Six agent protocols just launched. Three of them decide which products survive. Here is how to tell which three.

2026-06-07 · Source: Nate’s Substack · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

Six new agent protocols have launched in the past year, creating a complex landscape for builders and buyers of AI agents. While many expect large platforms to absorb these, the article argues that understanding their layered structure is crucial. Three protocols—MCP (for tools and data access), A2A (for inter-agent delegation), and AG-UI (for human-in-the-loop controls)—are identified as forming the core agent stack. These address fundamental questions about what an agent can use, who it can work with, and how humans maintain control. Other protocols like A2UI, AP2, and x402 exist but operate in higher layers where product requirements and payment rails are still evolving. Misunderstanding this layer map leads to agents failing at critical boundaries like security and supervision, or to ineffective vendor evaluations. The article provides a protocol map and guidance for strategy, auditing, and briefing.

Key takeaway

For AI Architects evaluating agent strategies or teams building new agent systems, you must prioritize understanding the core protocol stack: MCP, A2A, and AG-UI. Ignoring this layered map risks shipping agents that fail at critical boundaries like security, human approval, or cross-company supervision. Use the layer map to accurately evaluate vendor claims and design robust, controllable agent workflows, ensuring your systems meet essential operational requirements.

Key insights

Understanding the layered structure of agent protocols, especially MCP, A2A, and AG-UI, is crucial for effective agent development and evaluation.

Principles

• Agent substrate impacts customer experience more than model choice.
• Not all agent protocols solve the same problem.
• Avoid treating all agent protocols as equal bets.

Method

Map workflows to protocol layers, draft Agent Card boundaries for integration, audit for missing human controls, and produce strategy briefs for platform decisions.

In practice

• Evaluate vendor agent offerings using the layer map.
• Design agents with robust security for tool access.
• Implement human approval for long-running agent tasks.

Topics

• Agent Protocols
• AI Agents
• MCP Protocol
• A2A Protocol
• AG-UI Protocol
• Protocol Layering
• Human-in-the-Loop

Best for: CTO, VP of Engineering/Data, AI Product Manager, AI Engineer, AI Architect, Director of AI/ML

Related on AIssential

• Article: Architecting Agentic MLOps: A Layered Protocol Strategy with A2A and MCP — Layering A2A for communication and MCP for capabilities enables robust, extensible multi-agent MLOps workflows.
• The Agent Stack - Part 1: A Systems Map of Modern Agent Infrastructure — Deconstructing the overloaded "agent" concept into a defined stack clarifies system responsibilities and failure points.
• The Agent Client Protocol Overview — ACP standardizes AI agent-client communication, enabling seamless integration across diverse coding assistants and IDEs.
• EP202: MCP vs RAG vs AI Agents — AI development is shifting from LLM application to fundamental research, advanced prompt engineering, and AI-driven code generation.
• Use A2A to connect agents across different frameworks and teams — A2A standardizes agent communication, enabling collaboration and reuse across diverse frameworks and teams.
• How Pinterest Built a Production MCP Ecosystem — Standardized protocols and robust platform infrastructure are crucial for scaling AI agent integration with internal systems.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Nate’s Substack.