Claudio

Claudio is a blueprint for an AI chief of staff in a regulated organisation. The repository defines the operating model, security posture, memory boundaries, connector assumptions, and approval gates needed before such an assistant should touch executive work.

The intended interface is Slack. From approved work context, Claudio can prepare meeting material, synthesise priorities, draft documents, maintain decision and action logs, and coordinate follow-up. The conversational surface stays familiar even when the work crosses several enterprise systems.

Files for identity, tools, users, startup, runtime checks, memory, and setup make the assistant’s authority legible before any production system is switched on. The repository is an implementation baseline, not a simulation of a finished enterprise product.

The difficult design problem is usefulness without ambient authority. Claudio needs enough context to understand governance, liquidity, customer outcomes, executive time, and cross-functional dependencies. It also needs least-privilege access, auditable connector use, enterprise identity, scoped storage, retention rules, and approval for meaningful external action.

The project carries patterns from personal agents into an enterprise setting, but it fails closed when authority or policy is unclear. That constraint is the point, not a feature to be relaxed later.