CoworkOS Proposes a Harness for Long-Horizon Agent Work
Jiajie Jin and 39 other authors submitted the CoworkOS paper to arXiv on November 2, 2026. The authors describe it as a general-purpose harness that separates model-driven task strategy from runtime execution management, with synchronous or asynchronous delegation to subagents, hierarchical workspaces, persistent semantic memory, and support for human and agent revisions of shared documents, presentations, CAD models, and Blender scenes. They report that CoworkOS was highly competitive with Codex and DeepSeek Harness across 12 benchmarks and six categories of real-world applications.
The proposal addresses long-horizon productivity work by combining human-in-the-loop collaboration, shared evolving artifacts, delegation, persistent semantic memory, and resumable work after interruption.
Computer Science > Artificial Intelligence
arXiv:2611.03178 (cs)
[Submitted on 2 Nov 2026]
Title:CoworkOS:Scaling Agent Coworkers for Real-world Automation
Authors:Jiajie Jin, Guanting Dong, Chengjun Pan, Xin Deng, Zhuohui Sheng, Guoqiang Wang, Tao Gui, Yutao Zhu, Yujia Chen, Shihong Yi, Binze Hu, Zhixiong Yang, Zhichao Sun, Rui Wang, Han Han, Mingyu Ma, Junhao Huang, Xiaotao Wang, Huicheng Jiang, Yuyang Hu, Xinyu Yang, Wanghan Xu, Yajie Yang, Zhecan Wang, Honglin Guo, Xiaoyu Xing, Xiaoqian Zhang, Yulun Zhu, Xiaoxi Li, Jiahang Lin, Shaofan Liu, Shihan Dou, Zhiheng Xi, Qi Zhang, Yu-Gang Jiang, See-Kiong Ng, Yang Zhang, Tat-Seng Chua, Ji-Rong Wen, Zhicheng Dou
View a PDF of the paper titled CoworkOS:Scaling Agent Coworkers for Real-world Automation, by Jiajie Jin and 39 other authors