CUAWright: A Minimal Unified Interface for Digital Agents

2026-10-02

AI Summary

This study addresses the limitations of existing GUI frameworks, whose static interfaces constrain the programming capabilities of large language models (LLMs) and hinder dynamic tool construction. We propose a minimalist terminal framework implemented in approximately 3K lines of code. By abandoning fixed GUI interfaces in favor of Bash as a unified interaction layer and the file system as an evolutionary space, this approach enables LLMs to autonomously write code and dynamically create and manage tools. Experimental evaluations on benchmarks such as OSWorld demonstrate that our framework significantly outperforms GUI-based baselines, achieving up to a 44% improvement in success rate while reducing costs by 37.5%. These results validate the superiority of a minimal programmable terminal interface in unlocking model potential and facilitating efficient environment interaction.

Abstract

The prevailing approach to computer-use agents couples a model with a domain-specific harness: a browser or desktop environment equipped with human engineered tools that are fixed before task execution. As models'coding capabilities improve, the GUI native and static harness prevents them from direct programmatic operation on system state, as well as flexible construction of tools. To this end, we introduce CUAWright, a minimal terminal harness of roughly 3K lines of code that uses bash commands as its sole action interface, and a file system as its evolvable space for dynamically creating tools and managing the context. We conduct comprehensive experiments across a wide range of digital tasks, and demonstrate that by giving the agent a minimal, programmable interface, it achieves substantially stronger results compared to their GUI or hybrid CLI interface across a wide range of tasks. On OSWorld 2.0, CUAWright delivers a 33.2% relative improvement in partial reward while reducing estimated cost by 37.5% compared with the published GPT-5.5 baseline. On Online-Mind2Web and the long horizon Odysseys benchmark, CUAWright substantially outperforms GUI native harness by 4.7% and 44.0% in success rate, respectively. Furthermore, we found the gains extend to CAD applications that require accurate visual understanding and CLI interaction: on CADGenBench and BenchCAD, our unified harness yields 8.1%-41.6% relative improvements over other CLI based harnesses with GPT-5.5. Together, these results suggest digital environments are far more programmable than their GUI interfaces imply, and a minimal terminal-focused harness is the key for better performance and efficiency.

Problem

• computer-use agents
• domain-specific harness
• GUI interface
• tool construction
• programmable interface

Innovation

• Digital Agents
• Terminal Harness
• Programmable Interface
• Dynamic Tool Creation
• Bash Commands