MIT 协议的 DeepSeek Harness 插件：运行一支由独立 root agent 组成的团队，每个成员拥有自己的模型、上下文与工具，共享同一个 Workspace，由 Leader 指派任务并验证结果。
DSH 适配
原生运行时
作者声明
安全审计
未审计
最后核验
2026-09-06
许可证
MIT
01它能帮你完成什么？
Assemble a multi-agent team where each member is an independent root agent with its own model, provider, context, and tools, all sharing one Workspace.
A team of isolated agent sessions that plan, implement, test, review, and document on the same shared Workspace files.
DeepSeek Harness users who want to split complex work across specialized agents instead of overloading a single agent.
Create reusable assistants for planning, coding, testing, review, or documentation roles.
Configured assistants (provider, model, preset, default permission, reasoning mode, Skills, MCP Servers, role instructions) reusable across teams.
Teams building a library of role-based assistants under Settings → Agent Team.
Coordinate work through a Leader that creates and assigns tasks, tracks progress, and verifies results.
An explicit task/result/status flow where the Leader assigns work and members report progress, without sharing one expanding conversation.
Users who want a coordinating Leader and isolated member contexts rather than a single shared conversation.
02如何接入 DeepSeek Harness？
前置条件
- Node.js 22.19.0+ or 24.0.0+
- DeepSeek Harness 0.1.1-rc.2
- pnpm available on PATH (Harness uses it to manage Profile plugins)
安装步骤
- 01
$ npx @deepseek-ai/dsh plugin --profile web add @limuyang2/dsh-agent-team
- 02
$ npx @deepseek-ai/dsh web
- 03
Open the URL printed by Harness (normally http://127.0.0.1:3080/).
验证接入成功
- Open the URL printed by Harness, normally <http://127.0.0.1:3080/>.
回滚
- Stop Harness with Ctrl+C, then run: npx @deepseek-ai/dsh plugin --profile web remove @limuyang2/dsh-agent-team
03DSH 适配与能力边界
Installed as a DeepSeek Harness Profile plugin via the official dsh plugin CLI; each member runs as an independent Harness session sharing one Workspace.
Independent root agents
A role description, or a manually configured assistant→An independent root agent with its own model, session, context, permissions, reasoning mode, and tool activity
Leader task orchestration
A team goal sent to the Leader→Tasks created and assigned to members; progress and results collected and verified
Shared Workspace and Git preview
A team Workspace (Git repository or normal folder)→Shared files accessible to all members; Git changes and diffs previewable
Reusable assistant library
Settings → Agent Team assistant definitions→Reusable assistants with provider, model, preset, default permission, reasoning mode, Skills, MCP Servers, and role instructions
Agent Team reads the model catalog from the active Profile and never stores provider API keys
04适合谁？何时不该用？
适合
- DeepSeek Harness users who want to split complex work across specialized agents instead of overloading a single agent.
- Teams building a library of role-based assistants under Settings → Agent Team.
- Users who want a coordinating Leader and isolated member contexts rather than a single shared conversation.
不适合
- Requires DeepSeek Harness 0.1.1-rc.2 or later; the plugin is a Harness Profile plugin and depends on the Harness runtime.
05兼容性、维护与安全提示
- Requires DeepSeek Harness 0.1.1-rc.2 or later; the plugin is a Harness Profile plugin and depends on the Harness runtime.
- Requires Node.js 22.19.0+ or 24.0.0+ to run the Harness.
- Requires pnpm available on PATH; Harness uses it to manage Profile plugins.
- An assistant's permission setting is only the member's initial default; it can be changed at runtime per session.
- The Changes view requires a Git Workspace; normal folders still support file browsing but will not show Git changes.
MIT · current release 0.1.4 per README; repo latest_release tag absent
06常见问题
如何把 Agent Team 安装到 DeepSeek Harness？
运行 `npx @deepseek-ai/dsh plugin --profile web add @limuyang2/dsh-agent-team`，再用 `npx @deepseek-ai/dsh web` 启动 Harness，打开它打印的地址（通常是 http://127.0.0.1:3080/）。安装或替换插件后请重启 Harness。
Agent Team 会把成员变成某个父智能体的子智能体吗？
不会。Agent Team 不会把成员变成子智能体——每个成员都是独立的 root agent，拥有自己的模型、会话、上下文、权限、推理模式与工具活动。
成员可以使用不同的模型与供应商吗？
可以。你可以在一个团队里混用不同的供应商与模型，例如用 Codex 当 Leader、GLM 当编码成员，并只给每个角色它所需的 Skills 与 MCP Servers。
团队成员会共享对话历史吗？
成员共享同一个 Workspace，但不共享对话历史。这样既能让他们操作同一批文件，又能让各自角色与模型上下文保持隔离。
运行 Agent Team 有什么前置要求？
需要 Node.js 22.19.0+ 或 24.0.0+、DeepSeek Harness 0.1.1-rc.2，以及 PATH 上可用的 pnpm（Harness 用它来管理 Profile 插件）。
07相关的 DSH 工作流
archify
作者 tt-a1i
为编码智能体生成美观可验证的架构图、时序图与数据流图，输出自包含 HTML，支持动效与清晰导出。
openviking
作者 volcengine
为 AI 智能体打造的自进化上下文数据库，统一智能体记忆、知识 RAG 与技能。
nocobase
作者 nocobase
开源的 AI + 无代码应用搭建平台：AI 在久经生产验证的底层基础设施与所见即所得的无代码界面之上工作，帮你快速构建 CRM、ERP 等业务系统，兼顾速度与可靠性。
learn-harness-engineering
作者 walkinglabs
Harness 工程新手教程，从 0 到 1 系统学习智能体工作流框架。
08数据与来源
Build teams of independent AI agents inside DeepSeek Harness. Mix models and providers, assign one Leader, and let every…
Agent Team does **not** turn members into subagents. Every member is an independent root agent with its own model, sessi…
页面基于项目公开文档、仓库元数据和 DSH Plugins 的结构化解析生成；最后核验于 2026-09-06。发现错误？提交更正。
最佳 DeepSeek Harness 插件
从全目录挑出的 12 个值得优先安装的插件，覆盖各个分类。