NanmiCoder/dsh-agent-teams
34934Last commit Aug 15, 2026
dsh-agent-teams DSH plugin
dsh-agent-teams is a plugin for DeepSeek Harness that transforms a single session into a captain-led multi-agent team. It provides durable sub-agents, dependency-aware task management, direct messaging, and a live activity Web UI without requiring a separate workflow engine.
How to install the dsh-agent-teams DSH plugin
dsh plugin --profile web add @nanmicoder/dsh-agent-teamsCopying does not run this command. Review the repository and version before installing the dsh-agent-teams DSH plugin.
dsh-agent-teams DSH plugin data source
dsh-agent-teams DSH plugin snapshot date: Aug 16, 2026
discovered
What the dsh-agent-teams DSH plugin can do
- Captain-led delegation: the current session creates the team, assigns roles, and consolidates the final result.
- Durable members: sub-agents are continuable and can be woken for focused follow-up turns.
- Dependency-aware tasks: tasks move through explicit states and cannot be claimed before their dependencies finish.
- Direct messaging: members send durable mailbox messages directly to teammates or the captain.
- Live activity panel: a Web UI showing roles, current work, unread messages, task dependencies, and archived team history.
Where the dsh-agent-teams DSH plugin fits
- Reviewing code commits from multiple perspectives (performance, security, product) in one consolidated report.
- Decomposing a complex goal into dependency-aware tasks and assigning them to specialized sub-agents.
- Coordinating a team of AI agents to work on a multi-step project with role-based communication.
Who the dsh-agent-teams DSH plugin is for
- DeepSeek Harness users who need multi-agent coordination for complex workflows.
- Developers building AI-powered task pipelines that require dependency management and persistent sub-agents.
dsh-agent-teams DSH plugin limitations
- Only one active team at a time with a single captain.
- State is file-backed and not coordinated across concurrent processes editing the same team.
- The activity panel reports persisted state as-is; models may occasionally finish work without performing the expected task-state update.
dsh-agent-teams DSH plugin: from the repository README
Quoted from the NanmiCoder/dsh-agent-teams README, the upstream source of the dsh-agent-teams DSH plugin. Copyright remains with the original authors.
`dsh-agent-teams` turns the current DeepSeek Harness session into a captain that can assemble durable sub-agents, split a goal into dependency-aware tasks, and coordinate work through direct messages. Ask in natural language. The plugin provides the team protocol, nine coordination tools, persistent state, and a live Web UI—without requiring a separate workflow engine. <p align="center"> <img src="./assets/ui.png" width="100%" alt="DeepSeek Harness conversation with the AgentTeams live activity panel, members, tasks, dependencies, and reports"> </p> ## Why AgentTeams? | Capability | What it changes | | --- | --- | | **Captain-led delegation** | The current session creates the team, assigns roles, and consolidates the final result. | | **Durable members** | Members are continuable DSH sub-agents that can be woken for focused follow-up turns. | | **Dependency-aware tasks** | Tasks move through explicit states and cannot be claimed before their dependencies finish. | | **Direct messaging** | Members send durable mailbox messages directly to teammates or the captain—no relay required. | | **Live activity panel** | The Web UI shows roles, current work, unread messages, task depend
Read the full READMERepository license: MIT
dsh-agent-teams DSH plugin questions
How do I install dsh-agent-teams?
You need an existing DeepSeek Harness installation. Then run `dsh plugin --profile web add @nanmicoder/dsh-agent-teams` to install the plugin via npm. Alternatively, you can clone the repository and run `pnpm install && pnpm build && dsh plugin --profile web add .` to build from source.
Can I use multiple LLM providers for different team members?
Yes, the plugin supports heterogeneous teams. When calling `agent_teams_add_member`, you can specify `provider` and `model` fields to assign a different LLM provider per member. By default, the plugin snapshots the provider and model used by the captain's current step and reuses it for all members.
Does the plugin provide a user interface?
Yes, it includes a live activity Web UI that shows the team structure, roles, current work, unread messages, task dependencies, and archived team history. The panel reads from the persisted state under `<workspace>/.agent-teams/` and combines it with live sub-agent activity.
What happens if a team member is idle?
Members act only when they are woken. Durable mailbox messages remain queued while a participant is idle. When the member is woken again, it can process pending messages. The plugin's state is file-backed and persists across sessions.
Can I have multiple teams running at the same time?
According to the plugin boundaries, only one active team is supported at a time with a single captain. The state is not coordinated across concurrent processes editing the same team, so running multiple teams simultaneously is not recommended.