DeepSeek open-sources Super-Agent, a research prototype for team coding work

DeepSeek, the Chinese AI lab best known for its open-weight language models, has released the code of an experimental agent meant to handle software work on behalf of a whole team. The project, called dsh-superagent and nicknamed Super-Agent, is now public on GitHub under the permissive MIT licence, and The Next Web reported on it on November 1.

How the agent works

According to the repository's README, the agent is tied to a single GitHub repository and is packaged as a bundle for DeepSeek Harness. Requests can arrive through GitHub and, optionally, through Feishu, ByteDance's workplace messaging app. When a person on an allow-list asks for something, the agent opens a work item, handles it in its own session and working directory, and answers in the original conversation, for example with a pull request. A web panel lists only the items that need a human decision.

A separate design document sets out the guiding idea: the agent should cost people as little attention as possible. It acts only when asked or when a rule written by the team permits it, and such proactive work starts in a mode that only records what the agent would have done. It confirms a request once, stays silent while working and writes again only when a result is ready, a decision is needed or it is blocked. The document also tells operators that they remain responsible for what the agent posts under the team's accounts and advises against using it for bulk pull requests.

Positioning and caveats

The Next Web, citing Bloomberg, reports that DeepSeek is targeting Anthropic's Claude Code with this work, and that the team behind its agent tools was formed earlier this year to compete with US coding products. DeepSeek itself makes no such comparison in the repository.

The company is explicit about the project's status. The README describes it as a research prototype for reference only, "not an official DeepSeek product", and warns that the code is not production-ready, may contain bugs and differs from the version used internally. Some ideas in DeepSeek's accompanying article are described as future directions rather than features that exist today. Setup involves many steps, and The Next Web notes that DeepSeek tells users to review and adapt the code before running it.

No benchmark or independent evaluation of the agent's output has been published so far, so how well it performs on real team workloads remains unverified.