AXIS: Efficient Human-Agent-Computer Interaction with API-First LLM-Based Agents

Junting Lu , Zhiyang Zhang , Fangkai Yang , Jue Zhang , Lu Wang , Chao Du , Qingwei Lin , Saravan Rajmohan , Dongmei Zhang , Qi Zhang

🏛 Institutions: PKU , NJU , Microsoft
📅 Date: September 26, 2024
📑 Publisher: ACL 2025
💻 Env: Desktop
🔑 Keywords: framework API-first AXIS HACI agent OS

TLDR

AXIS is an API-first agent framework that prioritizes application APIs over direct UI actions and also expands API coverage through automated exploration. On Microsoft Word tasks, it reduces completion time by 65-70% and cognitive workload by 38-53% while maintaining 97-98% accuracy relative to human performance, and it motivates a Human-Agent-Computer Interaction design framework for agent-centric software.

Open paper Report issue