ZCode Dev Stack
BeginnerZ.ai's desktop coding agent with Goal Mode, running GLM, Kimi, Claude, or another provider's model.
Published 30 September 2026
About ZCode Dev Stack
ZCode is Z.ai's desktop coding agent, a free app for macOS, Windows, and Linux (Linux in beta) whose source is open under Apache-2.0. Its agent is tuned for GLM, Z.ai's own model family, and this stack defaults to GLM-5.3 on the GLM Coding Plan. The same app connects to Moonshot's Kimi, MiniMax, Xiaomi MiMo, Claude, and OpenAI's GPT models through their API keys, and to OpenRouter or any compatible endpoint, so GLM is the starting point rather than a lock.
The agent reads and edits the project, runs commands, drives a browser, and hands parts of a task to subagents. Its signature feature is Goal Mode: given a one-sentence objective, it keeps working in rounds and checks after each one for evidence that the goal is met, such as passing tests or the expected output. Project memory, skills, MCP servers, hooks, and plugins extend it, and scheduled automations run recurring tasks on the local machine. ZCode is desktop-first: the open-source repository can build a terminal interface and a local web mode, but there is no official packaged CLI.
ZCode runs on your own computer, which has to stay on and online while it works. Remote Control opens the current workspace in a phone browser from a QR code or link, one phone at a time, and Bot Channel does the same from WeChat or Feishu for longer-lived access. Both steer the desktop, which does the work, and ZCode has no hosted cloud or mobile app. Its remote development mode connects the desktop to a Linux server over SSH, to WSL, or to a running container, so the agent, commands, and git run there while models, login, and settings stay on the desktop. That session runs over the desktop's connection, so a server gives the agent a stronger or always-available machine, not a way to keep working with the laptop off.
Code lives on GitHub by default, or on GitLab. A server with private access, Docker, CI/CD, and an AI code reviewer are common additions.
Key Features
- ✓Desktop app for macOS, Windows, and Linux (Linux in beta), with source open under Apache-2.0
- ✓Goal Mode: the agent works in rounds until the evidence shows the objective is met
- ✓Subagents, browser automation, project memory, skills, MCP servers, hooks, and plugins
- ✓GLM-5.3 by default, plus Kimi, MiniMax, Xiaomi MiMo, Claude, GPT, OpenRouter, and compatible endpoints
- ✓Remote development over SSH, in WSL, or inside a running Docker container
- ✓Remote Control from a phone browser and Bot Channel from WeChat or Feishu
- ✓Scheduled automations and idle-time tasks on the local machine
When to Use ZCode Dev Stack
- →Running long refactors or fix-until-the-tests-pass tasks with Goal Mode
- →Getting the most from a GLM Coding Plan in the agent tuned for GLM
- →Working on a remote Linux box, WSL, or a dev container from a desktop app instead of a terminal
- →Checking on a running task from a phone browser while away from the desk
- →Teams that already coordinate in WeChat or Feishu and want to steer the agent from there
Pros
- The app is free and open source, with no separate subscription for the agent
- Goal Mode's evidence check suits long tasks better than a single-reply agent
- One GLM Coding Plan covers ZCode and other agents such as Claude Code, Cline, and OpenCode
- Remote development over SSH, WSL, or Docker is built into the desktop app
- Phone control opens in a browser, with no app to install
Cons
- No hosted cloud: phone control, bots, and automations all need the desktop running and online
- Remote development runs over the desktop's connection, so a server doesn't keep the agent working with your laptop off
- No official packaged CLI; the terminal and web modes have to be built from source
- Chat-app control covers WeChat and Feishu, which suits teams in China more than elsewhere
- No free usage: every task needs a Coding Plan or a provider API key
LLM Options for ZCode Dev Stack
ZCode Dev Stack powered by GLM
GLM is the default here and the model ZCode's agent is tuned for. The GLM Coding Plan runs from Lite at $18 a month ($12.60 billed yearly) to Max at $168, metered in five-hour and weekly credits, and works in other agents too. Pay-per-token GLM-5.3 costs $1.40 in and $4.40 out per million.
ZCode Dev Stack powered by Kimi
Moonshot is a built-in provider in ZCode, so Kimi connects with an API key. K3 handles the hardest agentic tasks at $3 in and $15 out per million tokens, and K2.7-Code covers everyday work at $0.95 and $4, a good second model when a GLM plan's credits run out.
ZCode Dev Stack powered by MiniMax
MiniMax is another built-in provider: MiniMax-M3 pairs a 1M-token context and multimodal input with $0.30 in and $1.20 out per million tokens, and its Token Plan from $22 a month is a flat-rate alternative that works across several coding agents.
ZCode Dev Stack powered by Xiaomi MiMo
Xiaomi MiMo is built into ZCode's provider list, and its MIT-licensed models do well on agentic coding for very little: $0.435 in and $0.87 out per million tokens for Pro, $0.14 and $0.28 for Flash, the cheapest way to run long Goal Mode loops.
