ANX Protocol: An Open Agent-Native Framework for AI Agent Interaction Replaces GUI Automation and MCP Skills

Researchers have proposed ANX, an open, extensible, agent-native protocol and top-level framework that integrates CLI, Skill, and MCP into a unified architecture for AI agent interaction.

The Problem

Current AI agent interaction methods have significant flaws:

• GUI automation — High token consumption, brittle
• MCP-based skills — Fragmented interaction, inadequate security
• No unified framework — Each module has independent defects

ANX's Four Core Innovations

1. Agent-native design (ANX Config, Markup, CLI) — High information density, reduces tokens, eliminates inconsistencies
2. Human-agent interaction — Dual rendering as agent-executable instructions AND human-readable UI
3. MCP-supported lightweight apps — On-demand, no pre-registration required
4. ANX Markup for SOPs — Machine-executable Standard Operating Procedures eliminating ambiguity

The 3EX Architecture

ANX uses a "3EX Decoupled Architecture" separating:

• Expression — How agents communicate
• Execution — How tasks are performed
• Experience — How interactions are experienced

This decoupling allows independent evolution of each layer.

Why It Matters

• Token efficiency — Reduces the massive token overhead of GUI-based agent interaction
• Multi-agent collaboration — ANX Markup enables reliable long-horizon tasks and agent-to-agent coordination
• Open standard — Extensible protocol anyone can implement
• Security — Built-in verification compared to GUI automation

Context

As AI agents become more autonomous, the interaction layer between agents and tools becomes critical. ANX proposes a protocol-first approach rather than retrofitting human GUIs for machine use.