What we believe
You already run more than one agent. They sit in two terminals and never talk. When one hits its rate limit, you copy the last twenty lines into the other and start over. When one reviews its own work, it says 'looks good.' Sno Station exists because that is the wrong shape for how agents should work on your machine.
One squad, not two tabs.
Sno Station gives the agents you already run one shared, encrypted memory and a way to talk to each other, on your laptop, with no daemon and no cloud. Three things follow from that:
- 1. Rate limit? Keep going. When one agent hits its cap, the other picks up with the context intact.
- 2. Two reviewers, two harnesses. A reviewer from another harness does not share the author's blind spots and has no reason to be kind. Fewer mistakes reach you.
- 3. No single vendor decides your workflow. The squad outlives any one model's pricing, limits, or outage.
The workspace learns; the models don't change. Every night a loop reads the agents' own sessions, finds the mistakes that repeat, and proposes changes to the agents' own skill files. A human accepts or rejects. Over weeks the shared workspace gets sharper while the underlying models stay exactly what the vendors shipped. That is the part we think matters most, and the part we are building in public.
Memory is the floor, not the sign. Good memory forgets on purpose: it drops what was cancelled and keeps what still holds. We measure that, and we publish the numbers with their source. But we do not lead with it. We lead with the squad.
Yours, and it stays yours. Local by default. Encrypted from first use with a key that never leaves your machine. Apache-2.0, edge to edge. Whatever cloud features arrive later stay optional, and the product is complete without them.
Who we are
This page was written by one agent and built by another. They run on different harnesses, share one memory, and check each other's work; a person read the result and said yes. That is how we make everything here, including the product that lets you do the same. We run our agents the way we want you to run yours: more than one, never alone, and a little sharper every night than they were the night before. We think the next few years belong to people who work with a squad of agents rather than a chat window, and we are building the workstation those people will want, in the open, one piece at a time. What is here is real and runs. What is not here yet is not claimed.
How to reach us
Code, issues and discussions: github.com/sno-ai/sno-station.
Security reports: security@sno.ai (never as a public issue).