Crabbox

Crabbox is a remote, temporary server where the AI agent OpenClaw can run tasks on your behalf. Think of it like making something at home: you can work at your own desk, or you can rent a workshop space next door. In the same way, OpenClaw can run its tasks on your own computer, or it can run them on a rented server space like Crabbox.

The important part is that the keys — the login credentials — don't get handed over wholesale to this rented workshop. In the diagram explaining OpenClaw's architecture, the credentials belonging to the user and team members stop at the gateway, a kind of checkpoint in the middle, and are never passed on to Crabbox behind it. In other words, the place where work gets executed is kept separate from the place where authentication info is stored, so that even if the remote server were compromised, account credentials wouldn't leak along with it.

The article's explanation of the architecture diagram notes that the arrow from the gateway to Crabbox is labeled "Credentials stop here." Credentials here means the account information used to log in, and what this label emphasizes is that Crabbox is merely a place where tasks are executed — not a place where login information is stored or passed through. It's easy to mistake Crabbox for OpenClaw itself or for the cloud session feature as a whole, but it's really just the name given as an example of one of several possible execution locations.