Related: This article is the product companion to 智谱 GLM-5.2 发布：1M 上下文旗舰模型，Code Arena 开源第一 — ZCode 3.0 is the desktop IDE where GLM-5.2’s long-context coding abilities are put to work in real development workflows.
ZCode is a full-featured Agentic Development Environment (ADE) built for Long Horizon Tasks. It helps AI Agents complete long-running, multi-step development work end-to-end with stability and control — from planning through to verification.
ZCode 3.0 is deeply tuned around GLM-5.2. The model, tooling, and execution workflow are tightly integrated, enabling complex development tasks to move from planning all the way to verification within a single continuous context.
With natural-language instructions, you can ask ZCode Agent to handle coding, debugging, testing, project preview, and change review. Through the desktop workspace, Remote, and Bot Channel, you can keep checking progress and adding instructions while long-running work continues.
Core Capabilities
ZCode organizes the development experience around the self-developed ZCode Agent, with emphasis on:
- Long-task execution — keep requirement understanding, planning, code changes, verification, and review inside the same task without context-switching overhead.
- Stable context retention — combine files, terminal output, browser context, execution modes, and Git state to maintain continuity across hundreds of thousands of tokens.
- Self-developed Agent workflow — tasks, permissions, context, tool calls, and review flow are all organized around ZCode Agent; model output maps directly to actual engineering operations.
- Continuous follow-up across devices — desktop, mobile Remote, and Feishu / WeChat Bot can all push the same workspace task forward from wherever you are.
- Safety confirmations — critical commands, file modifications, and high-privilege operations require user confirmation before execution.
Version 3.0 Highlights
Self-Developed ZCode Agent
Purpose-built for Long Horizon Tasks. Stability and context retention in long-chain tasks have improved dramatically over previous generations.
Unified Agent Workflow
Everything — tasks, context, permissions, file references, and review — flows through ZCode Agent. Long tasks stay continuous rather than fragmenting across sessions.
Feishu / WeChat Bot is Live
Bot integration has landed on Feishu and WeChat. Mention the bot to trigger ZCode Agent remotely and advance workspace tasks from your mobile chat.
Remote — Mobile Control
Phone-scannable Remote connects to the current ZCode workspace, mirroring the desktop experience for true Vibeworking — check progress and inject instructions from anywhere.
GLM-5.2 and Exclusive Benefits
GLM-5.2 is fully available in ZCode, supporting genuinely usable long-context capability — ideal for requirements that span more files, more steps, and longer sustained reasoning.
Users who have activated the GLM Coding Plan receive 150% of the standard usage quota within ZCode, so the same plan goes further in Agent development sessions. New users can start with the free Start Plan to try GLM-5.2 and GLM-5-Turbo directly.
Quick Start
- Download ZCode — macOS (Apple Silicon / Intel), Windows, and Linux beta are available.
- Configure your API Key — connect BigModel or Z.ai account to activate GLM models inside ZCode.
- Start a Goal — use
/goalto set a verifiable target and let ZCode Agent iterate autonomously until confirmed complete.
Downloads
| Platform | Package |
|---|---|
| macOS (Apple Silicon) | ZCode-3.1.2-mac-arm64.dmg |
| macOS (Intel) | ZCode-3.1.2-mac-x64.dmg |
| Windows | ZCode-3.1.2-win-x64.exe |
| Linux | Join the beta via the Feishu feedback group |
Full documentation is at zcode.z.ai.
Tags: ZCode, ZCode 3.0, Agentic Development Environment, ADE, AI Agent, Long Horizon Task, GLM-5.2, GLM, 智谱, BigModel, Z.ai, Coding Agent, Claude, Claude Opus, AI Coding, AI IDE, Developer Tools, Software Engineering, Automatic Programming, Code Generation, Code Review, Debugging, Testing, Long Context, 1M Context, Multi-Agent, Sub-Agent, Task Management, Remote Control, Bot Channel, Feishu Bot, WeChat Bot, Mobile Development, Vibeworking, Function Call, MCP, Plugin, Workflow Automation, Code Arena, SWE-Bench, FrontierSWE, Open Source Model, SOTA, Chinese LLM, Domestic LLM,国产大模型, 人工智能, 机器学习, 深度学习, 自然语言处理, 生成式AI, AI智能体, 开发环境, 软件开发, 自动调试, 持续集成
Free Start Plan for New Users
New users can claim a free Start Plan immediately after registration — no credit card required:
- GLM-5.2: 3,000,000 tokens free (expires June 22, 2026)
- GLM-5-Turbo: 2,000,000 tokens free (expires June 22, 2026)
Sign up via the invite link → get 20 million bonus tokens on top
