How Aboard Works
Understand the intent-to-result loop shared by users, Agents, Workbench, and Host.
Aboard separates understanding, planning, review, and execution so consequential work remains visible and recoverable.
Intent -> Context -> Plan -> Preview -> Review -> Confirm -> Execute -> Report -> Recover- Intent: a user states a goal and constraints.
- Context: Workbench and Host establish the relevant Workspace facts.
- Plan: an Agent or product workflow proposes typed changes.
- Preview: requirements, warnings, side effects, and expiration are calculated without mutation.
- Review: the user sees what will change and who will execute it.
- Confirm: the current principal gives explicit consent where required.
- Execute: the authoritative Host performs the mutation.
- Report: the system returns a structured result.
- Recover: failures, expiry, replay, and rollback have explicit paths.
The public site describes this model and never becomes a second runtime authority.
How is this guide?