DeepSeek has open-sourced dsh-superagent, an MIT-licensed research prototype that turns a GitHub repository into a self-running team of coding agents. The TypeScript bundle (v0.1.0) listens to GitHub — and optionally Feishu — and converts allowed requests into parallel work items, each running in its own agent session.
The design targets a real bottleneck: human attention. DeepSeek’s team argues that mainstream agents treat one session as the basic unit, but a person can actively follow only 3–5 sessions at once — so attention, not model capability, becomes the limit on multitasking. (Image Credits: MarkTechPost)
How dsh-superagent’s Router-Worker Design Works
The system has two layers. A long-running Router handles interaction and dispatch, while isolated Worker sessions do the actual work. The flow: a GitHub or Feishu event becomes a Trigger, an Eval decides whether it asks for new work, and authorized work becomes a Ticket with an immutable Mandate stating why it may proceed. Results return to the original conversation, often as a pull request, and a local web panel lists only what needs a person.
The core rule from the README: take as little human attention as possible, give as much value as possible. The agent confirms once, stays quiet while working, and writes again only with a result, a question or a failure.
dsh-superagent Benchmarks: 1.9x Faster, With Caveats
In DeepSeek’s own simulation, dsh-superagent finished tasks far faster under a noisy message stream than a single-session agent: p95 task completion of 15.7 minutes versus 29.5 minutes for the baseline — about 1.9x faster.
The honest caveats: it was tested only on synthetic messages and simulated tools, and real coding quality was not measured. It also runs on your own machine as a dedicated DeepSeek Harness profile, requiring Node.js 22.19+ or 24+, git 2.32+, Python 3.8+, tmux and GitHub CLI 2.48+.
The release shipped alongside a design post, “Exploring Super-agent Design: Towards a More Autonomous Agent System.”
Also Read: You.com Launches Inference Router: 12 Open-Weight Models With 99.9% Uptime SLA
Source: MarkTechPost — read the full story there.
