Adds a Managed Agents settings tab and MCP tools to dispatch long-running agent tasks to Ark cloud Managed Agents.
Install
# from npm (prebuilt)
dsh plugin --profile web add @volcengine/ark-managed-agents
# from a prebuilt release tarball
dsh plugin --profile web add "https://github.com/volcengine/ark-cli/releases/download/dsh-plugins%2Fv0.1.0/ark-managed-agents.tgz"
# from GitHub (first run asks for allowBuilds approval — follow the hint, retry)
dsh plugin --profile web add github:volcengine/ark-cli#path:/dsh-plugins/ark-managed-agents
Any plugin you install runs third-party code with your own permissions — it can read your files, use your credentials, and reach the network, and tool approvals don’t sandbox it. GitHub-sourced plugins also run build scripts at install time — pnpm blocks those until you allow them, so an install can stop with ERR_PNPM_GIT_DEP_PREPARE_NOT_ALLOWED or ERR_PNPM_IGNORED_BUILDS; dsh prints the exact key to add under allowBuilds in your profile’s pnpm-workspace.yaml, and the install works on the next run. Allowing a build is a trust decision: only install sources you trust, and pin a commit (github:owner/repo#sha).
README
This plugin publishes its README in Chinese only.
A DeepSeek Harness (DSH) plugin that integrates Volcengine Ark Managed Agents (AgentKit/Coze-style cloud agents) into the DSH chat UI. After install and an API key, you can browse your Ark managed agents and dispatch long-running agent tasks directly from DSH via MCP tools.
火山方舟 Managed Agents 插件。将方舟云端托管 Agent(AgentKit/Coze 风格)接入 DSH 对话界面;安装并配置 API Key 后,即可在 DSH 中浏览你的 Managed Agent、通过 MCP 工具派发长任务、取回执行轨迹与产物。
This package ships two runtime pieces:
本插件包含两部分:
- Server — A cordis plugin that launches a local stdio MCP server, exposing Ark Managed Agents operations as MCP tools (
mcp__ark_ma__*). 服务端:cordis 插件,启动本地 stdio MCP 服务,将 Ark Managed Agents 操作暴露为 MCP 工具(mcp__ark_ma__*)。 - Client UI — A web-profile injection that adds a Managed Agents settings section under Settings for API key, base URL and default model configuration. 客户端 UI:web profile 注入,在设置页新增 Managed Agents 区块,用于配置 API Key、Base URL 和默认模型。
Features / 功能
- Browse and search agents in your Ark account without leaving DSH / 在 DSH 内浏览、搜索 Ark 账号下的 Agent
- Dispatch tasks to any managed agent through the standard MCP tool interface / 通过 MCP 工具向任意 Managed Agent 派发任务
- Streaming responses with thought-step and tool-call visibility / 流式返回,展示思考步骤与工具调用
- Long-running task polling with automatic trajectory/result retrieval / 长任务自动轮询,取回完整轨迹与结果
- List and download agent artifacts / 列出并下载 Agent 产物
- Per-agent model and parameter overrides preserved from the Ark console / 保留 Ark 控制台配置的模型与参数覆盖
Install / 安装
From npm (recommended / 推荐)
npx -y @deepseek-ai/dsh plugin --profile web add @volcengine/ark-managed-agents
From GitHub Release
npx -y @deepseek-ai/dsh plugin --profile web add https://github.com/volcengine/ark-cli/releases/download/dsh-plugins%2Fv0.1.0/ark-managed-agents.tgz
Via awesome-dsh-plugin registry
npx -y @deepseek-ai/dsh plugin --profile web add volcengine/ark-cli#ark-managed-agents
From a local clone (development / 开发用)
npx -y @deepseek-ai/dsh plugin --profile web add ./dsh-plugins/ark-managed-agents
Restart DSH after install. The MCP server row carries a config: entry and only activates on the next host start. After restart, open Settings → Managed Agents to configure credentials.
