DeepSeek released an open-source prototype of a more autonomous coding agent named Super-agent, The Next Web reports based on Bloomberg coverage. The launch aims to challenge Anthropic's Claude Code by taking up as little human attention as possible while handling development tasks.
Key takeaways
- DeepSeek released its Super-agent prototype as open-source code under the MIT licence.
- The tool can monitor group chats, spot bugs, and work on fixes in the background.
- DeepSeek built the agent on top of its open-source DeepSeek Harness software.
- The project is currently a research prototype for reference only and not an official product.
What is the Super-agent prototype?
The Chinese AI company's Harness team detailed the new tool in a WeChat post on October 31 and published the code on GitHub, as reported by The Next Web. According to the post, Super-agent can watch developers' group chats and spot potential bugs even when nobody tags it. It then attempts to reproduce the problem and work on a fix in the background, a detail noted by Chinese outlet Yicai.
Super-agent runs on DeepSeek Harness, which is the company's open-source software designed to wrap its AI models so agents can tackle longer and more complex tasks. The company already runs the tool internally to fix bugs, respond to code reviews, and improve its own tools. The software works with GitHub and offers optional support for Feishu, the workplace messaging app owned by ByteDance.
How does the DeepSeek coding agent operate?
Documentation shows that the agent only starts working when an approved person asks it to or when a rule set by the team permits it. The GitHub page warns that it remains a research prototype for reference only rather than an official DeepSeek product, advising developers to review and adapt the code before use.
The Harness team was formed earlier this year to build agents that can compete with coding tools from US laboratories. Bloomberg reported that this release follows a $15bn funding round in October backed by Tencent and battery maker CATL, ahead of a potential stock market listing, though DeepSeek has not confirmed that round.
What it means for developers
Developers now have access to the code under the MIT licence, allowing anyone to use and alter it. While it is not a finished product, teams interested in autonomous coding workflows can test how the agent monitors chats and handles background fixes. Users must still ensure they set strict rules for when the agent is allowed to act, keeping human oversight in place for any code modifications.
Frequently asked questions
Super-agent is an open-source prototype of a more autonomous coding agent released by DeepSeek to compete with Anthropic's Claude Code.
The code was published on GitHub by DeepSeek's Harness team under the MIT licence, which lets anyone use and change it.
The agent can watch developers' group chats, spot potential bugs without being tagged, reproduce the problem, and work on a fix in the background.
No, the project's GitHub page describes it as a research prototype for reference only and advises developers to review and adapt the code.
Super-agent runs on DeepSeek Harness, which is the company's open-source software that wraps AI models to handle longer and more complex tasks.
Based on reporting by The Next Web. Read the original article
