AI that runs on your machine.
Radovar builds local-first AI. Creator, our desktop application, runs agents on your computer and keeps your code there.
Our product
Creator turns a local machine into an AI workstation.
Creator connects AI agents to a real project on disk. They read the files, propose changes, and run the checks, in the same place you already work.
Nothing is uploaded in order to run. That makes it usable in teams where sending source code to a third party was never an option.
Runs on the desktop
A native application, not a browser tab pointed at someone else server.
Works on real files
Agents open the actual project, not a copy pasted into a chat window.
Local models supported
Run against a model on your own hardware when the work cannot leave.
What we hold to, product after product.
Local by default
The useful path is the private one. Running locally is the normal mode, not a restricted mode you have to ask for.
Your code stays yours
No silent copies, no training on your work, no exceptions buried in a settings panel.
Real files, real results
An agent that cannot open your project cannot help with it. Ours work against the actual repository.
Built to be checked
Every change an agent makes is visible and reviewable before it lands. You stay the one who approves.
How Creator actually works.
Point Creator at a project you already have.
No migration, no upload step, no new home for your source. Creator opens the folder where the work already lives.
Agents work in place
They read, edit, and run against the real files on disk.
You review every change
Each edit arrives as a diff you approve or reject.
Choose the model, including one that never leaves the machine.
Creator runs against hosted models when you want the power, and against local models when the work cannot go anywhere.
Radovar is the company behind Creator.
We are a small technology company working on one question: how much of modern AI tooling can you get without handing over the thing you are working on.
Most tools answer that by asking you to trust a server. We would rather move the capability to where the work already is. That constraint shapes the whole product, from how agents get access to files to which models we support.
Creator is the first thing we have built to that standard. It will not be the last.
Talk to us about Creator.
Questions about the product, a team rollout, or how the local setup works.