DSH AgentTeams

An open-source AI agent coordination plugin for DeepSeek Harness, with editable team plans, mixed-model teams and local task records.

DSH AgentTeams coordinates AI agents inside DeepSeek Harness for coding, research and review tasks. A lead agent assigns work to specialists and combines their results, while you can review the proposed team before execution.

Editable plans: Review and revise agent roles, model choices and task dependencies before approving the team to start.
Mixed-model teams: Members can use different providers and models available through Harness, or inherit the lead agent's model settings.
Task coordination: Agents pick up work when its prerequisites finish, keep their sessions for follow-up tasks and exchange persistent messages with teammates.
Progress views: The team panel shows task ownership, agent activity and an interactive dependency graph, with access to member conversations and completed team archives.
Optional quality gates: Add verification and review tasks, with repair and repeat review when checks fail.

The plugin supports Harness's desktop, web and CLI interfaces and uses the MIT license. Team records stay in the workspace; model requests use the selected Harness provider. Interrupted work can resume, and recovery after a process restart retries stranded tasks.

One lead agent manages one active team at a time. Separate Harness processes don't coordinate changes to the same team records, and agents can finish work without updating its recorded task status.