SLAW separates control from execution. The control plane coordinates tasks, tracks state, and enforces governance. Execution happens wherever the agent adapter runs — locally, in a container, or remotely.
Control-plane vs execution model: the SLAW API coordinates work; agent adapters execute it in isolated workspaces.
The Operator (you) can pause, cancel, or override any agent at any time from the Board. Budget limits enforce token spend caps. Approvals require explicit Operator sign-off before agents can proceed with sensitive actions.