AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant

1Xi'an Jiaotong University 2Shanghai Artificial Intellegence Laboratory 3The University of Hong Kong
*Equal Corresponding Author.
(Stay tuned for upcoming website upgrades and the open-sourcing of our code!)

Demo1: Can you make Bing the main search thingy when I look stuff up on the internet?

Demo2: Help me access my outlook account with address "anonym-x2024@outlook.com" and password 'gTCI";=@y7|QJ0nDa_kN3Sb&>'

Demo3: Please create an SSH user named "charles" with password "Ex@mpleP@55w0rd!" on Ubuntu who is only allowed to access the folder "/home/test1".

Demo4: This essay requires one paragraph each for introduction, body and conclusion with single-space for introduction, double-space for body then one-and-a-half-space for conclusion.

AgentStore is a flexible and scalable platform for dynamically integrating various heterogeneous agents to independently or collaboratively automate OS tasks. It allows users to quickly integrate their own specialized agents into the platform, similar to the functionality of the App store. This scalable integration allows the framework to dynamically adapt itself to the evolving OS, providing the multi-dimensional capabilities needed for open-ended tasks.

Abstract

Digital agents capable of automating complex computer tasks have attracted considerable attention due to their immense potential to enhance human-computer interaction. However, existing agent methods exhibit deficiencies in their generalization and specialization capabilities, especially in handling open-ended computer tasks in real-world environments. Inspired by the rich functionality of the App store, we present AgentStore, a scalable platform designed to dynamically integrate heterogeneous agents for automating computer tasks. AgentStore empowers users to integrate third-party agents, allowing the system to continuously enrich its capabilities and adapt to rapidly evolving operating systems. Additionally, we propose a novel core MetaAgent with the AgentToken strategy to efficiently manage diverse agents and utilize their specialized and generalist abilities for both domain-specific and system-wide tasks. Extensive experiments on three challenging benchmarks demonstrate that AgentStore surpasses the limitations of previous systems with narrow capabilities, particularly achieving a significant improvement from 11.21% to 23.85% on the OSWorld benchmark, more than doubling the previous results. Comprehensive quantitative and qualitative results further demonstrate AgentStore's ability to enhance agent systems in both generalization and specialization, underscoring its potential for developing the specialized generalist computer assistant.

Methods

AgentStore consists of three main components: AgentPool, AgentEnroll, and MetaAgent. The AgentPool stores all feature-specific agents with distinct functionalities. AgentEnroll defines the integration protocol for adding new agents to the AgentPool. Finally, the MetaAgent selects the most suitable agent(s) from AgentPool to independently or collaboratively complete tasks.

Results on the OS-world Benchmark

ap-clip

Detailed success rates of previous methods and AgentStore on OSWorld, divided by domains. Methods marked with “*” represent our re-implementation of the corresponding agents to ensure their applicability. Due to the significant overlap of operations between the OS and Workflow domains in the original division, we merged these two domains into “OS*”.

BibTeX

@article{jia2024agentstore,
      title={AgentStore: Scalable Integration of Heterogeneous Agents As Specialized Generalist Computer Assistant},
      author={Jia, Chengyou and Luo, Minnan and Dang, Zhuohang and Sun, Qiushi and Xu, Fangzhi and Hu, Junlin and Xie, Tianbao and Wu, Zhiyong},
      journal={arXiv preprint arXiv:2410.18603},
      year={2024}
    }