ZCode Dev Stack powered by Claude
Claude is the pick for the hardest multi-file changes, connected through an Anthropic API key and billed per token: Sonnet 5 at $2 in and $10 out per million, Opus 5.5 at $4 and $20. Claude Pro or Max plans don't cover use in ZCode, and the agent is tuned for GLM first.
ZCode Dev Stack powered by OpenAI
OpenAI's GPT models connect through an API key billed per token: GPT-5.3-Codex, built for agentic coding, at $1.75 in and $14 out per million, or GPT-6 Sol at $2 and $10. A ChatGPT subscription doesn't cover use in ZCode.
These are highlighted picks. To see all the tools, check the LLM category.
Version Control Options for ZCode Dev Stack
ZCode Dev Stack with GitHub
GitHub is the default here: the largest pull-request ecosystem and the widest choice of review and CI integrations. ZCode's git panel works with any remote, so GitHub's advantage is everything around the repository rather than inside the app.
ZCode Dev Stack with GitLab
GitLab bundles source control, merge requests, and CI/CD in one self-hostable platform, the pick when code has to stay on your own infrastructure. ZCode only needs git, so everyday work in the app is the same on either host.
ZCode Dev Stack Add-ons
Each addition below extends this stack with a capability the base stack works fine without. None are required: include the ones your product actually needs when building this stack, and skip the rest.
Server Add-ons
Add a server when you want the agent to keep running somewhere other than your own laptop, reachable at any time and not tied to your machine staying on. Some agents also offer their own managed cloud sessions as an alternative to self-hosting; check the stack's own description for details.
ZCode Dev Stack with Hetzner
Hetzner is the default here: its smallest shared plan, CX23, has 4 GB of RAM for about €5.49/mo. ZCode documents no memory minimum and the models run at the provider, so 4 GB leaves room for its remote runtime, builds, and tests. The desktop still has to stay connected while the agent works there.
ZCode Dev Stack with DigitalOcean
DigitalOcean's Basic Droplet with 4 GB of RAM and 2 vCPUs costs $24/mo, billed per second up to that cap, with snapshots to roll back after a risky Goal Mode run. Its 1-Click agent images and Managed Agents cover other agents, so for ZCode it's a plain Linux box.
ZCode Dev Stack with Hostinger
Hostinger's KVM 1 has 4 GB of RAM at $6.49/mo introductory, renewing at $11.99/mo, enough for ZCode's remote runtime with the models served elsewhere. A budget pick on a plain Linux image; Hostinger's agent templates target other tools, not ZCode.
ZCode Dev Stack with Raspberry Pi
A Raspberry Pi 5 is a one-time hardware cost instead of a monthly bill, and ZCode's remote runtime ships for 64-bit ARM Linux. A 4 GB board is enough with the models served remotely, and 8 GB leaves room for builds. It suits a home dev box the desktop connects to over SSH.
These are highlighted picks. To see all the tools, check the Hosting & Cloud category.
Remote Access Add-ons
Add remote access when you want to reach an agent running on another machine — a VPS or a home server — without exposing it to anyone but you.
ZCode Dev Stack with Tailscale
Add Tailscale alongside a server so ZCode's desktop reaches it over SSH from anywhere, on a private address, without opening the SSH port to the internet. It's also a simple way to reach a home Raspberry Pi from outside the house. The personal tier is free.
Code Review Add-ons
Add code review when you want an AI reading every pull request before it merges: it comments on the diff so bugs, security issues, and inconsistencies surface before a human has to catch them.
ZCode Dev Stack with CodeRabbit
Add CodeRabbit when you want an AI reviewer on every pull request ZCode's changes end up in: it installs as a GitHub or GitLab app, summarizes the diff, and leaves line-level comments before anyone merges, with a free tier for open-source repos.
ZCode Dev Stack with Greptile
Add Greptile when review should read the whole repository rather than just the diff: it indexes the codebase so its GitHub and GitLab comments carry wider context, useful after a long Goal Mode run touches many files.
ZCode Dev Stack with Cursor Bugbot
Add Cursor Bugbot when a bug-focused reviewer is enough: it reviews GitHub or GitLab pull requests whatever wrote the code, though it bills through a Cursor account and plan rather than its own subscription.
CI/CD Add-ons
Add CI/CD when you want a dedicated pipeline for running tests, linting, or multi-stage builds before a deploy goes out. Many hosting platforms already redeploy automatically on every push on their own — a CI/CD tool adds the most value on top of that by gating the deploy on a passing test suite, and matters even more when the hosting choice does not auto-deploy at all, such as a self-hosted server.
ZCode Dev Stack with GitHub Actions
Add GitHub Actions when the repo lives on GitHub and every push should run the tests. ZCode has no CI mode of its own, so this is where its changes get checked independently of the machine that made them.
ZCode Dev Stack with GitLab CI/CD
Add GitLab CI/CD when the project is on GitLab and needs the same automated tests on every push, defined in a .gitlab-ci.yml file that can run entirely on your own runners.
