Jarvis v1 · open source · in development
A system that doesn't
stop at answers.
One system that understands context, plans work, uses real tools, and keeps you in charge. And this is only the beginning.
Be there when v1 drops. No filler, ever. Privacy
A useful assistant doesn't leave you a wall of text to interpret. It figures out where you want to go, splits the problem, picks the tools, runs the steps, and shows you what happened.
What “operational” means
01
Understand
Context from your files, notes, past conversations, and goals, not just the last message you typed.
02
Plan
A vague request becomes a concrete plan with steps you can read, edit, and reject before anything runs.
03
Act
Real tools: browser, terminal, code, files, and automations. Not another wall of text to interpret yourself.
04
Report back
What it did, what it couldn't do, and why, in plain language, with a log you can actually check.
The stack
Three layers, one system.
Nothing here is finished. That's the point of following a build. Each layer is labeled with its real state.
No magic claims
We'd rather show you what works than promise you what doesn't exist yet.
Every feature on this site is labeled: available, in development, planned, or research. If something is only an idea, we say so. The interesting part is watching it become real.
From the lab notes
“An agent isn't useful because it looks smart for five minutes. It's useful when it finishes a task, handles its own errors, admits its limits, and produces something you can verify.”Read about our research →
not research · not a roadmap item · an idea we're serious about
An operating system with AI in its bones.
macOS, Windows, and Linux were built for people clicking. We want to build the first OS natively integrated with AI and actually agentic: where the system doesn't wait for your next click, it works alongside you. It's far away. It's also where all of this is pointed.
FAQ
The questions worth asking.
Short answers, no spin. Anything missing? Write to us and we'll answer, then add it here.
Solen Lab
A small, open space for studying what actually makes AI systems useful.
Benchmarks, agent evaluations, open-source experiments. If you want to help build this from the very beginning, transparently and on your own time, the door is open.
Visit the Lab