Zhipu upgrades ZCode with Goal mode, subagents and remote control

Zhipu has rolled out a broad upgrade to its AI coding product ZCode, adding Goal, Subagents, Remote Control and Off-Peak Tasks as it seeks to turn a coding agent from a conversational assistant into a system that can autonomously deliver complex work. The company also said ZCode's user base has surpassed 1 million and reset weekly limits for all GLM Coding Plan users at 1:00 p.m. on Aug. 11 to mark the milestone. Goal mode lets users define a verifiable target, after which the agent can break down tasks, edit code, run commands and review tests in repeated cycles until the objective is met, with progress and elapsed time shown in a dedicated panel. Zhipu said its Z.ai Code Bench tests showed GLM-5.2 paired with ZCode delivered an overall pass rate 2.39% higher than GLM-5.2 paired with Claude Code on real-world tasks including full-stack development, bug fixing and feature implementation, with the edge more visible in complex cross-file work and final verification. Subagents add parallel task execution through built-in General-purpose and Explore agents, while custom sub-agents can be configured by users. Remote Control allows a phone to access the desktop ZCode window through QR codes, browser links and apps including WeChat, Feishu and Lark, while code continues to run on the original machine across local, SSH, WSL and Docker workspaces. Off-Peak Tasks are designed for non-urgent long-running jobs and do not consume GLM Coding Plan credits. Zhipu also said ZCode's context cache hit rate on GLM exceeds 98%, lifting effective token capacity by about 30%; combined with a temporary 1.5x quota bonus running through Aug. 31, usable volume approaches 1.8 times the standard level. Market sources cited in the report said Zhipu's MaaS open platform has nearly 7 million registered users, including 23,000 enterprise customers, and annual recurring revenue has grown 15-fold this year, though the company denied investor claims that year-end ARR is expected to reach $2 billion.

The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.