Six layers your agent has to handle. Most products have only thought about two. + a responsibility-layer audit.

Source: Nate’s Newsletter Date: 2026-05-12 URL: https://natesnewsletter.substack.com/p/agentic-commerce-protocol-war

Summary

This is the companion framing piece to the agentic commerce protocol war article (same URL, same post). Where the “one human click” signal focuses on the decomposition of the checkout action, this entry foregrounds the auditing lens: most products building agent-facing flows have only addressed two of the six responsibility layers (typically identity and payment credentials), leaving fraud detection, authorization scope, settlement liability, and dispute resolution as gaps that will surface when agents act autonomously at scale. The three protocols forming the emerging core stack are MCP, A2A, and AG-UI.

Implications

• Feeds the agent authorization / trust thread: The six-layer frame is a practical audit tool for any team building agentic workflows — gaps in the untouched four layers are where liability and security incidents will concentrate.
• MCP/A2A/AG-UI as necessary but not sufficient: The piece implicitly argues that adopting the protocol stack handles tool access and delegation, but not the commercial responsibility layers — those require separate design work that most teams haven’t started.
• Useful reference for evaluating agent product maturity: When assessing agentic products or frameworks, checking which responsibility layers they explicitly address (vs. deferring to the human) is a quick signal of production-readiness.