安装后必须重启 DSH(MCP 服务在 config: 行注册,需重启生效)。重启后打开 Settings → Managed Agents 配置凭据。
Configuration / 配置
Configure in the UI settings panel (preferred), or via environment variables (used as fallback):
在 UI 设置面板中配置(推荐),或通过环境变量配置(作为兜底):
| Setting / 设置项 | Env var / 环境变量 | Purpose / 说明 |
|---|---|---|
| API Key | ARK_MA_API_KEY / ARK_API_KEY |
Volcengine Ark API key. ARK_API_KEY is used as a shared fallback. / 方舟 API Key;ARK_API_KEY 作为通用兜底。 |
| Base URL | ARK_MA_BASE_URL |
API base URL. Default: https://ark.cn-beijing.volces.com/api/agentkit. / API 基础地址,默认指向 cn-beijing AgentKit。 |
| Default Model | ARK_MA_MODEL |
Model ID used when an agent does not specify one. / Agent 未指定模型时使用的默认模型 ID。 |
Exposed MCP Tools / 暴露的 MCP 工具
After restart, the following tools (prefixed mcp__ark_ma__) are available to the model:
重启后模型可使用以下工具(前缀 mcp__ark_ma__):
| Tool / 工具 | Purpose / 用途 |
|---|---|
ma_list_agents |
List agents in the configured Ark account / 列出账号下的 Agent |
ma_dispatch_task |
Dispatch a task to a specified agent / 向指定 Agent 派发任务 |
ma_get_status |
Poll a running task's status / 查询任务运行状态 |
ma_get_trajectory |
Retrieve a task's execution trajectory (thoughts, tool calls) / 获取任务执行轨迹(思考、工具调用) |
ma_get_result |
Retrieve a completed task's final result / 获取已完成任务的最终结果 |
ma_list_artifacts |
List artifacts produced by a task / 列出任务产生的产物 |
ma_download_artifacts |
Download task artifacts to local disk / 将任务产物下载到本地 |
toolCallTimeoutMs is set to 5 minutes (300000 ms) to accommodate long-running agent tasks.
工具调用超时设为 5 分钟(300000 ms),以适配长时间运行的 Agent 任务。
Requirements / 前置条件
- DSH ≥ 0.1.0
- Node.js ≥ 22.19.0 (ships with DSH)
- A Volcengine Ark account with Managed Agents (AgentKit) enabled
- An API key with access to the AgentKit API
Notes / 说明
- The MCP server runs as a local Node.js subprocess using
process.execPathwithELECTRON_RUN_AS_NODE=1, so it works under both thedshCLI and the DSH desktop app. MCP 服务以本地 Node.js 子进程运行(使用process.execPath+ELECTRON_RUN_AS_NODE=1),同时兼容dshCLI 与桌面版 DSH。 - Server path is resolved from the DSH profile directory at runtime — no hard-coded absolute paths. 服务路径在运行时从 DSH profile 目录解析,不依赖硬编码绝对路径。
failOnStartupError: falseis set so a misconfigured API key does not prevent DSH from starting; configure the key and restart when ready. 设置了failOnStartupError: false,API Key 配置错误不会阻止 DSH 启动;填好后重启即可。
License
Apache-2.0 © 2025-2026 Beijing Volcano Engine Technology Co., Ltd.
Links
More in this category
Tencent/WeKnora#dsh-weknora★ 21308
Four read-only tools over a WeKnora knowledge base: list knowledge bases, hybrid passage search, reassemble one document's chunks in order, and WeKnora's own cited RAG or ReAct-agent answer with a resumable session id.
superdesigndev/treg★ 1158
Tool catalog for agents: search ~2,600 external endpoints (SEO and SERP, backlinks, social, people and company enrichment, ad libraries, scraping) by the task you want done, read each one's parameters and per-call price, then call it with the credential injected server-side. Ships the skill plus an MCP row that stays disabled until TREG_TOKEN is set.
anysearch-team/anysearch-dsh★ 399
AnySearch-powered real-time web and vertical search provider for DeepSeek Harness.
EthanYoQ/Invoice-Downloader#dsh-invoice-downloader★ 375
Local IMAP invoice download, OCR, archive, and Excel reimbursement summaries for DeepSeek Harness.
omdsh-dev/dsh-data-agent★ 184
Let the AI connect to databases and write SQL for you.
zhaoolee/notes#dsh-plugin★ 158
Export DSH conversations as Smartisan Notes-style PNGs, or create and update Markdown notes in a configured account-scoped workspace.
Community comments
Comments are public GitHub Discussions. Loading them connects to GitHub and Giscus; a GitHub account is required to post.