The domestic large model application field has once again seen a fundamental iteration of productivity tools. On June 3rd, the AI startup Moonshot officially announced that its general-purpose local Agent product for knowledge workers, Kimi Work, has started internal testing. This feature will be released alongside the latest Mac and Windows desktop test versions of Kimi.

As an efficiency tool that claims to "transition from a conversation assistant to a work executor," Kimi Work is built on Moonshot's Kimi Code. Thanks to the evolution of the underlying capabilities, this application has powerful local Agent infrastructure. It not only allows AI to run scheduled tasks in the user's local computer environment, but also supports the installation and independent use of various operational skills (Skill).

image.png

In terms of professional skill reserves, the new product fully inherits the advantages of the online version of Kimi Agent, coming with high-frequency professional Skills such as website building and PPT creation right out of the box. In addition, the system deeply integrates professional-grade databases covering vertical industries such as finance, research, and law, and includes the Kimi WebBridge solution that supports browser invocation, greatly expanding the boundaries of local AI collaboration.

The most notable feature is its "team battle capability" for handling complex long-process tasks. Kimi Work has introduced support for Agent clusters for the first time. The system can autonomously create and schedule virtual expert teams consisting of up to 300 sub-Agents based on the complexity of the user's input task. Through parallel collaboration and division of labor among multiple agents, it helps users handle complex large projects that are more time-consuming and technically diverse across applications.

image.png

In terms of actual interaction experience, this update has completely broken through traditional development and operation barriers. Users do not need to understand any code, nor do they need to open terminals and write command lines, or configure the underlying environment in a complicated way. They just need to directly describe the final goal using ordinary natural language, and Kimi Work will automatically decompose the task, run the process in parallel, call tools across systems, and even create and organize folders on its own.

Ultimately, the system can deliver highly completed documents, spreadsheets, PPTs, and other types of core work products directly to users, fully covering various knowledge-intensive work scenarios such as office work, in-depth research, content creation, and daily data organization. With the implementation of local Agents, AI is truly evolving from a simple "chatting" generative dialogue box into a versatile digital employee capable of doing hard work in the computer operating system instead of humans.