Zhipu Ships ZCode, a GLM-5.2 Coding Agent With Chat-App Triggers

Zhipu's coding-agent desktop app is out under the name ZCode, and the interesting bit is not that it wraps GLM-5.2 but where you can invoke it from. According to the product page, you can invoke ZCode remotely through WeChat, Feishu or Telegram to trigger tasks without touching the desktop interface. That is a different shape of coding agent than the desktop-tethered pattern most Western tools have settled into.

The rest of the pitch is familiar. ZCode is described as having deep optimization for GLM-5.2 model capabilities, ships with 20+ programming tools including Git and a terminal, and has a 'Goal' construct for managing multi-step objectives across planning, execution, and verification. Multi-agent collaboration is mentioned but not spelled out on the product page, so read that phrase as a marketing label rather than a documented mechanism.

Pricing sits in three yuan-denominated tiers. Lite is ¥16.2 per month (regular ¥18) with a basic quota. Pro is ¥64.8 per month (regular ¥72) with 5x the Lite quota, MCP tools, and faster generation. Max is ¥144 per month (regular ¥160) with 20x the quota and peak-resource priority. The scanner draft that circulated earlier denominated those numbers in dollars, which materially misstated the affordability story.

The honest caveats: the product page does not publish benchmark numbers, does not spell out how the multi-agent handoff actually works, and says nothing about guardrails for the case where a chat message triggers a task that then writes files or runs git operations on your machine. Kai wrote on X that ZCode is 'Similar concept to Codex app, built around a Chinese model' with bot channels where you can @ the agent from Telegram or Discord, which is a fair characterization of the framing.

What is worth watching is whether the chat-trigger pattern gets copied outside China. If a developer can push a coding goal forward from their phone while walking around, without opening a laptop, that changes when and how coding work gets done, which is a bigger shift than any given model's leaderboard placement.