As autonomous artificial intelligence (AI) agents increasingly interact across decentralized networks, traditional communication protocols (e.g., HTTP, MCP) fail to provide intrinsic execution control, semantic validation, and liability boundaries. This paper introduces the Catalyst Intelligence Operating System (CIOS) and its core communication layer, the Onto-Protocol. Unlike transport-focused standards, the Onto-Protocol is a semantic contract framework designed to define agent intent, strictly enforce execution boundaries, and generate cryptographically verifiable audit trails. Furthermore, we present the CIOS Probabilistic Trust Model—a dynamic scoring engine that calculates operational risk and agent trust in real-time. By separating semantic meaning from execution and infrastructure, this framework establishes a foundational layer for sovereign, legally compliant, and monetizable machine logic.