AGENTIS: Mapping MCP and A2A Protocols to the ETSI ENI Multi-Agent Interface Model for Autonomous Network Management

ETSI GR ENI 056, published in 2025, formally recognises multi-agent frameworks as the architectural foundation for 6G core network management. Yet the same document identifies a critical gap: the protocols agents use to communicate — Anthropic's Model Context Protocol (MCP) and Google's Agent-to-Agent protocol (A2A) — remain unstandardised within ETSI's normative framework. Meanwhile, hyperscalers and LLM vendors are shipping agent coordination runtimes that operate outside any standards governance. This paper resolves that gap. We present the first rigorous mapping between the ENI 056 five-interface agent model and two dominant real-world protocols — MCP and A2A — demonstrating both their equivalences and their structural divergences. We show that MCP and A2A are not competing alternatives but complementary layers: MCP governs vertical agent-to-resource interactions while A2A governs horizontal agent-to-agent coordination. We propose a four-archetype network agent taxonomy (Orchestrator, Specialist, Sentinel, Gateway), a protocol selection framework tied to ENI closed-loop decision types, and a standards contribution pathway for normative adoption of inter-agent communication under ETSI ENI governance.