About Xymotrone
Xymotrone exists because most browser automation breaks the moment a page changes. Scripts are brittle; a model that can look at the page and decide the next action is not. Xymotrone gives that model a real Chrome session, a clear goal, and hard boundaries — then keeps a full record of what it did.
It is built for people who already have the pieces: a browser provider account, their own proxies, their own logins, sometimes their own model server. Xymotrone is the layer that ties them together and shows the work.
How we think about it
Every step is visible
The model's reasoning, the action it chose, and a screenshot of the page are stored for each step, so a run can always be replayed and explained.
Your keys stay yours
Browser provider keys, proxy credentials and stored logins live in your workspace and are only ever read server-side — never returned to the browser.
Limits by design
Step ceilings, run deadlines, allowed-domain lists and per-plan concurrency exist so an agent cannot wander outside the work you asked for.
Bring your own stack
Hosted models or your own Ollama, LM Studio or vLLM endpoint; any HTTP, HTTPS or SOCKS5 proxy. Xymotrone coordinates, it does not lock you in.
Try it on a real task
Create an account, connect your browser provider and proxies, and watch an agent work through a goal step by step.