ZCode + GLM 5.3: The Complete Guide to Z.AI's Coding Agent

Quick answer: ZCode is Z.AI's coding agent and the best place to experience GLM 5.3's long-horizon strength. It features Goal mode (plans, codes, tests, verifies until the target is met), a 98%+ cache hit rate (~30% more effective tokens), a 1.5× limited-time quota boost through August 31, and Remote Control to steer long tasks from WeChat or Feishu.

TL;DR

Feature	What it does
Goal mode	Runs until the goal is met — the long-horizon RL payoff
Cache hits	98%+ — repeated context bills at cached rates
Quota boost	1.5× through Aug 31 (stacks to ~180%)
Remote Control	Monitor/steer from WeChat or Feishu
Models	GLM 5.3 (all users upgraded), 5.2, 5-Turbo
Access	Included with GLM Coding Plan

What Is ZCode?

ZCode (zcode.z.ai) is Z.AI's own coding agent — the native home of GLM 5.3. Where other agents treat each prompt as a discrete unit, ZCode is built around long-horizon execution: it plans, writes code, runs tests, and verifies until the stated goal is actually met.

That's not a marketing line — it's the direct productization of GLM 5.3's training. The model was post-trained on environments that "approximate several days of work for an experienced engineer"; ZCode is the interface for that capability.

Feature Deep Dive

Goal Mode

Instead of answering and stopping, ZCode's Goal mode plans → codes → tests → verifies in a loop until the target is met. This is where GLM 5.3's AutomationBench strength (48.2 — best in Z.AI's official table) shows up in daily work: multi-step tasks that previously needed babysitting now run to completion.

98%+ Cache Hit Rate

ZCode aggressively caches repeated context (system prompts, repo structure, conversation history). Cached tokens bill at the lower cached-input rate, giving roughly 30% more effective tokens per quota. For long agent sessions — where the same context prefix recurs constantly — this compounds fast.

1.5× Quota Boost (Through August 31)

Limited-time: ZCode gives a 1.5× quota boost that stacks with cache savings, for up to 180% of your standard quota. If you're evaluating GLM 5.3, the cheap stress-test window closes at the end of August.

Remote Control

Monitor and steer long-running tasks from your phone via WeChat or Feishu — start a goal-mode run on desktop, check progress and intervene from mobile.

Setup

Subscribe to the GLM Coding Plan (z.ai/subscribe) — Lite $18/mo, Pro $80/mo, Max $168/mo.
Go to zcode.z.ai and log in with your Z.AI account.
Start a goal: describe the outcome ("refactor the auth module, keep all tests passing, update the docs") and let Goal mode run.
Bind Remote Control in settings if you want mobile steering via WeChat/Feishu.

All Coding Plan users are already on GLM 5.3 — no model selection needed, though you can switch between GLM 5.3, 5.2, and 5-Turbo as needed.

Cost Notes

Points-based quota: input, cached input, and output metered separately.
Off-peak = 50% points — outside 14:00–18:00 UTC+8 weekdays (all weekend included).
Stack the boost: 1.5× quota + 98% cache hits = up to 180% effective quota through Aug 31.

ZCode vs Other Agents

	ZCode	Claude Code / Cline / OpenCode
GLM 5.3 access	Native, auto-upgraded	Via devpack (`npx @z_ai/coding-helper`)
Goal mode	Built-in	Manual loops
Cache optimization	98%+ hit rate	Varies
Mobile steering	WeChat/Feishu Remote Control	Not built-in
Quota boost	1.5× through Aug 31	N/A

For heavy GLM 5.3 users, ZCode is the most feature-complete surface. For teams standardized on Claude Code or OpenCode, the devpack route works fine — you just miss Goal mode and the cache/boost optimizations.

FAQ

What is ZCode? Z.AI's coding agent, built around GLM 5.3's long-horizon capability — with Goal mode, 98%+ cache hits, and a 1.5× quota boost through Aug 31.

Is ZCode free with the Coding Plan? Yes — included. Plans start at $18/month (Lite).

How do I use GLM 5.3 in ZCode? Log in at zcode.z.ai — all Coding Plan users are already on GLM 5.3.

What is Goal mode? ZCode plans, codes, tests, and verifies until the stated goal is met — the productization of GLM 5.3's long-horizon RL training.

When does the 1.5× quota boost end? August 31, 2026.