As agentic systems move from experimentation into regulated, high-stakes, and multi-agent environments, the core problem is no longer merely orchestration. It is governability. This paper introduces the Action Claim as the correct pre-execution governance object: a contestable, machine-evaluable proposal that a specific world-state transition should be permitted, structured across declared, derived, and delegation-supplied fields. We argue that governing agentic systems requires not just a protocol for transmitting such objects, but an operational ontology — a typed, reusable semantic framework that defines what exists in the domain of agent action, how effects compose, and how governance systems can evaluate them before execution.