runs inference on hardware you own, so once the machine is paid for, running ten thousand tasks costs roughly what running ten costs. No subscriptions, no per-request fees.
The third is independence. When your models, your data, and your automations live on your own hardware, no provider can raise your prices, change the terms, deprecate a model you rely on, or cut off your access. You own the stack.
Where it's going
What ships today is the full system described above — running in production, demoable live, and capable offline. The first-party apps are built by our team.
What's ahead is opening that up. We're working toward a documented SDK and developer program so third parties can build their own apps, and eventually an ecosystem and app store around them. Those are planned, not present — today the apps are first-party. We'd rather be precise about that line than blur it.
In short
Galileo OS turns hardware you own into a personal AI computer: your models, your data, your network. Set up in days, reachable from any device, with no per-use bill and no dependence on the cloud.
If that premise interests you, we'd like to hear from you.
Talk soon! Mikael