ANX: Protocol-First Design for AI Agent Interaction with a Supporting 3EX Decoupled Architecture
Summary: arXiv:2604.04820v1 Announce Type: new
Abstract: AI agents, autonomous digital actors, need agent-native protocols; existing methods include GUI automation and MCP-based skills, with defects of high token consumption, fragmented interaction, inadequate security, due to lacking a unified top-level framework and key components, each independent module flawed. To address these issues, we present ANX, an open, extensible, verifiable agent-native protocol and top-level framework integrating CLI, Skill, MCP, resolving pain points via protocol innovation, architectural optimization and tool supplementation.
Core Innovations of ANX
ANX introduces four core innovations to enhance interaction and performance:
- Agent-native design: ANX Config, Markup, and CLI are designed for high information density, flexibility, and strong adaptability. This approach reduces token consumption and eliminates inconsistencies.
- Human-agent interaction: The combination of Skill’s flexibility allows for dual rendering as both agent-executable instructions and a human-readable user interface.
- MCP-supported on-demand lightweight apps: ANX enables the creation of apps without the need for pre-registration, streamlining the process for users.
- ANX Markup-enabled machine-executable SOPs: This feature eliminates ambiguity in instructions, ensuring reliability for long-horizon tasks and multi-agent collaboration.
3EX Decoupled Architecture
As the first in a series, this paper focuses on ANX’s design and presents its 3EX decoupled architecture, which includes ANXHub. This architecture is pivotal for ensuring seamless interaction between AI agents and their human counterparts. The design prioritizes native security by implementing measures such as:
- LLM-bypassed UI-to-Core communication, which helps keep sensitive data out of the agent context.
- Human-only confirmation processes to prevent automated misuse of the system.
Performance Analysis
Preliminary feasibility analysis and experimental validation have demonstrated the effectiveness of the ANX protocol. Key findings from form-filling experiments using Qwen3.5-plus and GPT-4o include:
- ANX reduces tokens by 47.3% (Qwen3.5-plus) and 55.6% (GPT-4o) compared to MCP-based skills.
- Token consumption is reduced by 57.1% (Qwen3.5-plus) and 66.3% (GPT-4o) when compared to GUI automation.
- Execution time is shortened by 58.1% (Qwen3.5-plus) and 57.7% (GPT-4o) relative to MCP-based skills.
Conclusion
The introduction of the ANX protocol marks a significant advancement in the development of AI agent interactions. With its innovative design and robust architecture, ANX addresses critical issues faced by existing methods, paving the way for more efficient, secure, and user-friendly AI applications. Researchers and developers are encouraged to explore the potential of ANX in their future projects.