AXI (Agent eXperience Interface) provides 10 design principles for building agent-native CLI tools that optimize for AI agent interaction. It addresses the high token costs and overhead associated with traditional CLIs and structured tool protocols (like MCP) by prioritizing token budget and agent ergonomics, leading to higher accuracy, lower costs, and faster execution times for AI agents interacting with external services.

How It Works

AXI tools are built around 10 core design principles aimed at minimizing token consumption and improving agent usability. Key aspects include a token-efficient TOON output format (saving ~40% tokens over JSON), minimal default schemas, content truncation with escape hatches, pre-computed aggregates to reduce round trips, and definitive empty states. These principles ensure agents receive concise, actionable information, reducing the need for multiple LLM calls and improving overall efficiency.

Quick Start & Requirements

• Install AXI Implementations:
  • Browser Automation: npm install -g chrome-devtools-axi
  • GitHub Operations: npm install -g gh-axi
• Build Your Own AXI: Install the AXI skill for guidelines and scaffolding: npx skills add kunchenguid/axi
• Prerequisites: Node.js and npm are required for installation and running the CLI tools.
• Documentation: Benchmark harnesses are located in bench-browser/ and bench-github/ directories.

Highlighted Details

• Browser Benchmark: chrome-devtools-axi achieved 1