Workspace-aware
Reference files, folders, past conversations, attachments, and project instructions without leaving the task.
ZCode is a desktop agentic development environment built around GLM-5.2. This guide explains how its Agent, workspace context, execution tools, and long-running task model fit into a real coding workflow.
Official downloads and product documentation open on zcode.z.ai.
Callback validation now uses the configured origin, focused tests pass, and the final diff is ready to inspect.
ZCode centers work on a persistent task. The Agent can combine project files, commands, tool results, execution controls, and Git state while a change moves from request to review.
Reference files, folders, past conversations, attachments, and project instructions without leaving the task.
Use terminal commands, browser context, integrations, and execution modes to turn an instruction into checked work.
Keep goals, edits, verification results, and follow-up instructions together across a sustained engineering sequence.
A useful agentic workflow is not just generation. It includes context, controlled execution, verification, and a reviewable handoff.
Read the Agent documentationSelect a project or work outside a project, then give the task a concrete outcome.
Reference files with @, past conversations with #, commands with /, and skills with $ when they help define the work.
Use the execution mode and model settings that match the task’s risk, complexity, and need for confirmation.
Review tool output, file changes, tests, and Git state before accepting or continuing the task.
ZCode exposes several ways to shape what the Agent knows, what it can do, and how the work continues.
The primary task interface for understanding requirements, editing code, running commands, and reviewing changes.
Organizes sustained work around an objective, completion checks, and state recovery.
Bring project references and retained project knowledge into future tasks with less repeated setup.
Extend local workflows with packaged capabilities, reusable instructions, commands, and tool connections.
Connect supported external tools and data sources through Model Context Protocol servers.
Continue checking and steering workspace tasks through supported remote and bot channels.
The official guide carries the current installers and platform-specific steps. A short first task confirms that the app, workspace, and selected model are connected correctly.
Choose the current build for your operating system from the official installation page.
Use the documentation to understand the Agent, execution modes, context controls, plugins, and integrations.
Four details to understand before choosing an installer and opening your first workspace.
Use the official install and Agent guides for the latest platform support, product behavior, and setup steps.