DeepSeek Open-Sources dsh-superagent, a Prototype Coding Agent Built to Spend as Little Human Attention as Possible

Most coding-agent launches sell capacity: more tasks, more parallel sessions. DeepSeek's new open-source prototype is pitched around a scarcer resource, the attention of the people supervising it. Its stated rule is to "take as little human attention as possible, and give as much value as possible," and much of the design follows from that: the agent acts only when asked or when a team's own rule allows it.

What happened#

DeepSeek published dsh-superagent on GitHub under the MIT licence; GitHub's records show the repository was created on October 31. The README describes it as "a proactive coding agent for one GitHub repository," packaged as a bundle for DeepSeek Harness. It listens to GitHub and, optionally, Feishu, ByteDance's workplace chat app. When an allowed person asks it for work, it opens a work item, handles it in its own agent session and directory, and answers in the original conversation, "for example with a pull request." A web panel shows what needs a person.

The README warns that "an agent that can act for a whole team can also waste a whole team's attention," so the agent speaks only when someone must act or can use a result. DeepSeek's agent tools team set out the design in an article published on Saturday, TNW and the South China Morning Post reported.

DeepSeek is aiming the agent at Anthropic's Claude Code, Bloomberg's Debby Wu reported, according to TNW. TNW also reported that DeepSeek released a preview of its own agent app in August.

The release lands as DeepSeek faces pressure in the US, where civilian federal agencies were ordered to remove its products last month, and as the company is reported to be closing a large funding round.

Limits#

• DeepSeek labels the code "Research prototype for reference only, not an official DeepSeek product," says much of its engineering was done by coding agents, and tells users to review and adapt it before running it.
• The README says the public code differs from DeepSeek's internal version and from its design article, and that some designs in the article are not implemented.
• DeepSeek publishes no benchmark or success-rate figures for the agent, and we have not run it.
• The Claude Code framing comes from Bloomberg's reporting, relayed by TNW; we did not read the Bloomberg article.

Sources#

• GitHub: deepseek-ai/dsh-superagent-prototype README (repository created October 31, 2026)
• TNW: DeepSeek open-sources Super-Agent, a coding agent that reads team chats (November 1, 2026)
• South China Morning Post: DeepSeek unveils AI 'super agent' in coding race with OpenAI, Anthropic (November 1, 2026)