~ / agents / zcode

Zcode

Z.ai's desktop agentic development environment, the official harness for GLM-5.2 / 5.3, with a Goal mode that runs long autonomous tasks to a verifiable end.

VendorZ.ai
FormDesktop ADE (macOS / Windows; Linux beta)
LaunchedZCode 3.0, June 2026 (v3.8.1, Aug 2026)
ARscorenot scored
Pricing tierClient free; GLM Coding Plan $18 / $72 / $160
Coding modelsGLM-5.2 / GLM-5.3 (deep) + Claude / GPT / Kimi K2.5 / DeepSeek / OpenRouter
Open sourceNo

Official links

Outbound, for trial and verification.

Official sitehttps://zcode.z.ai/
Pricinghttps://z.ai/subscribe
English interface / changeloghttps://zcode.z.ai/en/changelog

Capabilities

What it ships, in its own terms.

ZCode Agent is tuned for GLM-5.3 long-horizon work; Goal Mode (/goal) plans, executes and verifies; multi-agent runs agents in parallel; remote control drives sessions from Telegram, WeChat or Feishu; a built-in browser and PowerPoint support cover office tasks; AGENTS.md memory carries project rules across tasks.

Use cases

Scenario coverage (presence, not benchmark performance).

ScenarioLocal filesDocs / PPTExcel reportsAnalysisWeb searchMini-programsWeb building
Zcode??

✓ supported · ◐ partial or via a sibling product · ? unclear · — not in scope. Presence, not benchmark performance.

Users

Subscriber and active-user counts, source-labelled.

not published (new users get a 5-day trial at 5M tokens/day)Source: —

Third-party review

Independent coverage, not vendor marketing.

Third-party editorial: no /10 score; bitdoze (Jul 2026) qualitative review called v3.2 'finally stable'

Coding ability (benchmark): not scored — no DeepSWE or Terminal-Bench measurement.

An independent bitdoze review (Jul 2026) called v3.2 'finally stable enough for real work'; the GLM Coding Plan's 5-hour quota windows and peak multipliers drew Hacker News complaints. No DeepSWE or Terminal-Bench score.

Source: bitdoze / Community (Hacker News)

Compare further

The reference agents and the scored Chinese entry.

China big-tech hub

all entries vs Codex and Claude Code

comparison

Claude Code

the architecture leader

score 86

ChatGPT Codex

the reference OpenAI terminal agent

score 84

Qwen CodeOS

the scored open-source Alibaba CLI

score 45