These are highlighted picks. To see all the tools, check the CI/CD Pipelines category.
Containerization Add-ons
Add containerization when you want the app packaged the same way across local development, staging, and production, or need to deploy somewhere that isn't a managed serverless platform.
ZCode Dev Stack with Docker
Add Docker when the project's toolchain should live in a container: ZCode's remote development opens a running local container as its workspace, so the agent's commands and file changes happen inside it instead of on your machine.
These are highlighted picks. To see all the tools, check the Containerization category.
Frequently Asked Questions about ZCode Dev Stack
Can ZCode keep working when my computer is off?
No. Every way of reaching ZCode away from the desk steers the desktop app: Remote Control in a phone browser, Bot Channel in WeChat or Feishu, and scheduled automations all stop once the computer sleeps or the app closes, though a keep-awake switch covers overnight automations. Remote development moves the agent's work onto a server, but it runs over the desktop's SSH connection, so the laptop still has to stay on and connected. For work that continues with your machine off, agents with a hosted cloud, such as Claude Code or Codex, are the better fit.
Does ZCode work as well with Claude or GPT as with GLM?
It works, but GLM gets the most attention: the agent is tuned for GLM-5.3, and Z.ai's documentation and defaults assume it. Claude, GPT, Kimi, MiniMax, and MiMo connect through API keys and bill per token, since subscriptions like Claude Pro or ChatGPT can't be used here. Subagents inherit the main session's model unless you set another, so a cheap model can take the side tasks while a stronger one leads. Environment variables from your terminal don't carry into ZCode, so provider settings live in the app itself.
How does ZCode compare to Claude Code?
Both run one vendor's agent built around that vendor's models, but they differ in where you work and what runs without you. Claude Code is terminal-first, with a desktop app, IDE extensions, hosted cloud sessions, and an official phone app, and starts with a Claude Pro plan. ZCode is a desktop app with Goal Mode's evidence loop and built-in SSH, WSL, and Docker workspaces, and has no hosted cloud. The GLM Coding Plan also works inside Claude Code, so some developers use both on one subscription.
Which GLM Coding Plan tier do I need?
Lite covers steady daily use of one agent; Pro and Max raise the five-hour and weekly credit limits roughly six- and fourteen-fold for heavy or parallel work. Goal Mode loops and subagents spend credits faster than single-reply chats, so a long unattended objective can use up Lite's five-hour window. Credits go further outside weekday peak hours, and yearly billing cuts the price by 30 percent. The plan covers other agents too, so one subscription can serve ZCode and a terminal agent. The Pricing section has current figures.
What doesn't carry over to a remote workspace?
Models, login, and settings stay on the desktop, but everything the agent runs has to exist on the remote side: the project's toolchain, Git, and whatever a synced skill or MCP server depends on. Native Windows hosts aren't supported, SSH connections that rely on ProxyJump through an alias don't work, and connecting to WSL as root can leave files in the workspace owned by root. Automations can't target a remote workspace, and the phone can reconnect a remote workspace but not create one.
Scores
Popularity1/5
A recent launch with a small following so far compared with established coding agents.
Learning Curve2/5
A desktop app that signs in to a Coding Plan and starts working, with Goal Mode as the one new idea; remote workspaces and other providers add a little setup.
Flexibility3/5
Several model providers, skills, MCP, hooks, plugins, and SSH, WSL, or Docker workspaces, but everything runs through the desktop app, with no packaged CLI or hosted cloud.
Performance3/5
Goal Mode handles long tasks well on GLM-5.3, but results depend on the model, the product is young, and nothing runs once the desktop is off.
Portability3/5
The app is open source and models can come from several providers, but it's tuned for GLM, remote work depends on the desktop, and chat control is tied to WeChat and Feishu.
Tools in the ZCode Dev Stack Stack
ZCode Dev Stack Pricing
ZCode itself is free and open source, but it has no free usage, so the bill is the models. The GLM Coding Plan runs from $18/month for Lite to $168/month for Max, with 30% off billed yearly; other providers bill per token through an API key. GitHub or GitLab are free for individuals, Tailscale's personal tier is free, and an optional server runs about €5.49–24/month.
Apache-2.0 desktop app for macOS, Windows, and Linux; every task needs a Coding Plan or a provider API key.
Lite $18, Pro $80, Max $168 per month ($12.60, $56, $117.60 billed yearly), metered in five-hour and weekly credits; also works in Claude Code, Cline, and OpenCode.
API keys from Moonshot, MiniMax, Xiaomi MiMo, Anthropic, OpenAI, or OpenRouter, billed by each provider (MiMo Flash from $0.14/M input, Claude Sonnet 5 $2/M).
Free tier covers individual devs; GitHub Team is $4/mo, GitLab Premium is $29/mo for more seats.
Tailscale's personal tier is free for up to 100 devices.
4 GB leaves room for builds next to the agent: about €5.49 at Hetzner, $6.49 intro at Hostinger ($11.99 renewal), $24 at DigitalOcean; or a Raspberry Pi as a one-time cost.