Skip to content
  • Home
  • Plugin manual
  • dockyard-dsh

AITabby/dockyard-dsh

70Last commit Aug 15, 2026

dockyard-dsh DSH plugin

Dockyard DSH integrates official OAuth and CLI sessions from multiple providers (Codex, Antigravity, Grok, Claude, Cursor) into a unified account pool, model catalog, quota view, and native request path. It is a native DSH plugin that does not require a separate proxy gateway or provider-specific branches in the DSH core. The plugin provides a /dockyard command surface for managing accounts, scanning sessions, and selecting provider policies.

developer-toolsremote-execution

How to install the dockyard-dsh DSH plugin

dsh plugin --profile dockyard-dsh add github:AITabby/dockyard-dsh

Copying does not run this command. Review the repository and version before installing the dockyard-dsh DSH plugin.

dockyard-dsh DSH plugin data source

dockyard-dsh DSH plugin snapshot date: Aug 16, 2026

discovered

What the dockyard-dsh DSH plugin can do

  • Manage multiple provider accounts (Codex, Antigravity, Grok, Claude, Cursor) from a single DSH command surface.
  • Start provider-owned OAuth flows and securely import completed sessions into the account pool.
  • Scan existing official login states on the local machine without adding them automatically.
  • Support manual account selection, sticky session, round-robin, and failover account pool policies.
  • Read live provider model catalogs, reasoning tiers, plans, and quota windows.
  • Optional local visual page for diagnostics (not required for normal operation).

Where the dockyard-dsh DSH plugin fits

  • Developers who need to switch between multiple AI provider accounts (e.g., OpenAI, Gemini, Grok, Claude, Cursor) within a single DSH session.
  • Users who want to avoid running separate proxy gateways for each provider and prefer a unified authentication and request flow.
  • Teams that need to manage shared account pools with failover or round-robin policies across different provider APIs.
  • Advanced users who want to inspect live model catalogs and quota windows from multiple providers in one place.

Who the dockyard-dsh DSH plugin is for

  • Developers using DeepSeek Harness who need multi-provider account management.
  • Power users who run multiple AI provider CLI/desktop clients and want to consolidate their sessions.

dockyard-dsh DSH plugin limitations

  • macOS only; Windows is not supported and has not been fully verified in this release.
  • DSH itself is a developer preview; upstream breaking changes may occur.
  • Provider official CLI, client paths, OAuth fields, and quota interfaces may change; Dockyard reports unknown/missing fields rather than inventing them.
  • Claude and Cursor multi-account capability is limited by the official active session API; persistent offline credential storage is not supported.
  • The plugin depends on macOS Keychain and Swift helper for credential storage; non-macOS platforms will fail closed.

dockyard-dsh DSH plugin: from the repository README

Quoted from the AITabby/dockyard-dsh README, the upstream source of the dockyard-dsh DSH plugin. Copyright remains with the original authors.

**A macOS-only account-pool and native provider plugin for [DeepSeek Harness](https://github.com/deepseek-ai/deepseek-harness) (`dsh`).** [中文](#中文) · [English](#english) > **Current status / 当前状态:** Developer preview · **macOS only** · Windows is not supported or verified in this release. ## 中文 ### Dockyard DSH 是什么 Dockyard DSH 把多个官方 OAuth / 官方 CLI 会话接入 DeepSeek Harness，提供一个统一的账号池、模型目录、额度状态和 provider-native 请求入口。它是 DSH 的原生 bundle/plugin，不需要另起一个代理网关，也不把 provider 逻辑塞进 DSH 核心。 当前包含的 provider 模块： - **Codex** — 官方 OAuth 会话和原生 Responses 请求链路。 - **Antigravity** — 官方本机会话、实时模型目录、额度/credits 和原生 Gemini SSE 请求链路。 - **Grok** — 官方 OAuth CLI、实时模型目录和 provider-native streaming 请求。 - **Claude** — `claude` 官方 CLI 的订阅 OAuth 会话与原生请求适配。 - **Cursor** — `cursor-agent` 或 Cursor.app 的官方 active session 与原生请求适配。 如果对应的官方 CLI 或桌面端没有安装、没有登录，Dockyard 会返回明确的 unavailable/degraded 状态；不会用硬编码的账号、模型、版本、套餐或额度伪造可用结果。 ### 主要功能 - 在 DSH 内使用 `/dockyard` 命令管理账号和 provider。 - 启动 provider 自己的官方 OAuth 登录流程，并将完成的会话安全导入账号池。 - 扫描本机已有的官方登录态；扫描和新增账号是两个独立操作。 - 支持手动选择、sticky session、round-robin 和 failover 账号池策略。 - 读取 provider 返回的实时模型目录、推理档位、套餐和额度窗口。 - 所有命令、模型选择和 LLM 生成都读取同一个 Dockyard runtime，不维护第二套账号池或额度缓存。 - 可选的本地可视化页面用于观察状态；

Read the full READMEThe repository declares no license. Check with the authors before using it.

dockyard-dsh DSH plugin questions

Is Dockyard DSH compatible with Windows?

No, the current release is macOS only. Windows is not supported and has not been fully verified. The plugin depends on macOS Keychain, Swift helper, and macOS-specific CLI paths. If you are on Windows, you should wait for a future release that adds Windows backend and E2E validation.

Do I need to install DSH first before using Dockyard DSH?

Yes, Dockyard DSH is a plugin for DeepSeek Harness (dsh). You must install the DSH CLI first by running `npm install --global @deepseek-ai/dsh` and verify that `dsh --version` works. Then you can add the Dockyard DSH plugin via `dsh plugin --profile dockyard-dsh add .` (or `github:AITabby/dockyard-dsh` for a remote install).

How do I switch between different provider accounts with Dockyard DSH?

You can use the `/dockyard` commands inside DSH. For example, `/dockyard login <provider>` starts the OAuth flow, `/dockyard add <provider> <candidateId>` imports the session, and `/dockyard policy <provider> <policy>` sets the account selection policy (manual, sticky_session, round_robin, failover). Then `/dockyard use <provider> <accountId>` switches to a specific account for that provider.

Does Dockyard DSH store my tokens or credentials in plain text?

No. Original OAuth tokens or credentials are never written to Git, account pool snapshots, or page state. At runtime, only opaque credential references are passed. On macOS, the default credential store is the Keychain. Non-macOS platforms will fail closed rather than silently fall back to insecure memory storage.

Can I use Dockyard DSH without the local visual page?

Yes, the local visual page is optional and only for diagnostics. The core functionality (commands, account management, model selection, generation) works entirely through DSH's command interface. You do not need to run the local page for normal operation.