HarnessMatch
Menu

Coding agent

ZCode

Desktop agent workspace with durable goals, parallel subagents, and scheduled tasks.

Status
Active
License
Proprietary
Evidence
Documented, 17 sources
Product record checked
2026-07-27
Compare harnessesInspect sources

At a glance

A proprietary desktop agent environment with persistent goal state, independent completion checks, parallel and background subagents, scheduled tasks, scoped approvals, MCP, hooks, remote workspaces, and an integrated browser and review flow.

Good choice if

  • Developers who prefer a visual agent workspace over a terminal-only loop
  • Long-running goals that need explicit lifecycle state, repeated verification, and recovery
  • Scheduled or parallel desktop work combining subagents, MCP, hooks, browser context, and remote environments

Check before choosing

  • No documented headless CLI, CI entry point, or public API was found; scheduled automation still belongs to the desktop product
  • Docker isolation requires an already-running user-provided container, while local and SSH workspaces otherwise execute with the target account's privileges
  • Undo is documented for surfaced file changes, not arbitrary command, network, database, or system side effects
See 2 more considerations
  • Goal verification and trajectory auditing are vendor-described mechanisms, not independent proof of task success
  • Custom self-hosted compatible endpoints are supported, but the reviewed docs do not establish a packaged local-model runtime

Why it qualifies as a coding harness

This confirms category fit, not product quality. Every required criterion links back to first-party evidence.

Qualifies4 of 4 required criteria evidenced

  • Adaptive agent loop

    Documented

    The system repeatedly observes results and chooses the next action instead of following a fixed one-pass graph.

    EvidenceZCode Agent workflow

  • Repository tool execution

    Documented

    The system can use tools to inspect and change a repository or its execution environment.

    EvidenceADE tools

  • Task-aware context management

    Documented

    The runtime assembles, updates, compacts, retrieves, or persists task-relevant context while work proceeds.

    EvidenceTask and file management

  • Model-independent runtime control

    Documented

    Permissions, budgets, interruption, policy, or stop controls operate outside the model's own text generation.

    EvidenceSafety confirmation

Membership establishes category fit only. It does not score quality, safety, autonomy, model capability, or benchmark performance. · Read the membership rule.

How it works under the hood

Seven mechanisms mapped from first-party records. These labels describe what the harness provides, not how intelligent its model is.

7/7layers documented
  • Execution & isolationManaged isolationDocumented mechanism, not a performance score.
  • Tooling & integrationsExtensible + browserDocumented mechanism, not a performance score.
  • Context & stateManaged contextDocumented mechanism, not a performance score.
  • Lifecycle & recoveryManaged recoveryDocumented mechanism, not a performance score.
  • ObservabilityStructured tracesDocumented mechanism, not a performance score.
  • VerificationWorkflow-gatedDocumented mechanism, not a performance score.
  • Governance & permissionsPolicy controlsDocumented mechanism, not a performance score.

Public code audit

No official public repository was located for a code-level audit.

Measured configurations

No benchmark run passes the full metadata admission policy for this harness yet. Missing data is not scored as zero.

Benchmark policy and all runs

Capability support

Documented first-class product support, checked against the sources below.

External tools (MCP)
DocumentedProduct-supported MCP integrationSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Local models
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Agent parallelism
DocumentedDelegated or parallel agent workflowSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Runs without an open UI
Not documentedNo first-class support established by the current recordAbsence of current documentation is not proof that the capability is impossible.
Browser control
DocumentedBuilt-in or product-supported browser controlSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Isolated execution
DocumentedDocumented execution-isolation mechanismSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.
Undo file changes
DocumentedProduct-supported file or session rollbackSource · checked 2026-07-27The source establishes the mechanism, not its quality or availability in every mode.

Primary evidence

Each capability claim is tied to a first-party record and a verification date.

Product record checked 2026-07-27
ZCode Agent workflowDesktop-native task workflow, file and browser context, execution modes, Git state, model switching, and review flow without importing GLM model claimsOfficial docsGoal ModeSession-level goal state, resource budgets, automatic per-round verification, failed-closed completion, pause and resume, recovery, and trajectory recordsOfficial docsSubagentsBuilt-in and custom subagents, independent contexts, tool scoping, foreground parallel execution, and current Beta limitsOfficial docsMCP serversMCP service connection, tool availability, configuration, and workspace useOfficial docsSafety confirmationFive permission modes, exact action prompts, scoped allow and reject policy, risk visibility, review flow, and per-file open and undo actionsOfficial docsADE toolsTask workspace, existing Docker and SSH targets, built-in browser preview and element context, terminal, DevTools, and review integrationOfficial docsRemote developmentExecution placement in SSH hosts and existing local Docker containers, connection logs, target-account boundaries, and reconnection behaviorOfficial docsHooksJSON subprocess protocol, session and tool events, pre-tool policy decisions, post-tool diagnostics, stop checks, scopes, limits, and local-code warningOfficial docs
View 9 additional sources
Additional first-party evidence9 sources
Task and file managementTask status and grouping, repository file search, Git change filtering, review context, and generated repository orientationOfficial docsPlugin systemBeta marketplace packages spanning skills, commands, subagents, MCP, hooks, and language servers plus enable and configuration behaviorOfficial docsAgent SkillsSKILL.md format, user and project scopes, explicit invocation, enablement, generation, and imports from other coding agentsOfficial docsModel connectionsZ.ai subscriptions, first-party and OpenRouter API routes, Anthropic and OpenAI compatible custom endpoints, and self-hosted service supportOfficial docsUsage statisticsDevice-local session, token, message, activity, and model-use records plus remote Z.ai plan and tool-call quota reportingOfficial docsRemote ControlMobile access to a desktop-open workspace, task continuation, connection dependency, and remote-control limitsOfficial docsBot ChannelWeChat and Feishu task entry, workspace selection, task continuation, and desktop service dependencyOfficial docsDesktop installationVersion 3.5.2 desktop distribution for macOS and Windows, Linux beta status, onboarding, and workspace startupOfficial docsRelease 3.5.2 and recent automation changesVersion and release date, scheduled tasks, background subagents and shell tasks, built-in web integration, and session recovery fixesOfficial docs

Ecosystem discovery

OpenRouter apps Discovery signal only; usage rank is not used as a quality or capability score. Observed 2026-07-27.